Introduction

Years from now, how will the 7th Conference on Retroviruses and Opportunistic Infections be remembered?

For many who winged westward from other parts of the USA for the January 30-February 2 San Francisco meeting, the most cherished memory may be the escape from cruel winter. Native San Franciscans griped about damp dawns and drizzly afternoons. But to East Coasters who managed to wrangle a plane seat between the midweek blizzard and the weekend ice storm, this was not winter rain. This was the dewy, nymphen kiss of spring.

How can one man's winter wasteland be another's vernal Paradiso? Perspective, of course, explains everything. Comfort is a fluid notion, beauty mercurial, truth mutable. The eye of the beholder, and the date on his calendar, rule all.

If attendees dodging raindrops and vagabonds outside the San Francisco Marriott had little time to ponder Protean caprice, once entombed in the hotel's subterranean ballroom vaults, they could ponder little else. The forge of HIV research and, more, the candescence of clinical verity have lapped at the edge of yesteryear's crystalline theory and ice-clean assurances. Nimbly trading PowerPoint-programmed laptops, dissecting projected data with laser pointers, speakers at this Retro conference worked hard to construe the ambit of this icemelt puddle and to predict the direction of its spread.

These are molten times in HIV medicine and, to borrow the title of a Philip Glass song cycle, Retro 7 audiences heard many a song from liquid days.

The riptide rush found perhaps its clearest voice in William Powderly, MD (Washington University, St. Louis). Summing up a symposium on metabolic complications of antiretrovirals, this prudent and thoughtful investigator asked the question that, until then, seemed the sole estate of the most temperate British opinion leaders: Should an asymptomatic person with a high CD4⁺ count be treated immediately with combination therapy, or should that person wait until a time when clinical benefit outweighs the risk of side effects?

Then Powderly suggested when that "time" may be--when the CD4⁺ count sinks to 200 cells/mm³. The goal of therapy, he maintained, is "to let your patient live longer and prosper," not to suppress plasma viremia to undetectable levels. As American electoral politics ground clangorously from first to second gear, HIV clinicians wondered whether they should adopt the label of one noteworthy presidential aspirant and become compassionate conservatives, at least when it comes to timing antiretroviral intervention.

Was this apostasy, or a signal that Fortress Early Start can no longer withstand the torch of long-term failings by imperfect drugs? Certainly the Fortress did not lack for defenders at this conference. More than one study buttressed the hypothesis that starting early offers the best hope of shoring up that critical castle keep--an immune armory with weapons that still work.

Perhaps, then, the answer lies in STIs, those so-called strategic treatment interruptions that relieve long-suffering patients from the harrowing adversities called "events" or--less cheerfully--drug side effects. Indeed, juicy scraps of data actually hint that planned treatment holidays can be good for you, because they rouse sleepy sentries of immunity into action. But other research posed the possibility that the sandman soon reclaims these somn olent soldiers, or that anything less than complete and sustained viral control thwarts the chance of reconstitution.

Fluid also were findings on the benefits, and risks, of switching from one treatment to another. Every regimen has its drawbacks, and a combination considered benign to lipids, for example, may also by considered benign by HIV. Meanwhile, a few studies smudged the heretofore polished lipid profiles of nonnucleoside reverse transcriptase inhibitors (NNRTIs). And one big cohort study demonstrated the difficulty of tweezing apart the much-twined interplay of drug effects, lifestyle choices, and life itself.

And then there's that damnably persistent suggestion that you can keep your class-sparing strategies, and your STIs, because that much-touted, now-tarnished highly active antiretroviral therapy, HAART, does just great, if only people take the pills. One need not bend the truth to argue that the most important study of the meeting languished in an overheated poster hall where water tanks emptied fast as sweaty attendees pondered fluent findings on 84 Floridians.

Return to Article Contents

The Deadly Price of Poor Adherence

Among 42 treatment-naive individuals beginning a three- or four-drug regimen including a protease inhibitor (PI), an NNRTI, or abacavir, all 42 had a viral load below 400 copies/mL after 48 weeks in an intent-to-treat analysis [abstract 71]^*. Among another 42 people new to antiretrovirals, 68 percent had a sub-400 viral load 48 weeks after starting a three- or four-drug combination (P < 0.01 compared with the first group).

Why did the first group chalk up a letter-perfect score while the second group suffered the usual number of drug failures? It wasn't because group 1 was healthier when treatment began, reported Margaret Fischl, MD (University of Miami). On the contrary, the first group had a lower average starting CD4⁺ count (261 cells/mm³ versus 375 cells/mm³ in the second group) and a higher baseline viral load (4.0 logs versus 3.2 logs). And group 1 included more men (93 percent versus 79 percent), more African Americans (86 percent versus 17 percent), and more illegal drug users (31 percent versus 2 percent).

Everyone in group 1 made it to week 48 with fewer than 400 HIV RNA copies/mL because they all took their drugs in prison as directly observed therapy (DOT). No group 2 members were behind bars; all took their own medications in trials run by the Miami AIDS Clinical Research Unit.

The prisoners also beat the standard-of-care group in reaching and maintaining viral loads below 50 copies/mL at 48 weeks (85 percent versus 50 percent, P < 0.01), in CD4⁺-cell gains (+183 cells/mm³ versus +136 cells/mm³, P < 0.02), and in skirting grade 3 or 4 toxicities (15 percent versus 35 percent, P < 0.01). All of these results also came from intent-to-treat analyses.

One need not sharpen the scalpel to dissect conclusions from this inspired analysis. As Fischl's poster noted, her study "points out the substantial negative impact of noncompliance with current combination regimens." Even though the poster called group 2 the standard-of-care group, they weren't really. All were enrolled in clinical trials and so enjoyed the well-known blandishments that accompany care in such studies. Yet the DOT prisoners handily bettered their "standard-of-care" counterparts on every measure, and usually significantly so.

Fischl's study amply demonstrates that familiar regimens clinicians can construct today control replication very well indeed for at least 11 months, when people take the drugs without fail. "We have in our hands already potent regimens," Fischl told the Journal. They're not ideal, because they don't suppress viremia below 50 copies/mL in every last person who sticks with the regimen, because they have side effects, and because the country is probably not ready to lock up everyone with HIV infection and give them DOT.

But, oh, what a world of difference lies between that 85 percent with viral loads below 50 copies/mL after 48 weeks in Fischl's intent-to-treat analysis, and the 46 percent below that mark after 48 weeks in an intent-to-treat meta-analysis of 22 HAART trials by John A. Barlett, MD (Duke University, Durham) [abstract 519]. Bartlett understandably concluded that his study "underscores the importance of further development of more active and durable antiretrovirals for the treatment of HIV infection." No one would quibble with that opinion. Although Fischl concurred that today's treatments leave room for improvement, the improvement she stressed to the Journal was not activity or durability, but once-daily dosing.

It may be tempting to dismiss the importance of Fischl's findings because, even with potent once-a-day combos, DOT is possible only for prisoners and other special populations. But some clinicians have already reported that nearly faultless adherence with more complicated combinations doesn't depend on the ball-and-chain approach. In an illuminating report at last fall's Lisbon meeting, Sydney clinician Harry Michelmore, MB BS, kept viral loads below 50 copies/mL for 20 to 124 weeks in 29 of 30 individuals treated with an induction-maintenance strategy.¹ He calculated a 98 percent adherence rate in this group, and that result was no accident. Michelmore has crafted an effective plan to coax nearly faultless adherence from the people in his care.¹

At the Conference on Retroviruses, Steven Safren, PhD (Massachusetts General Hospital and Fenway Community Health, Boston) showed that even people who acknowledge poor adherence can improve their pill-taking habits quickly through a program called Life-Steps [abstract 74]. The 11 steps focus on nuts-and-bolts issues like getting to appointments, communicating with health care providers, coping with side effects, scheduling medications with meals and daily activities, and storing medications.²

The Mass General-Fenway team compared Life-Steps with medication monitoring via daily pill diaries in 83 persons. For moderately good adhererers, both techniques worked well. Fourteen people randomized to medication monitoring improved their adherence from 92 percent at baseline to 97 percent at week 2. And 13 people followed through week 12 maintained that 97 percent adherence rate. Thirteen people randomized to Life-Steps improved from 80 percent baseline adherence to 97 percent at week 2, a rate sustained at 96 percent among 11 people followed through 12 weeks (P < 0.05 for both week-2 measures).

But among people who came into the study confessing bad adherence habits, the more rigorous Life-Steps approach worked better. In this comparison, 26 people assigned to medication monitoring improved from an 84 percent baseline adherence rate to 90 percent at week 2 and to 93 percent at week 12 (n = 25). Thirty people who began the Life-Steps program improved from 74 percent adherence at baseline to 95 percent at week 2 (P < 0.01) and 94 percent at week 12 (n = 28). "Even minimal interventions (monitoring adherence) can be useful for helping persons improve adherence," Safren concluded. "However, those with already extant adherence problems may require stronger interventions than medication monitoring alone."

No one needs convincing that adherence has a measurable impact on viral load. Several studies prove that. Now research by Robert Hogg, MD, and colleagues in Vancouver also shows that death shadows bad adherence [abstract 73]. The British Columbia team sifted through records of 950 persons who began triple therapy with a PI or NNRTI between August 1996 and December 1998. Then they used a multivariate Cox proportional hazards analysis to pin down factors associated with the 64 deaths recorded by June 30, 1999.

Neither age, gender, baseline AIDS diagnosis, baseline viral load, intravenous drug use, nor physician experience predicted death. But, after adjusting for those factors, Hogg and colleagues did link two other variables to mortality: A lower baseline CD4⁺ count raised the risk of death by 35 percent for every 100-cell decrement (P < 0.001). And poor adherence, measured by drug dispensing records, upped the odds of death by 16 percent for every 10 percent fall in adherence (P < 0.001). Restricting the analysis to AIDS-free individuals did not change the results. It is difficult to imagine a more potent incentive to stick with an antiretroviral schedule.

Return to Article Contents

When Events Turn Adverse

Among motivated seropositive people with reasonably orderly lifestyles, the greatest threat to A+ adherence comes from the drugs themselves. Perhaps converts to medieval mortification of the flesh can contemplate a lifelong challenge of leaky bowels or melt-away facial fat. Most can't.

And the list of drug side effects (sometimes delicately dubbed "adverse events") just keeps growing. Besides devoting scores of posters and two slide sessions to metabolic and fat distribution abnormalities, this conference linked another health threat to antiretrovirals: osteopenia (see "Weaker Bones: Another HAART Side Effect.")

Return to Article Contents

Dystrophic lipids, dispersed fat

Meeting attendees anxious to learn the latest on metabolic and body shape changes brought on by antiretrovirals had 48 posters and six slide presentations of new data to absorb. No wonder they packed the Yerba Buena Ballroom when Kathleen Mulligan, PhD (University of California, San Francisco) and David Cooper, MD, DSc (University of New South Wales) bravely tried to summarize what's known so far about body composition and lipid changes (Mulligan) and about lactic acidosis and hepatic dysfunction (Cooper).

Mulligan's message was that much about lipodystrophy remains amorphous [abstract S20]. The "single-syndrome mentality" about body shape and lipid abnormalities in HIV infection "might have been premature," she allowed. Those 48 posters and six slide shows on lipodystrophy are probably talking about more than one syndrome, in which PIs, nucleoside reverse transcriptase inhibitors (NRTIs), and "other factors" all play roles. But, Mulligan said, no one knows exactly where this dysregulatory role playing takes place.

One possible nugget to emerge from all this work, Mulligan proposed, is that metabolic upsets and body fat fluctuations are not always directly linked. For example, studies of switching from PI to non-PI regimens often show quick changes in metabolic values, but slow or no changes in body composition (see Table 2). On the other hand, large cohort studies, such as the Australian Prevalence Survey discussed later [abstract 201], do find more severe body shape changes in people with greater metabolic abnormalities.

And because there's no case definition for this syndrome (or syndromes), no one can for sure how often it (or they) occur. David Cooper attempted to boil down the statistical stew on metabolic and lipodystrophic side effects according to which drug classes people had tried in reported studies [abstract S21]. Among 526 who had sampled only NRTIs, the overall rate measured 35 percent. The number rose to 49 percent among 1552 people whom researchers classified as having NRTI experience "with or without" a PI. Another 3972 folks who had definitely taken PIs with NRTIs had a 62 percent prevalence of these side effects. And the rate among 6192 people who had taken "any" antiretroviral weighed in at a hefty 56 percent.

These numbers seem to say that protease inhibitors worsen the side effects of nucleosides, or perhaps only that the side effects accumulate with cumulative time on treatment. Cooper tried at least to line up the PI and NRTI puzzle pieces, if not to assemble them into a coherent picture (Table 1). At this date, though, there's no escaping the conclusion that sizable proportions of people taking antiretrovirals eventually endure these abnormalities.

But it's no mean feat to discern the contributions of specific drugs, or even of different drug classes. That point resounded in a report on 1077 people seen at eight US clinical centers as part of the HIV Outpatient Study (HOPS) [abstract 23]. Kenneth Lichtenstein, MD, reported that clinicians at the centers evaluated these individuals for "physical signs of fat redistribution" and characterized those signs as "none or mild" or "moderate to severe." Then statisticians pecked through the numbers, correlating them with sundry treatment and nontreatment variables.

At first blush, this analysis seems to single out indinavir and stavudine (d4T) as prime culprits in fat changes. Those were the only two drugs whose use, at any time, heightened the risk of moderate to severe abnormalities. Taking indinavir upped the odds by 97 percent (P = 0.003), while d4T raised the risk by 82 percent (P = 0.004). But being HIV positive for seven years or more and having AIDS for four years or more also significantly increased the risk of bad fat changes in this multivariate analysis, by 75 percent (P = 0.007). Since indinavir and d4T have been popular drugs, could it be that people with a longer disease course--and thus a higher risk of fat abnormalities--are simply the most likely to have tried indinavir, d4T, or both?

Lichtenstein and his HOPS partners tried to answer that question by classifying cohort members according to time on d4T and indinavir plus the number of nondrug risk factors each person had, including longer time since HIV diagnosis, longer time since AIDS diagnosis, older age, months since nadir CD4⁺ count, and untoward changes in body mass index--all of which had separate and strong associations with fat abnormalities. It turned out that bad fat changes never showed up in people with none of these other risk factors, regardless of the time they spent taking d4T or indinavir. But as nondrug risk factors piled up, the chance of fat abnormalities grew apace.

What does all this mean? "The strong statistical correlations between these [risk] factors," the HOPS team ventured, "do not allow the distinction between direct toxicity of the drugs versus the consequence of their beneficial effects in patients with advanced HIV." Reviewing these findings in her symposium presentation, Kathleen Mulligan put it another way. Perhaps antiretrovirals should not be seen as a cause of lipodystrophy, but only as one of many risk factors.

In a cross-sectional analysis of 1337 HIV-positive persons from across Australia, John Miller from David Cooper's Sydney group confirmed the compounding, confounding influences of drug and nondrug factors [abstract 201]. Clinicians evaluated study participants between November 1998 and June 1999, then number crunchers weighed the findings against numerous lipodystrophy risk factors.

Taking NRTIs for more than 60 months increased the risk of lipodystrophy by 92 percent (P < 0.001), whereas taking a PI for 23 or fewer months nearly doubled the risk (P < 0.001). Taking PIs longer than 23 months quadrupled the chance of lipodystrophy (P < 0.001). Once more, though, drug effects could not be untangled from other statistically significant risk factors: older age, symptomatic HIV disease, and effective viral suppression (lipodystrophy risk increased with decreasing viral loads). This analysis also found a link between more severe physical symptoms and greater metabolic abnormalities.

Although the Australian and HOPS studies, and others, find worse body change symptoms in people taking PIs than in the PI naive, a study that relied on DEXA scans to track wayward fat suggested that PIs sometimes don't add to lipodystrophy woes [abstract 202]. UCSF's Kathleen Mulligan presented results of this retrospective cross-sectional analysis of 180 treated and untreated HIV-positive men, and 44 seronegative controls, who had a DEXA scan in the past seven years.

Mulligan and colleagues found that men who had taken a PI plus NRTIs did not have higher trunk-to-appendicular fat ratios than men who had tried only NRTIs. She discerned "not even a hint of migration" to a higher ratio among those taking drugs from both classes. Although that finding inspired her suggestion that PIs don't affect some fat levels in some people, interpreting these results is a little tricky. The fat change in half of the study participants was buffalo hump, a proportion much higher than that seen in other lipodystrophy cohorts, so this was an atypical population.

A cross-sectional analysis of 646 HIV-positive people seen at six hospitals in Paris supported the Australian and HOPS finding that lipodystrophy is worse in people who take PIs than in those taking only NRTIs [abstract 17]. This study, presented by F. Boufassa, MD, is instructive because the PI-naive and PI-experienced groups did not differ significantly in time since HIV diagnosis (89.9 versus 84.3 months, respectively) or length of antiretroviral treatment (41.6 versus 45.6 months, respectively). So how long people were HIV positive and how long they took antiretrovirals could not confound any distinction between PI and non-PI regimens.

Yet the Paris team found that 64 percent of people with PI experience had an average of three fat change symptoms per person, whereas 39 percent of those naive to PIs had an average of two symptoms per person. The study also linked PI experience with higher total cholesterol (5.6 mmol/L versus 4.7 mmol/L in the PI naive, P < 0.001) and with higher triglycerides (2.0 mmol/L versus 1.4 mmol/L in the PI naive, P < 0.001).

Comparing rates of lipoatrophy, fat hypertrophy, and mixed symptoms in people currently taking d4T versus zidovudine (ZDV), Boufassa and coworkers found significantly higher rates of all syndromes among those taking d4T (P < 0.001). But a different analysis of these same patients (see abstract 20 in the next section) suggests that total time on NRTIs, not the NRTI being taken at the time, may have a greater influence on lipodystrophy.

Table 2. What happens when you trade a PI for an RT inhibitor?

Lead author, site [abstract number]	Type of study	No. of participants; time on PI	Strategy	Follow-up after switch; reason for switch	BL VL (copies/mL); BL CD4+ (cells/mm3)	Maintained or improved VL or CD4+?	BL metabolic values	Substantial improvement in values?	BL fat abnormalities	Substantial improvement in fat abnormalities?	Comments
Switch to nevirapine
A. Carr Australia [205]	Randomized, open-label	80; >6 m	49 continue PI; 31 switch to NVP, ABV, ADV, HU	24 wk; LD	< 400; mean 450	YES (for VL) NO (for CD4+; -70 cells/mm3)	Elevated CH and TG in about three quarters	YES CH, LDL, TG all sig (P = 0.001); HDL and IR did not improve	High LD scores on rating scale	MIXED Sig in central fat; total muscle mass and peripheral fat also	HU may explain CD4+ in switch group; peripheral and lean tissue loss due to GI intolerance of HU, ADV?
L. Ruiz Barcelona [206]	Randomized, open-label	106;>9 m	54 continue PI + d4T/3TC; 52 switch to NVP + ddI/d4T	36 to 48 wk; LD	< 400 for > 6 m; >100	YES VL rebounds in 6% in PI group, 10% in NVP group; CD4+ equivalent in 2 groups	Mean BL mg/dL: TR: 267 CH: 224	YES CH and TR sig in NVP group; no GLU change in either group	--	NO No sig improvements in anthropomorphic measures through 36 to 48 wk; DEXA showed stabilizing measures in NVP group	NVP group had improved quality-of-life scores
P. Barreiro Madrid [538]	Randomized, open-label	138; >1 y	104 switch from PI to NVP, keep NRTIs; 34 continue PI	6 m; LD	< 50 for > 6 m; > 200	YES VL rebounds in 11% on NVP and 29% on PI (P < 0.05)	77% with CH > 200 mg/dL, 57% with TG > 200 mg/dL	YES, but CH and TG in both groups perhaps because of dietary advice	Body shape abnormalities clinically noted in 70%	MIXED Partial improvement by self report in half who switched to NVP	VL rebound in PI group due to poor adherence in 90%
P. Tebas St. Louis and St. Paul [45]	Prospective open-label	40; >6 m	Switch from PI to NVP; keep NRTIs	Median 30 wk; LD	<200 for >6 m, < 40 at switch; median 511	YES 1 rebound >200 copies/ mL at 24 wk, resuppressed with PI	Mean mg/dL: TG: 298 CH: 202 HDL: 34 LDL: 115 GLU: 99 IN: 14 µU/mL	YES At 24 wk, sig in HDL; sig in IN; near-sig in TG (P = 0.065) and GLU (P = 0.077); ITT normalized in 3/3	--	NO Trend to worse central to peripheral fat ratio by DEXA at 24 wk	7 switched from NVP to EFV because of rash

Lead author, site [abstract number]	Type of study	No. of participants; time on PI	Strategy	Follow-up after switch; reason for switch	BL VL (copies/mL); BL CD4+ (cells/mm3)	Maintained or improved VL or CD4+?	BL metabolic values	Substantial improvement in values?	BL fat abnormalities	Substantial improvement in fat abnormalities?	Comments
Switch to efavirenz
S. Gharakhanian Paris [46]	Prospective open-label	32; median 24 m	Switch from PI to EFV, keep NRTIs	10 m; LD	< 500; NR	YES 5/32 (16%) VL failures (see Comments); CD4+ in 75% (P = 0.002)	CH > 6.8 mmol/L in 22%, TG > 1.7 mmol/L in 84%, IR in 86%, IGT in 43%	NO No sig improvements: CH > 6.8 mmol/L in 22%, TG >1.7 mmol/L in 69%, IR in 66%, IGT in 33%	Fat atrophy in 16%, adiposity in 3%, mixed syndrome in 80%	NO No sig improvement	Among people who stayed on EFV for 10 m, 23/28 (82%) <50 copies/mL by bDNA
E. Bonnet Toulouse [49]	Prospective open-label	43; >1 y	Switch from PI to EFV, keep NRTIs	6 m; LD	< 50 twice in 3 m before switch and at switch; mean 512	YES 4/35 (11%) VL rebounds, none above 3540 copies; mean CD4+ to 524	Mean mmol/L: TG: 2.7 CH: 5.9 HDL: 1.1 LDL: 3.5 VLDL: 1.0	NO Levels of TG, CH, LDL, and VLDL slightly; HDL slightly	Trunk to total body fat ratio 0.61 (normal 0.51); leg to total body fat ratio 0.22 (normal 0.28)	NO Neither measure changed much; nor did leg, thigh, hip, waist, or chest diameter	Adherence better after switch to EFV
E. Martinez Barcelona [50]	Prospective open-label	20; NR	Switch from PI to EFV, keep NRTIs	6 m; LD	< 200 for median 14 m; 280	YES 1/20 (5%) VL rebound to 555 copies; CD4+ stable	CH >200 mg/dL in 70%, TG >200 mg/dL in 85%, fasting GLU >110 mg/dL in 40%	YES TG 31% (P = 0.03) IR index 28% (P = 0.03) CH and GLU changed little	19/20 with central adiposity; 20/20 with leg and face wasting; 19/20 with arm wasting; 18/20 with buttock wasting	MIXED Waist-to-hip ratio improved (P = 0.06); 11/20 (55%)self report partial improvement	3 switched from EFV to NVP because of side effects
P. Viciani Seville [48]	Prospective open-label	39; 18 m	Switch from PI to EFV, keep NRTIs	12 m; LD	< 200 for median 11 m, < 20 in 81% at switch; NR	YES 89% < 200 copies, 85% < 20 copies; CD4+ stable	Mean CH 218.9 mg/dL, mean HDL 37.5 mg/dL, mean LDL 119.4 mg/dL, mean TG 289.2 mg/dL	MIXED CH and LDL change little, mean TG to 198.7 mg/dL	--	NO No improvement in BMI; 70% self report improved body shape, 11% self report worse shape	--
ABV = abacavir; ADV = adefovir; BMI = body mass index; CBV = Combivir; CH = total cholesterol; EFV = efavirenz; GI = gastrointestinal; GLU = glucose; HDL = high-density lipoprotein ("good") cholesterol; HU = hydroxyurea; IGT = impaired glucose tolerance; IN = insulin; IR = insulin resistance; ITT = insulin tolerance test; LD = lipodystrophy; LDL = low-density lipoprotein ("bad") cholesterol; NR = not reported; NRTI = nucleoside reverse transcriptase inhibitor; NS = not statistically significant; NVP = nevirapine; PI = protease inhibitor; sig = statistically significant; TG = triglycerides; VL = plasma viral load; VLDL = very low-density lipoprotein.

Lead author, site [abstract number]	Type of study	No. of participants; time on PI	Strategy	Follow-up after switch; reason for switch	BL VL (copies/mL); BL CD4+ (cells/mm3)	Maintained or improved VL or CD4+?	BL metabolic values	Substantial improvement in values?	BL fat abnormalities	Substantial improvement in fat abnormalities?	Comments
Switch to abacavir
M. Opravil Switzerland and Italy [457]	Randomized open-label	164; mean 22 m	84 switch from PI to ABV (+CBV), 79 continue PI	48 wk; simplify, reverse side effects	< 50 for >6 m; 513 in PI arm, 522 in ABV arm	YES 9/84 (11%) rebound in ABV group; 5/79 (6%) rebound in PI group; CD4+ equivalent in both groups	--	YES CH sig in ABV arm; TG and HDL with ABV, but not sig	--	NR	Rebounds in ABV group linked to 184V, 215Y/F,* and 41L mutations despite good adherence in ABV arm
W. Rozenbaum Paris [47]	French LD substudy of randomized open-label multicenter trial	34; median time on ART 25 m for ABV group, 21 m for PI group	17 switch from PI to ABV, 17 continue PI; all keep same NRTIs	24 week; LD	< 50 for >6 m; median 593 in ABV arm, 535 in PI arm	YES No rebounds >400 copies in either group; median CD4+ to 721 in ABV group, 566 in PI group (NS)	Median mmol/L:TG: 1.5 for ABV group, 1.4 for PI group; CH: 5.3 for ABV, 6.1 for PI; GLU: 4.8 for ABV, 4.4 for PI; IN: 4.9 µU/mL for ABV, 4.5 µU/mL for PI (all differences NS)	MIXED CH sig in ABV group vs. no change in PI group (P = 0.01); TG slightly in both groups; tendency to stabilized IN metabolism in ABV group	26 signs of LD in ABV group, 29 signs of LD in PI group	YES Central obesity resolved in half in ABV group, remained stable in PI group; 14/26 signs resolved in ABV group, 8/29 in PI group; 2 new signs in ABV group, 10 in PI group	Fat abnormalities appeared to remain stable or improve in 24 wk after Switch to ABV
F.D. Goebel Europe and Canada [51]	Randomized open-label	210; >6 m	105 switch PI for ABV, 105 continue PI	24 wk; simplify (with LD substudy)	< 50 for >6 m; median 504 for ABV group, 507 for PI group	YES 3 VL rebound to > 400 with ABV, 2 with PI; CD4+ ~ 40 cells with ABV, stable with PI	--	YES In 31-person substudy, CH, TG, and IN sensitivity improve in ABV group, not in PI group (NS)	--	NO In 31-person substudy, 3 self reports of improvement in LD signs, 1 worsening with ABV; no improvements, 3 worsening with PI	Hints of improvement in metabolic values and LD signs within 24 wk of switch to ABV
*No one had 215 mutation at baseline. ABV = abacavir; ADV = adefovir; BMI = body mass index; CBV = Combivir; CH = total cholesterol; EFV = efavirenz; GI = gastrointestinal; GLU = glucose; HDL = high-density lipoprotein ("good") cholesterol; HU = hydroxyurea; IGT = impaired glucose tolerance; IN = insulin; IR = insulin resistance; ITT = insulin tolerance test; LD = lipodystrophy; LDL = low-density lipoprotein ("bad") cholesterol; NR = not reported; NRTI = nucleoside reverse transcriptase inhibitor; NS = not statistically significant; NVP = nevirapine; PI = protease inhibitor; sig = statistically significant; TG = triglycerides; VL = plasma viral load; VLDL = very low-density lipoprotein.

Return to Article Contents

Nucleosides and their side effects

One reason it's tough to implicate specific drugs in lipodystrophy, David Cooper noted in his symposium presentation, is that "long-term exposure" to antiretrovirals "tends to implicate the most recent drug." Indeed, Paris clinicians reported a highly significant correlation between time on antiretroviral therapy and lipodystrophy (P < 0.0001) [abstract 17 above]. Cooper's group reported the same finding [abstract 201 above]. And Cooper went on to list three other reasons why the watchword should be caution when trying to figure whether certain nucleosides are more toxic than others: (1) So far, in vitro toxic effects don't correlate with in vivo effects; (2) hydroxyurea complicates any interpretation, because it's used most with didanosine (ddI), often along with d4T; and (3) to date, clinical studies implicating specific drugs have been poorly designed and nonrandomized.

Nevertheless, circumstantial evidence against d4T came in several flavors at this meeting. With Cooper's reasonable caveats in mind, clinicians and people with HIV will have to look closely at data like these, weigh their merits, and--if the balance appears to tip against d4T--decide whether that risk offsets potential benefits of the drug.

Three studies implicated d4T in lipodystrophy or high lactate levels, another group reported equivocal results, and a fifth team found no lipodystrophy difference between d4T and ZDV.

C. Goujard, MD, offered a closer analysis of NRTI influences in the 149 PI-naive individuals [abstract 20] in the 646-person Parisian cohort scrutinized in Boufassa's comparison of PI and non-PI regimens [abstract 17 above]. Among individuals without clinician-diagnosed lipodystrophy, the Goujard study found that 74 percent were taking ZDV, whereas 52 percent with lipodystrophy were taking ZDV (P = 0.005). The tables were turned for d4T: 47 percent with lipodystrophy were taking d4T compared with 26 percent without lipodystrophy (P = 0.008). But those results can be parsed differently: equivalent numbers with lipodystrophy were taking either ZDV (52 percent) or d4T (47 percent).

And the story gets even more complicated. People with lipodystrophy had a significantly longer average treatment history (49.3 months) than did people without lipodystrophy (35.4 months, P = 0.002). And the lipodystrophy group had taken ZDV longer (45.4 months) than the nonlipodystrophy group (33.5 months, P = 0.01). People with lipodystrophy had taken d4T for an average 53.7 months, whereas those without lipodystrophy had taken d4T for 41 months, but this difference was not significant (P = 0.21). So it appears that treatment duration, not the NRTI of the moment, may be the more telling factor, at least in this cohort.

Another French study, this one reported by J.M. Molina, MD, and colleagues, fingered d4T plus ddI as a more culpable combination than ZDV plus lamivudine (3TC) [abstract 19]. These investigators tallied lipodystrophy rates among 66 treatment-naive people studied for 30 months in an open-label comparison of ddI/d4T, ZDV/3TC, and ddI/d4T followed by ZDV/3TC. This analysis is limited by the small numbers in each group. Twenty-five were taking ddI/d4T, 31 ddI/d4T followed by ZDV/3TC, and only 10 ZDV/3TC.

At the 30-month checkpoint, 52 percent assigned to ddI/d4T had one or more signs of lipodystrophy, compared with 19 percent randomized to the sequential NRTI regimens, and 10 percent originally assigned to ZDV/3TC. Rates of lipoatrophy followed the same pattern: 44 percent for ddI/d4T, 19 percent for the NRTI pairs in sequence, and 10 percent for ZDV/3TC. So ddI/d4T did substantially worse than the other regimens in each analysis, even though more people in the other arms added a protease inhibitor during the 30-month follow-up.

Another way to interpret this study is that people with better initial virologic control while taking two nucleosides (those in the ddI/d4T arm at 24 weeks³) had higher rates of lipodystrophy. As already noted, the Australian Prevalence Survey [abstract 201] did link lower on-treatment viral loads with more frequent lipodystrophy. After 30 months of treatment in the French study, however, neither mean viral load nor median CD4⁺ count differed significantly in the three treatment groups. Molina also noted a possible "center effect" bias. A subgroup of treatment centers "well aware of the study aim" reported twice as much lipodystrophy as other centers.

A 624-person cross-sectional study of people treated for more than three months with two NRTIs and one PI found that rates of atrophy in the face, arms, legs, and buttocks increased significantly with longer NRTI treatment and, independently, longer PI treatment (P < 0.001 for each measure with both NRTIs and PIs) [abstract 46]. But this study, reported by S. Gharakhanian, MD (Rothschild Hospital, Paris), could pick out no drug-specific risks in a multivariate analysis. The two most-used NRTIs in this cohort were d4T and ZDV.

After publishing a study pointing to d4T as a cause of lipodystrophy,⁴ Thierry Saint-Marc, MD (Herriot Hospital, Lyon) substituted ZDV for d4T and followed 36 persons for nine months [abstract 52]. Triglyceride levels fell 29 percent after the switch, and serum lactates dropped by 37 percent. Levels of cholesterol, glucose, and insulin hardly budged. Study participants gained subcutaneous fat in the abdomen (+32 percent) and midthigh (+36 percent); biceps, triceps, and suprailiac skinfolds expanded. Four people said their body shape returned to its pre-d4T outlines, while 11 claimed a "major improvement."

Saint-Marc's results are surprising and so far unconfirmed by other investigators. Limitations of the study are its nonrandomized plan and reliance on patient reports of overall improvement in body shape. (In a randomized Australian study in which nevirapine, abacavir, adefovir, and hydroxyurea replaced a PI in people with lipodystrophy (see abstract 205 below), patient reports of improved lipoatrophy did not correlate with objective measures of lipoatrophy.) And a study from Toulouse found no more improvement in lipid levels or fat changes among people taking ZDV than among those taking d4T after they traded a PI for efavirenz and continued the nucleosides [abstract 49, Table 2].

Nonetheless, a bunch of seropositive people in Lyon no longer have skinny thighs or high triglycerides. And, as Donald Kotler, MD, observed in his Medscape report on the meeting, Saint-Marc's study "is important, since many people have speculated that fat redistribution is irreversible once it has started, based upon the lack of reversal in PI switch studies."⁵ Saint-Marc's results, on the contrary, imply that lost subcutaneous tissue can return within a year, if the right biologic switch gets thrown. Could that switch be as simple as stopping d4T when so many other medications apparently contribute to lipodystrophy? Still, Kotler concludes, "these results must be taken into account when designing future switch studies."

What about that even more notorious side effect of NRTIs, lactic acidosis? Clinicians at the University of California San Diego Medical Center believe they may have nosed out an earlier stage of this deadly threat, which they call hyperlactatemia [abstract 56]. All 20 individuals in this case series took d4T as part of their regimen.

J. T. Lonergan, MD, and coworkers defined hyperlactatemia as a syndrome of abdominal pain and distention, nausea, and slightly elevated lactate levels (only seven of 20 had lactate levels above 5 mmol/L). They calculated the incidence to be 13 times higher than that of frank lactic acidosis, which is about 1.3 cases per 1000 person-years of NRTI experience. Symptoms of hyperlactatemia improved when the clinicians stopped all NRTIs. They suggested that nucleoside treatment may resume--minus the d4T--once symptoms resolve.

None of these studies says much about mechanisms. The dropping lactate levels in Saint-Marc's d4T switch study, and the absence of hyperlactatemia in people not taking d4T in Lonergan's cohort, do suggest a higher risk of lofty lactates with d4T, and thus perhaps a higher risk of lactic acidosis. Researchers consider lactic acidosis a mitochondrial toxicity, as David Cooper observed in his symposium lecture. Yet a University of Alabama cell study published after the conference found "evidence for a number of mitochondrial defects with [ZDV], ddC [zalcitabine], and ddI . . . but only ZDV induced a marked rise in lactic acid levels."⁶ And an in vitro study presented at the meeting by J.F. Morlese and colleagues at Chelsea and Westminster Hospital in London yielded evidence suggesting that mitochondrial toxicity is at most a bit player in the lipodystrophy drama [abstract 42]. Until mechanisms are better understood, anticipating, avoiding, and managing NRTI-associated lipodystrophy will be chancy.

Return to Article Contents

Nonnucleosides, lipids, and the liver

NNRTIs had managed to duck the recent shower of slings and arrows loosed by side effects researchers. Aside from rash with nevirapine, and usually evanescent CNS escapades with efavirenz, this class of drugs looked pretty clean. Now continuing scrutiny has unearthed some possibly unhappy lipid news with NNRTIs, though it's surely too early to count hypercholesterolemia as an NNRTI side effect.

Turning their attention from PI rivals to NNRTIs, James Lenhard, PhD, and Glaxo colleagues turned up evidence that efavirenz boosts lipids and balloons livers in corpulent lab mice [abstract 41]. The Glaxo team treated 15 age- and weight-matched mice for two weeks with 5 mg/kg of efavirenz, 50 mg/kg of efavirenz, or with the oral gavage vehicle without efavirenz. The luckless rodents randomized to the 50-mg group endured a 22 percent increase in liver mass compared with control mice (P = 0.00004) and a 26 percent jump in alkaline phosphatase levels (P < 0.001). Lenhard theorized that efavirenz may cause this hepatic toxicity by choking off expression of several genes that keep lipid metabolism humming in the liver. Mice who got the 50-mg dose of efavirenz also had significantly higher levels of cholesterol (P = 0.0064), triglycerides (P = 0.0096), and free fatty acids (P = 0.0006) than did control animals. Higher levels of HDL ("good") cholesterol in the 50-mg group (P = 0.056) contributed to the higher total cholesterol values. Lenhard is not sure why efavirenz has these effects on lipids, but he suggests they may result from efavirenz-induced drops in expression of PPAR and CPT-1, enzymes involved in fat catabolism and mitochondrial fatty acid oxidation, respectively.

So, do livers and lipids of HIV-positive people react to efavirenz just like mouse livers and lipids? The usual caveats about extending mouse results to humans apply. A person weighing 70 kg, for example, takes only about 9 mg/kg of efavirenz daily, not 50 mg/kg, though the drug does have a long half-life. Product information on efavirenz notes 10 percent to 20 percent elevations of total cholesterol in HIV-negative volunteers who took efavirenz, and "modest elevations" of triglycerides and cholesterol in nonfasting seropositive people taking this NNRTI. Cholesterol monitoring "should be considered," according to product information, which also recommends regularly checking liver enzymes in people thought or known to have hepatitis B or C infection.

A study reported last year by Graeme Moyle, MB, and his Chelsea and Westminster colleagues found that lipids kept rising after a switch from a PI to efavirenz.⁷ Moyle extended that research at the Conference on Retroviruses, reporting that "intervention range" cholesterol levels are not unusual in people taking efavirenz or nevirapine [abstract 21].

This cross-sectional review of 121 people taking efavirenz and 120 taking nevirapine found that 25 percent in the efavirenz group and 12 percent in the nevirapine group had cholesterol levels above 6.5 mmol/L (240 mg/dL) (P = 0.056). But Moyle pointed out several factors that cloud interpretation of these findings. First, cholesterol levels were rarely measured after a fast. Second, lots of people had PI experience, and those who did were more likely to have hypercholesterolemia (25 percent) than those still naive to PIs (14 percent). In fact, when these investigators did a multivariate analysis adjusting for gender, age, and duration of NNRTI and PI experience, they associated efavirenz with normal cholesterol values. Low CD4⁺ counts also correlated with normal cholesterol readings.

And in a different analysis of 135 people whose initial regimen included efavirenz or nevirapine with ZDV or d4T, Moyle's multivariate analysis correlated only age and triglyceride levels--not specific drug combinations--with heightened cholesterol. In fact, combos including efavirenz and d4T seemed to protect against soaring triglycerides. But Moyle couldn't explain that association and suggested it may be a "chance finding."

Meanwhile, three of four studies involving people who substituted efavirenz for a PI found no striking improvements in lipid levels six to 12 months later [abstracts 46, 48, and 49, Table 2]. One of those studies, in fact, found that levels of cholesterol, triglycerides, LDL ("bad") cholesterol, and VLDL cholesterol inched upwards in the six months after the switch to efavirenz [abstract 49]. One study found a significant drop in triglycerides six months after efavirenz replaced a PI (P = 0.03) but little change in cholesterol [abstract 50, Table 2].

Return to Article Contents

PIs, MIs, and other "coronary events"

No one should be surprised that research has yet to forge a firm link between PI therapy and heart disease. A host of risks may contribute to clogged arteries and heart attacks, risks shared or not shared by people taking PIs. And a few years of PI treatment, with consequent lipid leaps, may not be enough to send even some high-risk persons to the coronary care unit. Indeed, several studies found scant evidence of a heightened short-term coronary risk with PIs. But two studies did suggest that protease drugs may start monkeying with cardiomechanics early on.

Daniel Klein, MD, and colleagues marshaled the vast Kaiser-Permanente managed care database to compare rates of coronary heart disease (CHD) (myocardial infarction [MI], other forms of ischemic heart disease, and angina pectoris) in 4526 HIV-positive persons and a random sample of 45,123 people without reported HIV infection [abstract 33]. The seropositive group did have a significantly higher rate of hospital admissions for CHD than the control group (5.6 per 1000 person-years versus 3.3 per 1000 person-years, P = 0.03). That difference can probably be traced to plentiful nondrug CHD risk factors in the people with HIV infection: 59 percent had hypertension, 54 percent had hypercholesterolemia, 41 percent smoked, and 15 percent had diabetes mellitus.

The nature of the study precluded the possibility of linking high cholesterol levels with antiretrovirals on a person-by-person basis. But Klein's analysis turned up no difference in rates of heart disease before people started PI therapy and after they did. The age-adjusted CHD rate measured 5.5 per 1000 person-years in the pre-PI period and 5.7 per 1000 person-years since these people started PIs. No one younger than 35 had an admission for CHD while taking a PI. The rates per 1000-person years after PI treatment began measured 1.5 for 35- to 44-year-olds, 3.7 for 45- to 54-year-olds, 15 for 55- to 64-year-olds, and 24.8 for seropositive people older than 65. Those numbers match rates reported for the US male population in several large studies. Klein concluded that PIs don't inflate the short-term risk of heart disease.

In a rare efflorescence of antiretroviral industry bonhomie, researchers from Merck, Agouron, Abbott, and Roche teamed up to tote MI rates in people randomized to PI therapy or a nucleoside-only arm in four clinical trials [abstract 34]. Paul Coplan, ScD, and confreres reported no substantial differences in MI rates between the PI and non-PI arms, although the follow-up in this analysis lasted only about 12 months. Rates in PI arms ranged from 0.89 to 2.23 per 1000 person-years, compared with 0.0 to 2.97 per 1000 person-years in the non-PI arms. Person-years of follow-up in the PI arms stretched from 1129 to 3944. Again, the MI rates among PI takers were consistent with those reported in several gargantuan cohort studies of men and women in the US, and with MI rates in ACTG 193a, a two-year study of reverse transcriptase inhibitors.

But short-term equivalence with CHD rates in seronegatives and in the PI naive are cold comfort to people with lipid levels so high they've had to make lipid lowerers part of their HAART regimen. So the question arises whether a few years of PI treatment touch off morphologic changes that presage cardiovascular disease. One study found no higher rate of coronary artery plaques in PI takers than in seropositive PI-naive individuals. Whereas another study discerned a hint of carotid intimal thickening in women with hypercholesterolemia while taking PIs, a third group definitely correlated thicker carotid arteries with HAART. A fourth team found brachial artery endothelial dysfunction in people with PI experience.

The plaque study, by M. Depairon, MD, and coworkers at Vaudois University Hospital in Lausanne, compared 27 HIV-positive PI-naive individuals with 131 people who had an average 27 months of PI experience [abstract 30]. The groups didn't differ in obesity, history of smoking or hypertension, or average viral load or CD4⁺ count. But the PI group had a lower average CD4⁺ nadir (212 cells/mm³) than did the control group (344 cells/mm³). Ultrasound picked up plaques in femoral or carotid arteries of 87 people, whereas 71 were plaque free. In a multivariate analysis, only older age (P < 0.001) and smoking (P < 0.05) correlated with plaques. PI treatment did not. Total cholesterol above the intervention cutoff of 6.2 mmol/L did increase the odds of having a plaque 3.4 times, and that correlation leaned toward statistical significance (P = 0.09). As the Lausanne team noted, longer follow-up of this cohort will be necessary to determine whether that cholesterol jump eventually does raise the risk of plaques in people taking PIs.

Because carotid artery blockage correlates with clogged coronary arteries, Judith Currier, MD (University of California, Los Angeles) and colleagues also trained their ultrasound equipment on these vessels in 16 premenopausal women taking protease inhibitors for an average of 11 months and 16 age- and race-matched seronegative women [abstract 32]. Currier and coworkers excluded smokers and women with a history of hypertension or diabetes from the study. Mean HDL ("good") cholesterol was higher in the control group (P = 0.007), whereas average triglycerides were higher in the seropositive women (P = 0.011). The HIV group also had higher average cholesterol and LDL levels than the control volunteers, but these differences were not significant.

Despite these bad omens in the PI takers, Currier found no overall difference between the two groups in carotid intima media thickness. When she ranked the PI-treated women into escalating cholesterol strata, she did note "a suggestion" of thicker arteries in women with higher lipid levels than in their matched controls (correlation coefficient 0.489, P = 0.064). But the small size of this study and short duration of PI treatment make it hard to figure whether that suggestion will translate into a real problem.

A similar and somewhat larger case-control study by clinicians in Paris did, however, find thicker carotid arteries in people taking HAART than in healthy controls matched for age, sex, body mass index, and smoking [abstract 31]. N. Cheminot, MD, and colleagues reported significantly greater intima media thickening in the 36 HAART-treated individuals (P = 0.03), 32 of whom were taking PIs. Multivariate analysis showed that older age (P < 0.01), greater waist circumference (P < 0.05), and higher cholesterol levels (P = 0.03) correlated with abnormal arteries. Duration of antiretroviral therapy and type of therapy did not correlate.

James Sosman, MD, and colleagues from the University of Wisconsin also pinned down a PI-related risk for heart disease, impaired endothelial cell function measured by flow-mediated vasodilation of the brachial artery [abstract 29]. The Wisconsin team compared 22 people taking PIs for an average 30 months with 10 people taking non-PI regimens. The groups were similar in age, lean body mass, waist-to-hip ratio, blood pressure, duration of antiretroviral therapy, time since HIV diagnosis, and CD4⁺ count. The PI group had a somewhat higher average viral load (22,583 versus 3515 copies/mL), but this difference was not significant.

Other differences were significant. The PI group had significantly lower flow-mediated brachial artery vasodilation (P = 0.002), and higher levels of cholesterol (P = 0.008) and triglycerides (P = 0.029). Sosman and coworkers calculated that 65 percent of the variance in vasodilation could be explained by PI therapy and wider brachial artery diameter in the PI group.

Return to Article Contents

Making Side Effects Go Away

What can be done about antiretroviral side effects? If it's too late for prevention, William Powderly suggested four options in his side effects symposium lecture:

• Additional therapy, such as lipid-lowering drugs
• Behavior modification, for example through exercise or nutrition counseling
• Switching treatment
• Stopping treatment

The latter two options have certainly been the most studied. The conference featured only two studies of the first two approaches.

Behavior modification wins hands down as the safest strategy. And studies presented so far generally show some benefits. But even dazzling results from diet and exercise studies probably wouldn't boost the behavioral approach much. Most people with HIV infection probably feel the same way about diet and exercise as most people without HIV infection: Why go through all that trouble when you may get the same results by taking a pill, switching a pill, or stopping a pill? To have an impact, the obvious answer to that question must be backed up with dietary or exercise counseling.

The conference featured no diet studies per se. But in a Spanish study that randomized 138 people with lipodystrophy to continue a PI or switch to nevirapine, both groups enjoyed six-month drops in triglycerides and cholesterol [abstract 538, Table 2]. Why? P. Barreiro, MD, surmised that the parallel improvements can be traced to dietary advice given to all study participants, all of whom had taken a PI regimen for at least one year.

The meeting's sole exercise study came from Kevin Yarasheski, MD, and colleagues at Washington University in St. Louis [abstract 54]. Sixteen weeks of progressive weight training, they found, lowered high triglycerides and added lean body mass in 17 seropositive men. But it had no effect on levels of total cholesterol, HDL cholesterol, LDL cholesterol, insulin, C peptide, proinsulin, or glucagon.

The individually supervised workouts included three upper-body exercises and four lower-body exercises initially repeated three to four times per session, during three to four sessions per week. As the men gained strength, they hefted increasingly heavier weights four to five times per session, during four sessions per week. The men's average age was 42, their body mass index averaged 22 kg/m2, and their whole-body adiposity averaged 19 percent. Fifteen of the 17 were taking PI-containing HAART, and two had never taken PIs.

After 48 exercise sessions, average fasting serum triglycerides dropped from 288 mg/dL to 209 mg/dL (P = 0.05). Exercise lowered triglycerides in 11 of the 17 men, and nine of those 11 had hypertriglyceridemia before starting the program. Whole-body lean mass increased by 1.25 kg (P = 0.01), as did lean mass in the trunk and arms (P < 0.04). And the men got stronger. Muscle strength increased by 22 to 35 percent (P < 0.0001). But exercise did not help the men shed whole-body adipose tissue or arm or trunk fat.

If 16 weeks in the gym doesn't dent cholesterol concentrations, what about statins? These lipid lowerers can get the job done, of course, but they roam the same metabolic byways as PIs, so interactions are likely. Carl Fichtenbaum of the University of Cincinnati offered the first thorough analysis of pharmacokinetic clashes between these two groups of drugs [abstract LB6]. He and an army of ACTG colleagues tested pravastatin, simvastatin, and atorvastatin, all at doses of 40 mg daily in HIV-negative volunteers also taking 400 mg of both ritonavir and saquinavir twice daily.

Levels of active atorvastatin increased 74 percent, simvastatin concentrations shot up 2676 percent, and pravastatin levels dropped 47 percent. You won't have to consult your PharmD to figure that simvastatin should not be given with ritonavir/saquinavir, or its dose should be reduced. But why tinker with doses when atorvastatin or pravastatin are less hair-raising options? Fichtenbaum advised that atorvastatin "may be used with caution" with ritonavir/saquinavir and that pravastatin is "safe for further study." Clearly, much work remains to be done.

Return to Article Contents

Another round of PI switch studies

Although the impact of nonnucleosides on lipid levels--either as first-line therapy or after a PI--remains far from certain (see page 76), trading a PI for an NNRTI has appealed to many clinical investigators. The ICAAC and Lisbon meetings last fall showcased 15 studies in which people dropped their PI in favor of a nonnuke, and three studies in which abacavir replaced a PI.⁸ PI-induced viral load and CD4⁺ count improvements almost always persisted after the switch. But short-term changes in lipid and glucose measures proved more of a mixed bag, and body shape abnormalities tended to improve little in the first six to 12 months after the switch.

Trends in 11 studies at the Conference on Retroviruses were also mixed. And that's no surprise, since many of these reports were updates of earlier studies. One new study earned attention because of its size (though still modest) and randomized plan. Results of this 11-site Australian switch study, presented by Andrew Carr, MD, confirmed earlier findings: Virologic and immunologic gains can be maintained after switching from a PI; some lipid measures improved; many fat abnormalities did not [abstract 205]. But central adiposity did wane when people swapped a PI for a handful of reverse transcriptase inhibitors.

• Switch to nevirapine. The Australian study randomized 80 people with PI- associated lipodystrophy to continue their PI regimen or to continue only their nucleosides and trade the PI for nevirapine, abacavir, adefovir, and hydroxyurea. Everyone had a viral load below 400 copies/mL for at least six months, and no one had tried abacavir, adefovir, or any NNRTI before. Forty-nine people stayed with their PI, and 31 switched.

• After 24 weeks of follow-up, 6 percent in the switch group and 19 percent in the PI group had viral loads above 50 copies/mL, a nonsignificant difference. Most switchers maintained low viral loads, even though 11 stopped adefovir, eight stopped hydroxyurea, three stopped abacavir, and two stopped nevirapine because of side effects. The CD4⁺ news wasn't so good for the switch group. The mean baseline of 450 cells/mm³ slipped by 70 cells, possibly because of the hydroxyurea. And that loss was statistically significant (P = 0.002).

  People who said good-bye to their PI did shed central fat, measured as square centimeters of subcutaneous abdominal fat (P < 0.001), centimeters of waist size (P = 0.003), kilograms of central abdominal fat (P = 0.002), or central lipodystrophy severity score (P = 0.002). But total muscle mass also fell in the switch group, a change Carr suggested may be attributable to GI intolerance of hydroxyurea and adefovir. Switchers also lost more peripheral fat than people who kept their PI. Aided by hindsight, one might suggest that the switch regimen in this study was more ambitious than necessary, though people who tolerated three or four of the new drugs would dismiss that cavil.

  Lipid profiles improved significantly in the switch group, as levels of triglycerides, total cholesterol, and LDL cholesterol all dropped significantly (P = 0.001 for all). Levels of HDL cholesterol, C peptide, and insulin resistance did not change significantly within 24 weeks of the switch.

  Three other studies swapped a PI for nevirapine because of lipodystrophy [abstracts 45, 206, and 538, Table 2]. Two of these studies [206 and 538] randomized people to continue their PI regimen or trade it for the NNRTI. The third study [45] shifted everyone to nevirapine and followed them prospectively. Together these three studies tracked 196 people who stopped a PI and started nevirapine. Along with the 31 switchers in the Australian study, that brings the nevirapine switch sample to a respectable 227.

  After follow-up stretching from 30 to 48 weeks in the three studies, all yielded good evidence of preserved virologic and immunologic gains. In a randomized, 138-person Spanish study [abstract 538], there were significantly more viral load rebounds in the PI continuation group than in the nevirapine group (P < 0.05), a difference the investigators attributed to much better adherence in the nevirapine group.

  Lipid and glucose values also improved impressively in the three nevirapine studies. Changes in various lipid and insulin scores often became statistically significant within six months of the drug switch (Table 2). As noted, total cholesterol and triglyceride concentrations fell in both arms of one randomized Spanish study [abstract 538], probably because everyone got dietary advice and many of them followed it. Results like that make one wonder how much well-structured dietary intervention would rein in runaway lipids in people who stick with their PIs. It also makes one wonder why diet isn't tracked more closely in switch studies, or reported if it is tracked.

  Physical signs of lipodystrophy didn't improve much in any of the nevirapine studies. Even after 48 weeks of follow-up in the other randomized Spanish trial [abstract 206, Table 2], body measurements detected no meaningful reversal of fat abnormalities. DEXA scans showed, at least, that fat problems weren't getting worse with nevirapine. But those problems may reach a point where they simply can't get any worse no matter what drugs a person takes. In a prospective study by clinicians in St. Louis and St. Paul [abstract 45], DEXA documented worsening central-to-peripheral fat ratios 24 weeks after the switch to nevirapine.

• Switch to efavirenz. Four prospective open-label studies tracked a total of 134 people who traded a PI for efavirenz because of lipodystrophy [abstracts 46, 48, 49, 50, Table 2]. Postswitch follow-up ranged from six to 12 months.

• Once more, PI treatment gains generally remained intact in these studies, all of which kept people on their baseline NRTIs. Of course, not everyone with an undetectable viral load while taking a PI will maintain that status indefinitely after a switch, just as everyone who continues a PI will not stay aviremic forever. In a study at the Rothschild Hospital in Paris, for example, five of 32 people (16 percent) had a virologic breakthrough within 10 months of starting efavirenz [abstract 46]. In a study from Toulouse, four of 35 (11 percent) had rebounds within six months, but none to levels above 3540 copies/mL [abstract 49].

  A modest surprise in the efavirenz studies is that metabolic measures generally didn't improve much. Perhaps, as other studies have suggested (see page 76), efavirenz has its own ill effects on lipids or for some reason can't undo PI-induced damage. Or perhaps the drug's tendency to boost HDL ("good") cholesterol masks a lipid benefit because total cholesterol levels don't drop much (although HDL fell slightly after a switch to efavirenz in one study [abstract 49, Table 2]).

  In any event, only one of these four studies, involving only 20 persons [abstract 50], found a significant lipid improvement with efavirenz, a 31 percent drop in triglycerides after six months (P = 0.03), along with an improved insulin resistance index (P = 0.03). In the other studies, cholesterol, LDL, and triglyceride levels usually changed little after a switch to efavirenz, and sometimes concentrations rose [abstract 49].

  None of the efavirenz studies saw much in the way of body fat improvements. The best news came from clinicians in Seville, who found that 70 percent of 39 individuals had improved body shape 12 months after switching to efavirenz, according to self reports [abstract 48]. Another 11 percent in this cohort said their shape got worse. The investigators reported no improvement in body mass index. In a 20-person study, waist-to-hip ratio improved with efavirenz, but not quite significantly (P = 0.06), after six months of efavirenz [abstract 50]. Eleven of these 20 people said they thought their bodies looked better after the switch.

• Switch to abacavir. Three studies, all of them randomized and two of them fairly large, looked at the NRTI abacavir as a PI replacement [abstracts 47, 51, and 457, Table 2]. But only the smallest of the three [abstract 47], a 34-person substudy of a larger trial, looked specifically at metabolic and body shape changes. In one of the other trials [abstract 51], a 31-person substudy focused on lipodystrophy.

• Once more, over the short term, abacavir generally preserved viral load and CD4⁺ gains enjoyed with a PI. Nine of 84 (11 percent) endured a viremic rebound within 48 weeks in a Swiss-Italian study [abstract 457]. In a European-Canadian study, only three of 105 people (3 percent) saw their viral load climb back over 400 copies/mL within 24 weeks of switching to abacavir [abstract 51]. Investigators linked the rebounds to familiar mutations in all cases (184V, 215Y/F, and 41L), despite reportedly good adherence. Only two of 105 persons in the group that continued their PI had viral rebounds. In two of the three studies [abstracts 47 and 51], CD4⁺ gains in the abacavir switch group outpaced those in the PI continuation arm, but the differences were not statistically significant in either case.

  In two of the three abacavir studies, cholesterol loads dropped significantly within 24 to 48 weeks after people switched [abstracts 47 and 457]. Triglyceride levels improved, but not significantly, in all three studies. People who continued their PI therapy generally saw no improvements in these measures.

  In the 34-person study at the Rothschild Hospital in Paris, central obesity improved in about half of the 17 who switched to abacavir, and 14 of 26 physical signs of lipodystrophy were said to improve [abstract 47, Table 2]. Some signs also improved in the 17 people who continued PIs, but those people also reported new fat abnormalities during the 24 weeks of follow-up. A 31-person substudy of the big European-Canadian abacavir trial noted only three self reports of improved physical signs within 24 weeks of the switch to abacavir [abstract 51].

Return to Article Contents

Fishing for the bottom line

Who can plumb this ocean of data? Table 2 attempts to gauge major trends, though any such attempt necessarily obscures subtleties of individual studies that may explain as much, or more. The table tries to answer three questions, the same three posed in a review of switch studies presented in the autumn of 1999⁸:

1. Does switching to a PI-sparing regimen maintain virologic suppression and CD4⁺-cell gains?
2. Does switching substantially improve out-of-whack lipid and glucose levels?
3. Does switching substantially improve fat abnormalities?

The answers are a tad arbitrary, especially for questions 2 and 3. Generally, a "substantial" improvement here means one marked by statistically significant improvements in objectively measured parameters. Self reports of better body shapes are given less weight, because such reports can be inaccurate.

Enough equivocation. On to the final scores. If one reviews all switch studies presented at ICAAC, Lisbon, and the 7th Conference on Retroviruses, here's what one finds (Table 3):

1. 22 yeses and 2 nos
2. 13 yeses and 6 nos
3. 2 yeses and 9 nos

The overall trends are more consistent than one might expect. If viral load is well controlled for six months or more with a PI, dropping that PI in favor of a reverse transcriptase inhibitor only occasionally leads to erosion in virologic gains. Postswitch CD4⁺ counts were wobbly in two studies that included hydroxyurea as part of the new regimen, but otherwise CD4⁺ counts stayed good or got even better. Switch regimens also tended to improve bad lipid values. Table 3 suggests that efavirenz may be an exception to that rule, but the results are hardly clearcut. For example, about one third of patients were taking efavirenz in the two "nevirapine or efavirenz" studies that charted better metabolic values after the switch (row one of Table 3).

The physical stigmata of lipodystrophy heal slowly--if at all--within a year of switching to a PI-sparing regimen. Why don't things get better faster? A thoughtful analysis by E. Bonnet, MD, and coworkers from Purpan Hospital in Toulouse, lists some of the possibilities [abstract 49]:

1. Improvement may take longer than six or 12 months. (Indeed, anecdotes attest to physical improvements more than a year after stopping a PI and starting an NNRTI.)
2. PI-linked lipodystrophy may be irreversible. (Perhaps in some people, but there certainly are reports of reversion, with before-and-after photos and fancy objective measures to prove it.)
3. Stopping only the PI may not be enough if NRTIs contribute substantially to lipodystrophy. (Good point. In most reported studies, backbone NRTIs were continued when the PI was dropped. Maybe, as French investigator Thierry Saint-Marc would argue, you have to dump the d4T as well [abstract 52]. But in Bonnet's study, people taking ZDV did no better than those taking d4T after the switch to efavirenz. Randomized trials will have to sort out whether start or switch regimens with certain NRTIs do better than others in averting or reversing lipodystrophy.)
4. Metabolic disorders and physical changes may be caused by treatment-induced stifling of viral replication, not by specific drugs. (Donald Kotler has made this argument.⁹ If true, there's little hope for avoiding this syndrome and simultaneously harnessing HIV.)
5. HIV itself may cause lipodystrophy. (Probably not, Bonnet added, because the syndrome seems to correlate with low viral loads, not high ones. And Bonnet did DEXA scans on treatment-naive HIV-positive people "for years" and never found any with lipodystrophy.)
6. Efavirenz, the switch drug in Bonnet's study, may maintain lipodystrophy. (This is the argument advanced by Glaxo's James Lenhard [abstract 41 above] Again, only randomized switch studies with different NRTIs can hope to approach an answer.)

The Toulouse group's final point is its most cogent yet. "None of these hypotheses is fully satisfactory," they wrote in their poster report. "The mechanism of lipodystrophy is probably multifactorial and still needs to be completely elucidated." The only point here that one might quibble with is the word "probably."

Return to Article Contents

When to Start, Re-revisited

The many side effects of antiretrovirals threaten people with HIV infection not only because high lipids may lead to heart disease, or osteopenia to broken bones, but also because they intensify the urge to stop treatment. As UCSF's Carl Grunfeld, MD, PhD, observed, lipodystrophy is a public health threat because people are abandoning antiretrovirals for fear of heart attacks and hollow faces.

Whether planned treatment pauses bring more benefits than risks remains uncertain (see the next section). At the same time, clinicians must reevaluate the possibility that a decided benefit of starting later--avoiding side effects--handily outweighs the risks of delayed therapy. At an American Foundation for AIDS Research symposium last fall, only one of five HIV clinicians on a roundtable--Wafaa El-Sadr, MD--said she had not become more conservative in starting antiretrovirals. Her reason? She was much more conservative than her colleagues two or three years ago, and now they're coming around to her point of view.

Another clinician who forsook aggressive intervention a few years back, Keith Henry, MD, of Regions Hospital in St. Paul, has just tacked his theses on the cathedral door in an Annals of Internal Medicine dissection of US treatment guidelines.¹⁰ The prime goal of therapy, he wrote, should not be making plasma virus vanish, but maintaining "the long-term health of the patient while avoiding drug-related toxicity and preserving viable future treatment options."

In an editorial reply, Oren Cohen, MD, Executive Secretary of the US Department and Health and Human Services (HHS) guidelines panel, agreed that the "long-term strategic approach" espoused by Henry "is undoubtedly where the field should go."¹¹ In fact, the latest update of the HHS guidelines,¹² Cohen added, emphasize that "the goals of antiretroviral therapy must extend beyond simply suppressing plasma viremia as much as possible for as long as possible." In its millennial edition of treatment guidelines,¹³ the International AIDS Society-USA Panel stuck with its 1998 recommendation to begin treatment at 350 cells/mm³, the same number cited by the British HIV Association.¹⁴ But the HHS brain trust still favors a 500-cell threshold.¹²

Whether guidelines guide or are guided by clinical practice is another question. And although one might not call this growing compassionate conservatism a sea change in antiretroviral planning, it signifies at least a cyclic ebbing of the tide.

The University of Washington's William Powderly gave the strongest voice to this shift at the conference, citing 200 cells/mm³ as an appropriate start signal as health- and life-threatening side effects of treatment continue to emerge. On the other side of the intervention divide, many still argue that the cardinal threat to the health of HIV-positive people remains HIV itself. And this meeting did not lack reports attesting to the dangers of delayed treatment. Is this evidence enough to slow the pendulum's slice toward later intervention? Here are the numbers:

Return to Article Contents

Arguments for starting sooner

Ann Meibohm, PhD, and colleagues at Merck lined up results from four trials of that HAART trendsetter--ZDV, 3TC, and indinavir [Table 4, abstract 521]. The four studies involved folks at four different points in disease progression: acute infection, "early" disease, "intermediate" disease, and "late" disease. The 48-week RNA and CD4⁺ results pretty much reflect what one would expect: The earlier treatment starts, the better these numbers look.

Of course this type of analysis has limitations. The acute and "early" groups were virtually identical in CD4⁺ count, though viral loads were much higher among acute patients often still in the throes of primary infection. More important, perhaps, is that people with "intermediate" disease had mostly crossed the 200 CD4⁺-cell threshold generally cited by the most conservative clinicians. Yet these people made up the legendary Merck 035 cohort, the ZDV-experienced group whose good response catapulted triple therapy onto its standard-of-care pedestal. Surely, advocates of delayed intervention would latch onto their inclusion in this analysis as platinum proof that treatment can wait. Of course the "late" group, who began ZDV/3TC/indinavir at a median CD4⁺ count of 15 cells/mm³, fared poorly compared with the other three groups.

Another limitation of this analysis is its 48-week cutoff. That prevents a good assessment of longer-term PI side effects. But the side effects data presented merit analysis. People in the "late" group had the highest 48-week rate of "serious, drug-related clinical adverse experiences," 9 percent--not a surprise given their advanced disease. But rates of serious side effects differed little in the other three groups: 4 percent for the acute group, 2 percent for the "early" group, and 3 percent for the "intermediate" group. So one might conclude that a HAART regimen will cause similar problems no matter when it begins in people without the most advanced HIV disease. But that may be stretching the point, because there weren't many serious side effects in any of these three groups (Table 4).

Richard Moore, MD, and colleagues from Johns Hopkins looked at this same question from a different perspective [abstract 522]. They trawled through medical records of 674 people in their AIDS clinic who began HAART (with efavirenz or one or two PIs) after January 1, 1996. Sure enough, people who started a potent regimen at a higher CD4⁺ count had a better shot at a viral load under 400 copies/mL and a better chance of sustaining that response for more than six months. Among 184 people who began at a CD4⁺ count above 350 cells/mm³, 81 percent got their viral load below 400 copies/mL and 43 percent kept it there. Among 132 persons who started with 200 to 350 CD4⁺ cells/mm³, respective percentages were 74 percent and 34 percent. And among 357 people who started with fewer than 200 CD4⁺ cells/mm³, 65 percent went below 400 copies/mL and only 28 percent stayed there for more than six months.

A multivariate analysis singled out four factors that independently correlated with a sub-400-copy response: Beginning HAART at a CD4⁺ count above 350 cells/mm³ increased the odds by 90 percent compared with starting at a T-cell count under 200 cells/mm³ (P = 0.003). Starting treatment at a viral load above 100,000 copies/mL trimmed the chance of a sub-400 response 31 percent compared with starting at or below 100,000 copies (P = 0.04). Injecting drug use cut the odds of getting below 400 copies 32 percent compared with not injecting drugs (P = 0.04). And missing more than three antiretroviral doses in the past two weeks nearly halved the chance of a sub-400 response compared with missing fewer than three doses (P = 0.02).

The Johns Hopkins team concluded that their results suggest an advantage to starting HAART at a CD4⁺ count above 350 cells/mm³. But delayed-treatment advocates could easily use the same data to turn the tables on this argument. At least in an inner-city cohort like this, more than half of whom injected illegal drugs, the 43 percent six-month sub-400-copy rate in people who began HAART above 350 cells/mm³ doesn't inspire confidence in early treatment. That result means 57 percent in this group needed a second-line combination before six months passed, and everyone knows second-line therapy tends to fail faster than the first regimen.

Arguments made by Keith Henry in his antiretroviral guidelines criticism may apply to a cohort like this.¹⁰ Rather than starting treatment with a PI or an NNRTI and risking resistance in someone with more than 350 cells/mm³, options for such individuals would include not beginning treatment and monitoring closely while working on problems like adherence, waiting for development of more tolerable once-daily regimens. Or one might opt to dampen viral replication with a nucleoside-only regimen slow to elicit resistance mutations.

Perhaps a stronger case for early intervention can be made from the already discussed retrospective population study in British Columbia [abstract 73]. In a multivariate analysis of 950 people starting PI or NNRTI regimens from August 1996 to December 1998, only a lower pretreatment CD4⁺ cell count and poor adherence correlated with a higher risk of death. After adjusting the calculation for age, gender, baseline AIDS diagnosis, baseline viral load, injecting drug use, and physician experience, these investigators found that every 100-cell lower CD4⁺ count at baseline raised the odds of death 35 percent (P < 0.001).

Table 4. When to Start: 48-week HAART Data in Four Groups at Different Disease Stages

	Acute infection (study 042)	Early disease (study 060)	Intermediate disease (study 035)	Advanced disease (study 039)
n	47	199	33	108
Median baseline CD4+ cells/mm3	546	574	133	15
Median baseline RNA copies/mL	91,720	7720	41,900	76,900
<500 copies/mL at 48 wk (%)*	90	83	80	52
<50 copies/mL at 48 wk (%)*	83	80	67	47
CD4+ from baseline at 48 wk§	212	164	187	126
Serious, drug-related clinical   adverse experience, n (%)	2 (4)	3 (2)	1 (3)	10 (9)
*Calculated in intent-to-treat analyses in which withdrawal "for therapy-related reasons" was counted as a failure. § Calculated in an intent-to-treat analysis in which "the last observation was carried forward to time points after withdrawal for patients who withdrew for therapy-related reasons." Source: Ann Meibohm, PhD, abstract 521.

Return to Article Contents

LPRs, TCRs, and early Rx

Most agree that, at various points in HIV disease progression, various immune defenses erode, perhaps past the point where they can be rebuilt. Only in the past few years, though, has research begun to define potential points of no return. Two small studies at the San Francisco meeting suggested that earlier therapy shores up immune bastions better than later intervention.

C. Tortajada, MD, and coworkers in Barcelona tracked T-cell subset changes for 12 months in 32 antiretroviral-naive people beginning treatment with a PI- containing regimen [abstract 337]. Seventeen started with a median CD4⁺ count of 694 cells/mm³ (all had more than 500 cells/mm³), while 15 began with a median CD4⁺ count of 39 cells/mm³ (all had fewer than 100 cells/mm³). So this analysis is not as compelling as it may have been had it included a third group with a median count of, say, 200 cells/mm³. Still, the results are edifying.

Everyone in both groups reached and maintained a viral load below 200 copies/mL for 12 months. When treatment began, the 17 early-stage individuals had normal allotments of naive and memory CD4⁺ cells, and they stayed normal through 12 months of treatment. People in the late-stage group, however, did not regain normal consignments of CD4⁺ naive and memory cells after a year of suppressive therapy. Activated (CD38⁺) CD8⁺ cells fell to normal levels in the early-stage group, but remained high in the late-stage group, a finding implying ongoing replication below the 200-copy mark. Stores of naive CD8⁺ cells rose significantly in the early-stage group (P = 0.05), while the cache of memory CD8⁺ cells dwindled significantly (P = 0.004). In the early group, these subsets approached a normal distribution after 12 months of treatment, whereas they remained unbalanced in the late-stage group.

The one area in which late-stage individuals nearly matched their early- stage counterparts was in churning up a lymphoproliferative response (LPR) to cytomegalovirus (CMV) antigen. In the early-stage group, 40 percent had LPR responses to CMV after 12 months, compared with 33 percent in the late-stage group. Only one person in either group had a marginal LPR response to HIV p24 after treatment (a stimulation index of 4), and that person was in the early group. These results indicate, then, that beginning treatment very early restores some order to T-cell subsets, or preserves what's already there. Whether such an investment pays long-term clinical dividends that offset potential long-term treatment side effects cannot be determined by a study like this, or maybe by any feasible study.

Another nonrandomized study of 17 treatment-naive people who began a PI regimen at CD4⁺ counts above 500 cells/mm³ logged immunologic benefits with early HAART [abstract 330]. This report, by Lena Al-Harthi, PhD, in Alan Landay's group at Rush Medical College in Chicago, did find LPR responses to HIV p24 in 12 of these 17 people after 48 weeks of treatment. Indeed, five of the 17 had T cells that proliferated in response to p24 before treatment began. These findings contradict reports that HIV-specific CD4⁺ responses disappear after primary HIV infection in nearly everyone except untreated long-term nonprogressors. Louis Picker, MD, and his University of Texas colleagues also reported last year that they spotted plenty of HIV gag-specific CD4⁺ cells in slow progressors and in others with chronic HIV infection.¹⁵ So what's the story?

As Oxford's Andrew McMichael, PhD, proposed at the conference, the difference between Picker's findings and those by Harvard's Bruce Walker, MD, and others, probably lies in the assay [abstract L8]. Picker used an intracellular cytokine assay and flow cytometry--not a standard lymphoproliferative assay--to document that CD4⁺ cells can recognize HIV. "It looks like there are CD4⁺ T cells detectable in these intracellular cytokine assays that are specific for HIV," McMichael explained, "but there is some problem with the proliferation in those proliferative assays, which may or may not represent a real deficit in vivo." But that doesn't explain Al-Harthi and Landay's findings, because they used the standard lymphoproliferative assay. If Al-Harthi and Landay are right--and if those proliferative responses can't be provoked when treatment begins at a lower CD4⁺ count--this would be a potent argument for starting early.

On the other hand, G.P. Rizzardi, PhD, in Giuseppe Pantaleo's Lausanne lab, saw only evanescent LPR responses in a similar group of 41 early-stage treatment-naive individuals [abstract 336]. Treatment with abacavir and amprenavir for 72 weeks yielded big gains in CD4⁺ cells. But only 40 percent in this cohort had boosted LPR responses to envelope-depleted HIV and to p24. And the investigators labeled that increased response "transient." Together, these results make it hard to figure exactly what standard LPR assays are saying about people with chronic infection.

A study by Uma Malhotra, MD, in Julianna McElrath's group at the University of Washington, Seattle, found HIV p24-specific lymphoproliferative responses in only four of 46 people (9 percent) with acute HIV infection [abstract 610]. Among 18 who took ZDV, 3TC, and indinavir for 52 weeks, 16 (89 percent) could generate such responses, whereas none of 30 untreated primary infection patients could. The story differed with envelope proteins. Malhotra found env-specific responses in only two of 46 people (4 percent) with primary infection. But those responses gradually waned, despite antiretroviral therapy.

The University of Washington group also found that treatment of primary HIV infection did not completely balance an out-of-kilter T-cell receptor (TCR) repertoire. They measured disruption of 20 percent of TCR subfamilies at baseline in six individuals, and 10 percent of those subfamilies remained disrupted after six to 12 months of treatment. "Evolution to a broader repertoire may occur with treatment," Malhotra concluded, "but restoration of immunity [reflected in a treatment-induced median CD4⁺ count of 888 cells/mm³] does not imply immediate restoration of the repertoire."

This recently published work¹⁶ indicates that treating HIV infection even at the earliest point possible leaves chinks in the immune armor. Perhaps those holes can be plugged by strategies that complement HAART, such as planned treatment withdrawals or immunotherapies. But if immune restoration is so tough when treating people in the first year of infection, can it be any easier when treating more advanced disease? That, too, remains an open question, Malhotra and McElrath noted. "‘Very early' therapy may be most optimal for preservation of the repertoire," they wrote in their conference poster. "However, factors other than duration of infection may influence response to therapy."

Return to Article Contents

Make room for Markov

Long-term trials of early versus delayed treatment--which are once more being discussed--might help unravel the enduring riddle of when to start. Then again, they might not. Researchers say any such trial would have to enroll thousands of participants and last five or 10 years to yield meaningful answers. And who wants answers in 2005 or 2010 about drugs prescribed in 2000? So what's a clinical investigator to do?

Roll out a Markov model. That's the tack taken by Pablo Tebas, MD, and colleagues at Washington University in St. Louis, the University of Minnesota in Minneapolis-St. Paul, and Northwestern University in Chicago [abstract 523]. Yes, we know some clinicians love reading about Markov models only a little more than they love paying malpractice insurance, but this study is different. In this reporter's experience, it's the first Markov model report that actually explains what a Markov model is.

"A Markov model," we learn, "defines a finite series of ‘states' in which an individual can be found." In this study, those states involve response to antiretroviral therapy. A person with HIV infection starts in a treatment-naive state. After treatment begins, that person can have a detectable or an undetectable viral load (fewer than 50 copies/mL in this study). An undetectable viral load can stay that way or become detectable. When viral load becomes detectable, a second regimen can make it undetectable, or not. And so on up to a point--after three regimens in this analysis--where a person has multidrug-resistant virus.

Every transition from one state to another has a certain probability. To plug those probabilities into the model, Tebas and colleagues parsed results from a top-10 list of antiretroviral trials, including some golden oldies--Merck 035, INCAS--as well as more recent hits--SPICE, ACTG 343, Trilège. Then the St. Louis-Minneapolis-Chicago trio posed three questions:

1. What would be the virologic outcomes of a theoretical 10,000-patient population that starts therapy immediately after diagnosis?
2. What would happen if, instead of starting therapy immediately, patients started gradually when they reached a specific threshold?
3. If therapy begins immediately, how much better does the response rate have to be, compared with delayed therapy, to justify immediate treatment?

In answer to question 1, the model predicts that about 4000 of 10,000 people who begin treatment immediately will maintain a viral load below 50 copies/mL for 10 years on their first regimen. Drastically smaller proportions who move on to a second or third regimen have undetectable viremia after 10 years. But, overall, 5764 of 10,000 people are aviremic at year 10. This optimistic number shows that these modelers did not stack the deck against the durability of immediate treatment.

What happens if treatment is delayed, that is, if only 5 or 10 or 15 or 20 percent of infected people begin treatment immediately? The 5 percent analysis, which assumes that 5000 people never get treated, does worse than immediate treatment for the entire population: Only 3232 of 10,000 people have undetectable viremia at year 10. But if 10 or 15 or 20 percent, instead of everyone, start treatment immediately, the respective numbers undetectable 10 years later are 6464, 6331, and 6156. All of those rates easily outdo the 5764 below the sub-50 copy mark if everyone starts treatment immediately.

Finally, in answer to question 3, the model calculates that the response rate to immediate treatment must be 10 to 20 percent better than the response to delayed treatment to justify immediate intervention. That's a big gap.

Tebas and coworkers (who include Keith Henry and William Powderly) frankly admit that there's only so much you can take away from this type of analysis. The strongest statement these Markovians make is that their study "provides a theoretical basis for questioning the current aggressive early use of therapy and should help prompt studies that look at when and how to start antiretroviral therapy."

And they list the study's limitations. The analysis to answer question 2 assumes that the response rate of people who delay treatment is the same as if they start early. That's a big assumption, and perhaps an untenable one. The Meibohm and Moore studies at this meeting [abstracts 521 and 522 above], and scores of published clinical trials and cohort studies, show that people who begin treatment with higher T-cell counts and/or lower viral loads have better virologic responses.

Also, this Markov analysis does not consider "potential benefits of starting therapy immediately like eradication . . . or better preservation of the immune system." Although most are probably willing to write off the benefit of eradication as less than "potential" right now, better immune system preservation with earlier treatment remains a real possibility and the focus of much research.

The practical value of this analysis can also be questioned because many clinicians these days don't consider immediate treatment (except for people with acute infection) when someone has a reasonable and stable CD4⁺ count and a low viral load. So the "immediate treatment" in this comparison is something of a straw man.

But no one can nit-pick another conclusion advanced by Tebas and coworkers. "Delaying treatment," they note, "would decrease the total mount of drug used and the amount of exposure which could result in cost savings and less toxicity." And toxicity, right now, is the crux of the problem.

Return to Article Contents

Resistance and reservoir surprises

Three studies may give some clinicians pause when considering early intervention with today's most potent regimens. Research from two groups added to the horde of data showing that standard HAART doesn't quite grind out the last embers of replication and--this is new--those embers can spawn resistant virus. The third study suggested that the most-mulled reservoir of viable virus, resting CD4⁺ T cells, may not pose the biggest threat during interrupted HAART.

Until now, research had failed to spot evidence of resistance mutations that evolved in virus persisting below levels of detection in plasma. That has changed. Lisa Frenkel, MD, and coworkers at the University of Washington, Seattle, snooped for resistant virus in peripheral blood mononuclear cells (PBMCs) of 10 children six to 12 months before they started taking three, four, or five antiretrovirals, and periodically during 1.5 to 3 years of treatment [abstract 754]. All of the children maintained viral loads below 50 copies/mL during these resistance studies.

PI-associated resistance mutations emerged in all of the kids taking three antiretrovirals while their viral loads stayed under the 50-copy level. In three children taking four or five drugs, however, pre-HAART nucleoside-related mutations (L74V, M184L, and T215Y) reverted to wild-type. Frenkel proposed that this shift to wild-type in PBMCs during four- or five-drug HAART "suggests a loss of PBMCs that supported ‘recent' replication of PBMCs infected early" in the disease course. As for the new mutations in children treated with three drugs, they preceded rebounds above 50 copies/mL.

Richard D'Aquila, MD (Harvard) and colleagues from across the country scrutinized virus from PBMCs in six persons and from resting memory CD4⁺ cells in six others [abstract 238]. Five of the 12 people (10 were adults) had viremic blips above the 50-copy mark while taking ZDV, 3TC, and one or two PIs, and all five of them picked up new 3TC- or PI-linked mutations. Virus remained wild-type in the seven people whose viral load never nosed into detectable territory. In one of the five people with new mutations, phenotyping verified decreased susceptibility to the PIs being taken.

D'Aquila noted, though, that none of the five people with blips and new mutations endured a full-fledged viral rebound during six to 12 months of follow-up after the mutations were spotted. But he believes these findings "show drug selection pressure on residual replication." As a result, he gave a nod to the notion of "proactive switching," that is, changing a regimen before a full rebound gets a head of steam. Though D'Aquila stopped short of endorsing that strategy without reservation, he believes it should be studied.

These two studies differ in that resistance may well have emerged during the viremic blips in D'Aquila's cohort, whereas resistance emerged during "fully suppressive"--but obviously still inadequate--therapy in Frenkel's group. Frenkel's intriguing distinction between the children taking three antiretrovirals and those taking four or five hints that slightly more potent regimens may help slam the door on any viral evolution that can lead to resistance--at least in kids. Together the studies show that HIV can change its genetic stripes at vanishingly low viral loads, far below those that can be canvassed by current genotypic resistance assays.

Continuing study of people who interrupted treatment with HAART and IL-2 deepened the mystery of viral reservoirs that contribute to rebounds if treatment stops. Tae-Wook Chun, PhD, at the US National Institute of Allergy and Infectious Diseases (NIAID) reported that the kinetics of those rebounds in nine people did not correlate with the size of the pool of latently infected CD4⁺ cells, which he had painstakingly measured before stopping treatment [abstract 239]. Chun also compared env sequences from virus in the resting T-cell reservoir before treatment stopped and from virus sampled during the rebound. In four of six persons studied, the rebounding virus proved genetically distinct from the latent virus.

"The latently infected resting T cells do not account entirely for the early rebound in plasma viremia," Chun concluded. "This is not to say that this reservoir is not important. It's obviously an important reservoir, given that this reservoir can give rise to infectious HIV in vitro. Perhaps we need to pay more attention to other reservoirs of HIV throughout tissue compartments and peripheral blood."

Return to Article Contents

A blind clue to when to start?

Digesting data from studies addressing the when-to-start conundrum yields no easy answers. And for a good reason--there are no easy answers. But clinicians who find themselves rethinking this question because of the surging threat of side effects may do well to flip back through the hospital library files of Nature Medicine. Flip back five years. That's when Joseph Margolick, MD, from Johns Hopkins, espoused an intellectually appealing theory that rarely gets mentioned in when-to-start debates.

What Margolick found is that the so-called blind homeostasis of T cells fails at a generally consistent time before the onset of AIDS.¹⁷ Blind homeostasis simply means that the total number of T cells--CD4⁺s plus CD8⁺s--stays constant for years during HIV infection because, as CD4⁺ counts fall, CD8⁺ counts rise. But then total T counts begin to dive as CD8⁺ levels also start crashing. That turnaround, called the inflection point, precedes clinical AIDS by 1.5 to 2 years. And the inflection point has nothing to do with when a person became infected. Individuals A, B, and C can get infected 2, 5, or 10 years before blind homeostasis fails, and AIDS can (usually) be diagnosed in all three persons a little more than 1.5 years later.

Margolick figured that out by checking CD4⁺ and CD8⁺ counts of gay men enrolled in the Multicenter AIDS Cohort Study (MACS), the same bunch of guys whose frozen plasma samples clued John Mellors in to consistent rates of disease progression depending on viral load. But the blind homeostasis theory didn't catch on the way plasma RNA did as a benchmark for starting treatment.

Undaunted, the Johns Hopkins group and MACS collaborators from other sites haven't dropped the idea. At the 5th Conference on Retroviruses, they presented data showing that a more rapid fall in total T cells after homeostasis fails correlates with a quicker arrival of AIDS.¹⁸ And at this year's conference they spelled out results of a study correlating T-cell drops with changing viral load [abstract 200]. While homeostasis is preserved, they found by analyzing 841 pre-HAART plasma samples, RNA levels climb at a rate of 0.29 log (about twofold) annually. The T-cell inflection point came at a viral load of about 5 logs (100,000 copies/mL). After homeostasis failed, viral load stopped rising at such a fast clip, settling back to a yearly gain of only 0.06 log (about 1.15-fold).

Is blind homeostasis the neglected key to when treatment should start? Of course not, although it may be tempting to conclude that one could periodically measure CD4⁺ and CD8⁺ counts, then start antiretrovirals when the total count begins its slide from the inflection point. In theory, one would still have more than a year before the first AIDS-defining symptom appears, even if the viral load stood at a lofty 5 logs.

Readers of this Journal know, however, that they can't pin all their treatment plans on one number--whether it's this inflection point, or 200 or 350 CD4⁺cells/mm³--because some people will fall outside expected progression ranges, some people will want to start earlier, and so on. But Margolick's well-researched theory gives clinicians and people with HIV another--possibly powerful--signal of when to start.

Return to Article Contents

The Essence of the STI

No topic attracted more attention at the 7th Conference on Retroviruses than STIs--structured treatment interruptions. No surprise there. At this pass in the epidemic's evolution, what strategy exudes more allure? Here are The certain benefits of STIs

1. The person being treated gets to stop gut-wrenching drugs.
2. The person treating gets to stop giving lectures on adherence.
3. The former treatee feels better almost immediately.
4. The former treater feels the warm glow of gratitude.

Beyond that, the benefits become less certain, though possibly even more alluring. Stopping treatment--in a structured way with scrupulous follow-up that permits quick restarts--may actually reacquaint CD4⁺ T-helper cells with HIV and jazz up those lackadaisical lymphocytes (CTLs) intended to kill infected cells.

But it's much too early to figure whether STIs will live up to their considerable advance billing. Although several studies of STIs during successfully treated chronic infection offered a smattering of hopeful results, the verdict was hardly unanimous (Table 5). And two looks at people who stopped treatment during failing HAART raised several cautionary flags. The best news came from two tiny studies of STIs during successfully treated primary infection.

Return to Article Contents

Stopping treatment that's still working

The largest trial in this category has results on 96 people who stopped treatment with CD4⁺ counts above 300 cells/mm³ and viral loads below 50 copies/mL after taking HAART for about two years [abstract 458]. The plan calls for enrolling 200 people who have never endured a HAART failure and who will stop treatment for two weeks four times, followed by eight weeks back on therapy after each stop. Finally, they will abandon HAART until their viral load clambers back over the 5000-copy mark. One potential shortcoming of this strategy, noted by the University of North Carolina's Joseph Eron, MD, is that the two-week interruptions may not be long enough to jar the immune system into action.

Bernard Hirschel, MD (University Hospital, Geneva) presented results on 96 people who've had one STI, 54 who've had two, and 23 who've had three. With each successive treatment interruption, more people are showing HIV-specific CD4⁺ responses by two assays, but the number involved is still small (Table 5). And despite this apparent riling of T-helper cells, the study team saw no clear trend toward broadening CD8⁺-mediated immunity. Neither have investigators picked up any hint of rebound blunting so far. Hirschel proposed that STIs will induce remissions in only a minority of those who try this tactic. Still, the eventual size of this study raises the prospect that it will answer some questions with authority.

Franco Lori, MD (RIGHT, Pavia and Washington, DC) cajoled nine members of his PANDA cohort to try an STI, and he compared them with eight people who took a break from suppressive PI regimens [abstract 352]. The PANDAs have been taking ddI plus hydroxyurea, and nothing else, for over two years and have maintained low but usually detectable viral loads.¹⁹ Viral loads in the nine PANDAs in this study ranged from fewer than 50 copies/mL to 1500 copies/mL. Six of eight people in the PI group had viral loads under 50 copies/mL for 67 weeks or more, and the other two had viral loads below 400 copies/mL.

Ten PANDAs had already tried an eight-week STI. Their viral loads peaked six weeks into the STI, then fell without the help of drugs. The nine PANDAs who quit ddI/hydroxyurea in the PI comparison study saw their viral loads rebound to an average 4000 copies/mL two weeks into the STI, then drop spontaneously to around 2000 copies/mL. The eight people who stopped their PI-based HAART all had progressively rising viral loads with no spontaneous drops. At the eight-week mark, the average viral load increased 0.27 log above baseline in the ddI/hydroxyurea group and 2.25 logs above baseline in the PI HAART group. Viral loads of everyone in both groups slipped to pre-STI levels when treatment resumed.

These results will encourage those who argue that the best antiretroviral regimen controls viremia at a modest but detectable level and so constantly tickles immune cells with a trickle of measurable replication. HAART regimens that push viral loads under 50 copies/mL or lower, this argument goes, may keep virus from immune sentinels for too long.

In this presentation Lori didn't report that T cells began recognizing HIV during treatment with or interruptions of ddI/ hydroxyurea, or that a cytotoxic T lymphocyte (CTL) response kicked in. He did report apparently maturing CD4⁺ responses to HIV during ddI/hydroxyurea treatment in an earlier report on this cohort.¹⁹ If such responses can be tied conclusively to long-term control of viremia with ddI/hydroxyurea in these people, Lori will add to questions about whether everyone needs to start treatment with HAART that includes a PI or an NNRTI.

On the other hand, a selection bias may have favored the ddI/hydroxyurea group in this comparison. Certainly not everyone responds to ddI/hydroxyurea as well or as long as members of the PANDA cohort. So that group may have had an advantage over the PI comparison group because they had already been "selected" as longer treatment responders. They had an average of 184 weeks of treatment under their belts, versus 106 weeks in the PI arm.

A study by clinicians in Barcelona and colleagues in Lausanne and New York found that viral doubling times increased during three interruptions of PI triple therapy, from 2.0 days, to 2.9 days, to 3.2 days (P = 0.03) for the first versus the last doubling time) [abstract LB11]. Four of nine people who completed two STIs had a spontaneous decrease in viral load during the second interruption. Levels of HIV-specific CTLs and HIV-specific CD4⁺ cells correlated significantly with these spontaneous drops (P = 0.05 and P = 0.001, respectively), a result suggesting that rejuvenated T cells helped control viremia. The glass-half-empty view of these results is that most study participants did not enjoy spontaneous control of replication during STIs.

Another Barcelona group reported that only two of 12 people with a HAART-induced viral load below 50 copies/mL for 24 months or more generated CD4⁺ cells that recognized HIV during the first of two STIs [abstract 354]. And those T-helper responses vanished when HAART resumed. The average time to a viral load rebound above 20 copies/mL measured 14 days during the first STI and barely increased, to 15 days, during the second STI. Adding IL-2 to HAART for five of 12 persons after the first STI made no difference in rebound rate.

Researchers in the lab of Brigitte Autran, MD, PhD (Pitié-Salpêtrière Hospital, Paris) offered the least sanguine report on STIs during HAART-treated chronic infection [abstract 356]. They studied three people whose viral load stayed below 20 copies/mL and whose CD4⁺ count stayed above 400 cells/mm³ for more than two years while taking a PI and two NRTIs. The Paris team stopped treatment for seven to 21 days three to four times in each person, generally restarting when the viral load edged over 1000 copies/mL.

One person never generated HIV p24-specific CD4⁺ cells during three seven-day interruptions and never had a viremic rebound. This person's fourth STI was allowed to continue until viral load rebounded, but HIV-specific CD4⁺ cells still could not be detected. The two other individuals had rebounds to 1000 copies/mL during three or four STIs that lasted seven to 18 days. The first of these two individuals hatched some HIV p24-specific CD4⁺ cells during each STI, but those cells always disappeared even before HAART resumed. The second of these two persons produced HIV-specific CD4⁺s only during the first STI.

Although treatment interruptions can produce CD4⁺ cells that recognize HIV, Autran and colleagues concluded, those cells "were rapidly destroyed by virus replication and did not help expand CD8⁺-T cell responses to HIV" during the rebounds. That failure "raises concerns about STI strategies aiming to restore HIV-specific T-cell responses during chronic HIV infection."

Autran's poor results with three or four successive STIs and the second Barcelona group's difficulty in eliciting HIV-specific responses [abstract 354] certainly show that two or three or four STIs, at least brief ones, won't work for everyone. If the large Swiss study confirms these clues that STIs help only a handful of people with chronic infection, STIs will be called drug holidays again, and they will be invoked mainly to give people a break from side effects.

Even if that happens, though, many will surely abide by the precept that stopping treatment is almost never a good idea. The thinking here is that relieving pressure on the virus allows it to begin repopulating reservoirs from which it may be draining or that resistance may emerge if certain components of a regimen stay in the body when treatment stops because those drugs have long half-lives. Professors of this school of thought could cite three conference studies to bolster their claims. The two summarized in the next section both found that intermittent suppression of viremia resulted in significantly lower CD4⁺ improvements over time than if viral load always stayed under 500 copies/mL. The third study involved a replenished reservoir (see abstract 306 on page 92).

Table 5. How well are structured treatment interruptions (STIs) working so far?

Lead author, site [abstract number]	No. of participants; baseline values*	STI strategy*	Follow-up	Pluses	Minuses	Comments, conclusions
STI during successful treatment of chronic infection
C. Fagard Switzerland and Spain (SITT study) [458]	105 so far, 200 planned; median CD4+ 740, all > 300 (n = 97); VL < 50 (80 <10); mean 24 m on HAART	In people with <50 copies/ mL on first HAART (without 1st STI, NNRTI), stop HAART for 2 wk, treat for 8 wk, for 4 cycles; then stop HAART until VL > 5000	96 completed 1st STI, 54 completed 2nd STI, 23 completed 3rd STI	HIV-specific CD4+ response by Elispot in 1/5 at 2nd STI, in 13/13 at 3rd STI; HIV-specific CD4+ response by SI in 0/19 at 2nd STI, in 9/19 at 3rd STI	No clear tendency to broadening of CD8+- mediated immune response; no clear trend to in VL rebounds so far	Fixed 2-wk STI may not permit maximum stimulation of antibodies
F. Lori Italy (PANDA study) [352]	9 on ddI/HU vs. 8 on PI-containing HAART; mean 184 wk on ddI/HU vs. 106 on HAART; mean CD4+ 495 on ddI/HU, 560 on HAART; mean VL 549 on ddI/HU, 97 on HAART	Stop ddI/HU or HAART for 8 wk or when VL to >10,000 or CD4+ to < 200	8 wk	0/9 on ddI/HU, vs. 4/8 on HAART, had to restart Rx before 8 wk; after initial VL during STI after ddI/HU, VL spontaneously ; no spontaneous after rebound in HAART group; VL to pre-STI levels in all in both groups when Rx resumes	No evidence of viral control during 1st STI after 67 to 139 wk on HAART (see "Pluses" column); CD4+/CD8+ratio in HAART group but not in ddI/HU group during STI (P = 0.01)	STI feasible in people with chronic infection; selection bias may favor ddI/HU arm
F. Garcia Barcelona [LB11]	10 on d4T/3TC + RTV or IDV; VL <20 for >8 m; CD4+ >500	STIs at wk 0 and 28, restart if VL > 200 or after 4 wk; follow-up after 3rd STI (1 lost to follow-up after 1st STI)	>12 wk after 3rd STI	Viral doubling time 2.0 d at 1st STI, 2.9 d at 2nd STI, 3.2 d at 3rd STI (P = 0.05 3rd vs. 1st STI); spontaneous in VL in 4/9 during 2nd STI; VL set point <baseline in 6/7 12 wk into 3rd STI; CD4+ and CTL LPR to HIV in 5/7 after 3rd STI; no resistance mutations emerged during STIs	No spontaneous in VL in 5/9 during 2nd STI	HIV-specific CTLs (P = 0.05) and HIV-specific CD4+s (P = 0.001) strongly correlate with spontaneous in VL during STI in those who had a spontaneous in VL
L. Ruiz Barcelona [354]	12 with RNA < 50 for >24 m; CD4+ 1202; CD4+/CD8+ ratio 1.4 vs 14 who continued HAART	STI for 30 d or if VL >3000; resume HAART for 3 m in 7, HAART + IL-2 in 5; then STI for 30 d or if VL to >3000	Through resumed HAART after 2nd STI	2 had no VL rebound >20 during 1st STI, 1 had no rebound during 2nd STI; resumed HAART suppressed VL in all; no resistance mutations emerged; CD4+ LPR to recall antigens in 4 of 12 during 1st STI	Delay to VL rebound did not change from 1st STI (14 d) to 2nd STI (15 d); IL-2 after 1st STI did not affect rebound rate; CD4+ LPR to p24 in only 2 of 12 after 1st STI, not sustained when HAART resumed	"HIV-specific helper T-cell responses may require subsequent cycles of STI to keep viral replication under control."
E. Papasavvas, Philadelphia [353]	5 with VL <50 on ddI/d4T or PI-containing HAART vs. 5 untreated HIV-positive controls	STI with close follow-up and tracking of HIV-specific CD4+ and CD8+ responses	Variable; STI >118 d in 1/5; 4/5 resumed HAART	SI of CD4+ to HIV p24 1.82/wk (P = 0.001); median 2.7 in SI from baseline in 21-33 d; CD4+ response followed by in IFN--secreting CD8+ against env antigens	CD4+ and CD8+ responses "variably preserved" after resuming Rx	CD4+- and CD8+- mediated responses against HIV improved by exposure to virus during STI; in 2 with longest STIs, spontaneous control of VL <1080 copies/mL
G. Carcelain Paris [356]	3 with VL <20 and CD4+ >400 for > 2 y with 1 PI + 2 NRTIs	STI for 7 d, then 2 m of HAART, then 2 or 3 STI until VL >1000	Through resumed HAART after 3rd or 4th STI	Resumed HAART suppressed VL in 3/3	1 person had no VL rebound or HIV-specific CD4+ LPR in 3 7-d STIs, VL rebound in 4th STI but no CD4+ LPR; 1 person had small VL rebounds in 3 STIs with HIV-specific CD4+ LPR; 1 person had small VL rebounds in 4 STIs but HIV-specific CD4+ LPR only in 1st STI; no HIV-specific CD8+ LPRs in 3/3	"HIV-specific [T-helper type 1 CD4+] cells producing IFN-g in chronically infected patients . . . were rapidly destroyed by virus replication and didn't help expand CD8+ T cell responses to HIV during the virus rebounds. . . . this raises concerns about STI strategies aiming to restore HIV-specific T-cell responses during chronic HIV infection."
Lead author, site [abstract number]	No. of participants; baseline values*	STI strategy*	Follow-up	Pluses	Minuses	Comments, conclusions
STI after successful treatment of primary infection
M. Altfeld Boston [357]	7 with VL control for 370 to 685 d of HAART begun during primary infection; CTL frequencies with HAART and dwindle slightly with continuing Rx	STI after >1 y of VL control during Rx of primary infection; 2 STIs in 2 persons, 3 STIs in 1	>100 d after 1st STI	In 1st STI, VL rebound in 7/7 led to significant, persistent in magnitude and breadth of CTL response; VL to below limit of detection in 7/7 when HAART resumed; spontaneous control of VL to <5000 to 10,000 for >17 wk in 3/3 during 2nd and 3rd STIs		STIs breadth and magnitude of CTL response after VL controlled > 1 y by HAART given during primary infection; better CTL responses associated with spontaneous control of VL in 3/3
X. Jin New York [346 and LB12]	4 with VL control for 2.5 y	After VL <50 for >2 y of HAART started within 90 d of primary infection, give canarypox vaccine with gag, pol, env, and nef genes + gp160 vaccine 4 times; STI 1 wk after last vaccine	>4 m after STI	Vaccinations arouse CTL responses against >1 HIV antigen in 2/4, and CTL responses against only 1 antigen in 2/4; 2 with broader CTL responses have longer time to VL rebound (68 and 85 d) than 2 with narrow CTL responses (13 and 23 d)	8 vaccinations elicit broad CTL responses in only half of those studied so far	After STIs of 4 to 8 m, VL at 2.52 and 3.75 logs in delayed rebounders and 3.55 and >4.7 logs in rapid rebounders; HIV vaccination during HAART begun early in infection produces different responses in different individuals
Lead author, site [abstract number]	No. of participants; baseline values*	STI strategy*	Follow-up	Pluses	Minuses	Comments, conclusions
STI after virologic failure of HAART
S.G. Deeks San Francisco [LB10]	17 on failing PI HAART; CD4+ 245, VL 4.6 logs; on PI 36 m; VL failure for 31 m; median 56-fold in susceptibility to PIs	STI with close monitoring; patients encouraged to restart Rx for VL to >1 log or CD4 of 50	36 wk	Reversion to PI susceptibility in 16/17 in wk 6 to 11 of STI	No return to NRTI susceptibility in 7; median CD4 94; median VL 0.82 log in 12 wk of STI; resistant virus persists in PBMCs in 4/8 after wild-type reversion in plasma 12 to 36 wk after STI	Rapid in VL and in CD4+ during STI just after shift to wild-type; replication capacity of virus after shift to wildtype; "antiretroviral therapy in long-term virologic failure is associated with continued virologic and immunologic benefits . . . due in part to residual antiviral activity of the drugs the patients are taking, despite high-level resistance"
R.W. Price San Francisco [306]	5 on failing PI HAART in Deeks study (above), 1 observed during change to new Rx; median CD4+ 179.5, mean plasma VL 4.2 logs, mean CSF VL 2.06, plasma/CSF ratio 2.01	STI with lumbar puncture within 2 wk before STI then ~ 3-6 wk during 12-wk STI	Through 12-wk STI	2 had parallel in CSF and plasma VL during STI, and plasma/CSF ratio remained constant	4 had >100-fold in CSF VL--much greater than in plasma VL--and plasma/ CSF ratio narrowed as a result; 3 had abnormal but asymptomatic in CSF WBCs	"These observations may have broad pathogenic implications in showing that STI is accompanied by heretofore unrecognized changes in viral exposure and cell traffic patterns."
*CD4+ counts in cells/mm3; plasma viral loads (VL) in copies/mL. CSF = cerebrospinal fluid; CTL = cytotoxic T lymphocytes; HAART = highly active antiretroviral therapy; HU = hydroxyurea; IDV = indinavir; LPR = lymphoproliferative response; NNRTI = nonnucleoside reverse transcriptase inhibitor; NRTI = nucleoside reverse transcriptase inhibitor; PI = protease inhibitor; RTV = ritonavir; Rx = treatment; SI = stimulation index; VL = plasma viral load; WBCs = white blood cells.

Return to Article Contents

Is continuous suppression essential?

Jean-Pierre Aboulker, MD, and colleagues who worked on the Trilège induction-maintenance trial prospectively compared 217 people whose viral load remained below the 500-copy mark through this 72-week study with 136 people whose viral load rose above that mark one or more times [abstract 343]. The study participants whose viral loads meandered into 500-plus territory were those randomized to a two-drug maintenance arm after induction with ZDV, 3TC, and indinavir.

When it became clear that maintenance wasn't working, those people all restarted triple therapy and resuppressed their virus at different points of follow-up. To that extent, the viral loads of the maintenance group mimicked rebounding loads of people during an STI, even though no one in the Trilège study stopped taking drugs completely, so HIV wasn't let completely off the leash.

Over the 72-week follow-up, mean CD4⁺ counts rose by 0.31 cell per day in the group with continuous viral suppression, compared with 0.18 cell per day in the intermittent suppression group, a highly significant difference (P < 0.0001). At those rates, total CD4⁺ gains during Trilège came to 221 cells/mm³ in the full-suppression group and to 154 cells/mm³ in the intermittent-suppression group (P < 0.001). Levels of activated (HLA DR⁺) CD8⁺ cells remained stable in the 217 people who maintained viral suppression but rose by 101 cells/mm³ in the 136 with interrupted suppression (P = 0.02). Aboulker and coworkers believe these results "suggest that a permanent control of HIV replication could be necessary for faster immune reconstitution." But results might differ in people whose on-treatment CD4⁺ gain has hit a plateau.

Benita Yip, MD, and colleagues in Julio Montaner's Vancouver group addressed the question of incomplete suppression in a retrospective population study [abstract 335]. She compared viral loads with CD4⁺ counts in 465 adults who began antiretrovirals in British Columbia between August 1996 and May 1998. The study analyzed three groups: 112 nonresponders whose viral loads always stayed above 500 copies/mL after treatment started; 100 partial responders whose viral loads fell below that mark at least once, then climbed above it at least once; and 253 full responders whose viral loads sank below 500 copies/mL and stayed there. The median follow-up for the entire group measured 64 weeks.

Yip figured the median viral load changes throughout the observation period at -0.37 log in the nonresponders, -2.27 logs in the partial responders, and -2.56 logs in the full responders. Two-way differences between each of the groups were all statistically significant. Absolute CD4⁺ changes from baseline and CD4⁺ percentage changes from baseline through the end of observation also differed significantly in every two-way comparison. For nonresponders, partial responders, and full responders, median CD4⁺ changes were -20 cells/mm³, +150 cells/mm³, and +240 cells/mm³ (P = 0.003 for the difference between partial and full responders). Respective median CD4⁺ percentage changes measured 14 percent, 57 percent, and 104 percent (P = 0.042 for the difference between partial and full responders).

The Vancouver investigators also found a negative correlation between overall CD4⁺ changes and viral load in all three groups, with correlation coefficients of -0.44 (P < 0.001) for nonresponders, -0.31 (P = 0.002) for partial responders, and -0.16 (P = 0.011) for full responders. They argued that their results support the contention that "maximal suppression of viral replication should remain as the primary goal of therapy."

Neither study, of course, set out to probe the impact of intermittent viremia and CD4⁺ plummets during STIs. But it would take a long, randomized comparison of STIs versus uninterrupted therapy to determine what dangers may lie in allowing replication to go uncontrolled for weeks or months when continued treatment could keep the virus under wraps. Lacking such a trial, clinicians will have to base their opinions on any evidence that may arise, tangential or otherwise.

Richard Harrigan, MD, and the Vancouver group offered another retrospective study aimed a little more directly at showing what may--or may not--actually happen during an STI. Specifically, they wondered, how common is undetectable viremia after people stop taking antiretrovirals? Though it doesn't happen every day, they reported, so-called spontaneous plunges in viral load are not as rare as relics of the true cross.

Harrigan and coworkers reached that conclusion by once more shuffling through records of HIV-positive people in British Columbia, all of whom receive their antiretrovirals through the British Columbia Centre for Excellence in HIV/AIDS [abstract 351]. This analysis began with 238 people in the 1958-person database who stopped antiretroviral therapy for 90 days or more for any reason. Thirteen of these 238 (5.5 percent) had viral loads that stayed below 500 copies/mL for 90 days or more after they quit taking their drugs. All 13 had positive ELISAs and Western blots, and only two had been treated during primary infection.

None of these people took anything that might be mistaken for HAART. The zestiest regimen was ZDV, ddI, and 3TC. Five people had taken ZDV monotherapy, and the other seven had taken a dual-NRTI combo. Time on treatment ranged from 30 to 1987 days, and most of these lucky 13 still had viral loads below 400 copies/mL the last time anyone looked.

The investigators concluded that about one in 20 people has a viral load set point near the detection limit of standard assays. Treatment during primary infection may explain low set points in some people, but certainly not in everyone. "Isolated cases of undetectable viral load after treatment is stopped," they added, "may not be evidence of a unique therapeutic effect." One can assume that a small but certain percentage of "responders" in STI studies simply share the good fortune of these Canadians.

Return to Article Contents

Stopping after HAART flops

The notion that STIs may pay dividends when people quit a failing regimen emerged only a year ago, when Veronica Miller, PhD (JW Goethe University, Frankfurt) showed that treatment breaks fostered a return of wild-type virus in some people, and that those people got more oomph from a combination started after the break.²⁰ That idea got a closer look by Steven Deeks, MD, and colleagues at San Francisco General Hospital and the Gladstone Institute [abstract LB10].

Whereas Miller's cohort consisted of people who decided on their own to ditch their drugs, Deeks offered people that enticing option so that he could study them prospectively. He rounded up 17 folks with viral loads above 2500 copies/mL for one year or more while taking a PI combo. Their average CD4⁺ count stood at 245 cells/mm³ before the STI, and the group viral load measured 4.6 logs (about 40,000 copies/mL). They had been taking one PI or another for about three years. When Deeks phenotyped viral samples before the STI, he found an average 56-fold loss of susceptibility to the PIs being taken.

Then everyone stopped taking their drugs. Between 5.8 and 11 weeks later, and abruptly in each person, viral samples again became susceptible to PIs. Only one of 17 people failed to achieve this shift to sensitivity. But virus remained resistant to NRTIs in seven people. An assay that measures replication capacity of virus assigned a 22.3 percent fitness rate (compared with wild-type virus) at baseline. In other words, the viral population in these people during poorly suppressive PI therapy replicated sluggishly, only about one quarter as well as virus that has never seen an antiretroviral. Fitness improved to 67.1 percent 12 weeks into the STI (P = 0.004 versus baseline fitness).

During this 12-week upheaval in fitness and susceptibility, some more familiar measures were also taking wild swings. The median CD4⁺ count swooned by 94 cells/mm³, confirming a similar plummet in Veronica Miller's study. The median viral load jumped 0.82 log (about 6.6-fold). Twenty-four weeks after the shift to sensitivity, plasma virus remained susceptible to PIs. But virus extracted from PBMCs was genotypically identical to virus during the last days of PI treatment in four of eight persons studied. So resistant virus persists in blood cells of some people long after PI pressure is relieved. These investigators did not determine how often resistance persists in deeper reservoirs.

The quick viral load rise and CD4⁺ drop after treatment stopped led Deeks to conclude that "antiretroviral therapy in long-term virologic failure is associated with continued virologic and immunologic benefit . . . due in part to residual antiviral activity of the drugs the patients are taking, despite high-level resistance." In his presentation abstract, Deeks added his appraisal of well-planned holidays: "The safety and efficacy of STIs as a therapeutic strategy remain to be determined."

(In a 1170-person EuroSIDA cohort analysis, Veronica Miller showed that failing PI therapy has not only virologic and immunologic benefits, but a clinical payoff as well [abstract 454]. She counted 258 new AIDS events or deaths for an incidence of 44.3 per 100 person-years of follow-up in individuals with CD4⁺ counts below 50 cells/mm³. The event rate differed starkly between people still taking a PI regimen (37.5) and those who weren't (68.4, P = 0.001). Miller calculated a 43 percent reduction in risk of progression or death, independent of viral load or CD4⁺ count, for people taking a PI regimen, even though the regimen failed to lift the CD4⁺s above the 50-cell mark. That enduring benefit, she noted, must be "weighed against the risks associated with further viral evolution.")

Richard Price, MD (University of California, San Francisco) turned his attention to viral loads in cerebrospinal fluid (CSF) of five people in Deeks's STI study, plus another individual who agreed to a spinal tap during a treatment break [abstract 306]. In two of the six, CSF and plasma loads rose in tandem during the STIs. But four people had more than 100-fold leaps in CSF load, substantially more than their plasma load gains. Three of these four also had lymphocytic pleocytosis--a rush of white cells into plasma. That abnormality remained asymptomatic.

STIs, Price warned, can be "accompanied by heretofore unrecognized changes in viral exposure and cell traffic patterns." It will be interesting to learn whether interruptions of suppressive HAART, rather than the failing regimens these people were taking, result in the same reflooding of the CSF reservoir and other viral strongholds.

Return to Article Contents

The primary time for an STI?

The best STI results at the meeting came from studies of people whose HAART started during primary infection, although both studies are small and the clinical impact of these results cannot be judged.

Marcus Altfeld, MD, in Bruce Walker's Massachusetts General group detailed findings on seven people whose primary infection yielded to HAART given for periods ranging from 370 to 685 days [abstract 357]. HAART increased CTL frequencies in these people, but those levels dwindled somewhat during 12 months of treatment. Viral loads rebounded in everyone during the STI, but the viremic burst significantly enhanced CTL frequencies (P = 0.007 when treatment started again), and the breadth of these responses widened. These higher frequencies persisted for the 100 days of follow-up reported so far after HAART resumed (P = 0.0005).

Two of these seven individuals have had two STIs, and one has had three. Viral loads remained between 5000 and 10,000 copies/mL for 17 weeks or longer during the second or third STI in all three of these study participants. The Massachusetts General team attributed this control of replication without drugs to the burgeoning CTL responses. Two of the three had resumed treatment while the third remained in his second STI at the time of this report.

Altfeld and his colleagues cited no fewer than six studies linking CTL activity with decreased viremia or nonprogressing HIV disease. If that association holds up in this primary infection cohort, some of these individuals could enjoy--let's be cautious--long drug-free periods.

Xia Jin, MD, and colleagues at New York's Aaron Diamond AIDS Research Center pushed the STI envelope a bit farther [abstracts 346 and LB12]. After HAART kept viral loads below 50 copies/mL for more than two years in eight people treated within 90 days of infection, these investigators gave everyone four shots of two vaccines--vCP1452, a canarypox vector carrying snatches of four HIV genes, and a recombinant gp160 (envelope) vaccine. The eight shots stirred up CTL activity in four of six people whose results Jin presented in a poster [abstract 346]. Two had increased CTL activity against more than one viral antigen, and two had activity only against gag. These four individuals volunteered to stop their antiretrovirals [abstract LB12].

Two of the four had rapid viral load rebounds, 13 and 23 days after stopping treatment. Viral doubling time in these two individuals was a quick 1.4 days. But virus didn't bounce back in the other two people until 68 and 85 days into the STI. Their respective viral doublings times were more languorous: 4.5 and 3.2 days. And these two slow rebounders were the same two with CTL responses to more than one viral antigen. CTLs in the quicker rebounders had perked up only when prodded with gag.

At the time of Jin's presentation, the STIs of the slow rebounders had stretched to four and eight months, and their viral loads lingered at 3.75 logs and 2.52 logs (about 5600 and 330 copies/mL). Viral loads of the two quicker rebounders, both four months into their STIs, stood at 3.55 logs and more than 4.7 logs (about 3500 and 50,000 copies/mL).

The Aaron Diamond team issued an appropriately restrained conclusion, noting only that "therapeutic vaccination can enhance cellular immune responses that are temporally associated with suppressed HIV-1 rebound kinetics after discontinuation of HAART." Even in a small study like this, however, it's clear that different people respond differently to immune boosts like vaccines and STIs.

Return to Article Contents

. . . And Much More

Although the four issues just reviewed--adherence, side effects, when to start, and when to stop (with an STI)--accounted for the lion's share of controversy at the Conference on Retroviruses, they hardly monopolized the slide sessions or poster halls. This seventh edition of the meeting featured about 100 more presentations than last year's get-together--and not because the scientific committee felt extra nice this year. Indeed, 40 percent of poster submissions got turned down.

Return to Article Contents

Phenotypos

The most interesting other clinical research reported at the meeting involved phenotypic resistance testing. Results of the two resistance studies, both using the Virco phenotyping assay, reached different 16-week results. But that was not a surprise. One trial enrolled people in whom only one PI regimen had failed [abstract 237], whereas three quarters of participants in the other study had used two or more PIs [abstract 786].

In the successful study, presented by Calvin Cohen, MD (Community Research Initiative, Boston), 38 percent whose clinicians planned their next regimen with the help of phenotypic results reached a viral load below 400 copies/mL in a 16-week intent-to-treat analysis, compared with 23 percent in the standard-of-care arm (P = 0.02). In an on-treatment analysis, the proportions with sub-400 viral loads at 16 weeks were 58 percent with phenotyping and 37 percent with standard of care. The poor overall 16-week results are surprising because healthy majorities of people in both arms had three or more active drugs available to them when they traded regimens.

In the other study [abstract 786], only 20 of 61 people in the standard-of-care arm and 30 of 54 in the phenotyped arm had three or more active drugs available. Sixteen weeks after the new regimen started, investigators counted virologic failures in 69 percent in the standard of care arm and 85 percent in the phenotyping arm. They defined failure as (1) not attaining at least a half-log drop in viral load, (2) a half-log rise from the lowest viral load attained in the study, or (3) return of viral load to baseline.

Even among people who had three or more active drugs at their disposal, though, the 16-week virologic failure rate stood at 70 percent. Why such an unhappy result? D. Melnick, MD [Kaiser Permanente, Springfield, Virginia] and colleagues from Virco and Glaxo Wellcome justly blamed "the limited ability of this assay to detect ‘archived' resistant virus or minority species in plasma." The same phenomenon may explain the modest virologic success in the other trial.

Anyone needing more convincing that phenotyping can be tricky had only to ponder posters displayed by Abbott's Dale Kempf, PhD [abstract 731] and Glaxo's Randall Lanier, PhD [abstract 788]. Both showed that the numbers used to label virus susceptible or resistant to certain antiretrovirals can be misleading.

Kempf analyzed 50 percent effective concentrations (EC₅₀s) among people who began taking ABT-378 plus low-dose ritonavir after a PI failure [abstract 731]. The 48-week virologic response rates turned out to be as high among people with "reduced susceptibility" to ABT-378 at baseline (<4-fold higher EC₅₀) as among those with phenotypically sensitive virus (<4-fold higher EC₅₀). The reason, Kempf maintained, is that mean plasma levels of ABT-378 stay more than 30-fold above the EC₅₀ for wild-type virus. So modest drops in susceptibility (that is, modest increases in EC₅₀) do not rile this drug.

Lanier used the ViroLogic assay and the Virco assay to rate susceptibility to abacavir in viral samples of people enrolled in four trials of this NRTI [abstract 788]. Among people with only the M184V mutation, most responded virologically to abacavir when either phenotyping test averred intermediate or decreasing susceptibility to the drug. "Clinically relevant cutoff values for individual antiretrovirals," Lanier concluded, "must be established before phenotypic resistance tests can be most effectively incorporated into routine clinical practice."

Return to Article Contents

Changing opinion

The American composer Philip Glass should be commissioned to write the theme song for antiretroviral research. It's not just that Glass came to fame as a so-called minimalist whose simple-sounding music grows ever more complex.²¹ It's not that some describe his style as "monotonous and repetitive,"²² then suddenly propelled by "startling rhythmic changes." It's not even that he penned the music for a song named "Changing Opinion" in his cycle "Songs From Liquid Days."²³

The real reason Glass merits the commission is that his music can make sounds that the notes don't immediately account for. Describing the rehearsal of a piece that relies on his trademark techniques, additive process and cyclic rhythms,²¹ Glass revealed how he discovered this eerie arrival of unanticipated sonics. "During rehearsal we noticed sustained sounds emerging from the repeated beats of the music," he explained. "I stopped the music and asked who was singing. And nobody was singing; it was just a psychoacoustical effect of the music."²¹

People who practice HIV medicine should keep an ear cocked for psychoacoustical effects. They're real, all right. But they're based on the coincidence of circumstance. With Glass's music, overlapping, intersecting rhythms are the circumstances that produce these tones. With HIV medicine, overlapping, intersecting rhythms also make unexpected music.

A fortissimo example at the Conference on Retroviruses came as the insistently intrusive beat of antiretroviral side effects played over the dulcified tones of suppressed viremia and disappearing opportunists. The psychoacoustical effect was a warmer embrace of therapeutic caution.

And that embrace seems sensible, maybe even essential. As St. Paul clinician Keith Henry argued in his dissertation on antiretroviral guidelines, flawless adherence can prove too high an expectation for people confronted with a decades-long treatment course.¹⁰ Doesn't it make more sense, he asked, to spend these uncertain, liquid days working on adherence and waiting for less complicated regimens that use more tolerable drugs?

That may be exactly the right tack, as long as clinicians remember that the voice telling them to wait is psychoacoustical. Part of the interlocking rhythm that produces this voice is long-term viral suppression--and its consequent clinical benefits--which many people can attain with regimens already available. The 42 inmates (of 42 studied) who kept their viral loads under 400 copies/mL for 48 week with directly observed HAART in Margaret Fischl's study [abstract 71] show how potent today's treatments are. And there are plenty of people not living in penitentiaries and not on DOT who can tell a similar, and much longer story.

Of course many of those highly suppressive regimens come with highly depressing side effects. And that calls forth another psychoacoustical exercise practiced at this conference: the ongoing PI switch studies. They are nearly unanimous in showing that PI-induced suppression can be sustained, at least for a year or so, with combinations that include only reverse transcriptase inhibitors. And often, but far from always, high lipid levels and other measures of contorted metabolism begin to ease. But the bitter physical signs of lipodystrophy linger. In some cases, a few fear, they may never fade.

Still, the psychoacoustical voice here says switch. But the interlaced rhythms that yield this acoustic all but defy disentwining: successful suppression, overlaid by scary side effects, bissected by a promise of easier adherence without PIs, running headlong and full circle into fear of stopping that effective PI combo. Assurances that nonnucleosides are lipodystro-free took a small setback at this meeting. And nucleosides surely share some culpability in this syndrome. With mechanisms still so poorly understood, switching remains close to guessing.

At last we come to the most bedeviling acoustic acrostic, the structured treatment interruption. No other development in HIV management deserves a more careful audition. And the tune some listeners hear is caution. Caution is certainly the song sung by NIAID director Anthony Fauci, MD, when considering STIs. According to The New York Times, his take on STI findings at this meeting is that, "if there will be any success with interrupted therapy, it will be among those treated within weeks after they became infected."²⁴ And even in that rare cohort, he added, only a minority may profit from STIs.

That cold shower for STI enthusiasts may seem excessively circumspect to some, but Fauci is hardly alone in his view. Although leaders of the largest STI study so far, the Swiss trial presented by Bernard Hirschel [abstract 458], characterized their preliminary findings as "modestly encouraging," Hirschel himself suggested that STIs won't work for most people.

And in online audio comments, University of Colorado investigator Daniel Kuritzkes, MD, adroitly pointed out the rhythmic crosscurrents accounting for this psychoacoustic.²⁵ Labeling STI results at the meeting as "mixed," Kuritzkes reminded listeners that rehearsing the STI melody depends on abandoning the healing tonic of viral suppression. When physicians treat influenza, he elaborated, they would not countenance rhythmically withdrawing treatment just to keep antiflu immunity piqued. Should physicians be stopping antiretrovirals to keep antiretroviral immunity on edge when the benefits remain unproved?

For the cautious, the answer will come only with completion of large STI trials. In the meantime, the nearly irresistible charm of this tactic should itself arouse intense scrutiny of its merits. The opinions cited here deserve deep thought precisely because everyone wants so much to see STIs as the best medicine for primary infection, chronic infection, and uncontrolled infection. Many once saw HAART as a similar panacea.

Return to Article Contents

References and Notes

* Abstracts, and many posters, from the conference can be read at http://www.retroconference.org.

1. Michelmore H, Li YM. Highly successful induction/maintenance ARV therapy--three years experience. Presented at: Seventh European Conference on Clinical Aspects and Treatment of HIV Infection; October 23-27, 1999; Lisbon. Abstract 453. For details on Michelmore's adherence-evoking tactics, see: Mascolini M. Crossing the waters to the post-HAART era. J Intl Assoc Physicians AIDS Care 1999;5(12):29. Readers curious about Dr. Michelmore's approach but without a copy of this issue of the Journal are encouraged to e-mail this reporter at the above address for an outline of the Michelmore plan. Or watch for the book Dr. Michelmore is writing.

2. Safren SA, Otto MW, Worth JL. Life-Steps: applying cognitive-behavioral therapy to HIV medication adherence. Cognit Behav Pract. In press.

3. Molina JM, Chêne G, Ferchal F, et al. The ALBI trial: a randomized controlled trial comparing stavudine plus didanosine with zidovudine plus lamivudine and a regimen alternating both combinations in previously untreated patients infected with human immunodeficiency virus. J Infect Dis 1999;180:351-358. Available at: http://www.Journals.uchicago.edu/JID/Journal/issues/v180n2/981456/981456.html. Accessed February 29, 2000.

4. Saint-Marc T, Partisani M, Poizot-Martin I, et al. A syndrome of peripheral fat wasting (lipodystrophy) in patients receiving long-term nucleoside analogue therapy. AIDS 1999;13:1659-1667.

5. Kotler D. Lipodystrophy: return of the splitters. Medscape HIV/AIDS. Available at: http://hiv.medscape.com/Medscape/CNO/2000/ retro/Story.cfm?story_id=1009. Accessed February 29, 2000.

6. Pan-Zhou X-R, Cui L, Zhou X-J, et al. Differential effects of antiretroviral nucleoside analogs on mitochondrial function in HepG2 cells. Antimicrobial Agents Chemother 2000;44:496-503. Available at: http://aac.asm.org/cgi/content/abstract/44/3/496. Accessed February 29, 2000.

7. Moyle GJ, Baldwin C, Dent N, Gazzard BG. Maintenance of viral suppression but incomplete resolution of metabolic and clinical lipid abnormalities in persons switched from protease inhibitor (PI) based to efavirenz (EFV) based therapy. Presented at: Seventh European Conference on Clinical Aspects and Treatment of HIV Infection; October 23-27, 1999; Lisbon. Abstract 112.

8. Mascolini M. Crossing the waters to the post-HAART era. J Intl Assoc Physicians AIDS Care 1999;5(12):20-31.

9. Kotler DP, Rosenbaum K, Wang J, Pierson RN. Studies of body composition and fat distribution in HIV-infected and control subjects. J Acquir Immune Defic Syndr 1999;20:228-237.

10. Henry K. The case for more cautious, patient-focused antiretroviral therapy. Ann Intern Med 2000;132:306-311. Available at: http://www.acponline.org/Journals/annals/15feb00/henry.htm. Accessed February 29, 2000.

11. Cohen OJ. Antiretroviral therapy: time to think strategically. Ann Intern Med 2000;132:320-322. Available at: http://www.acponline.org/ Journals/annals/15feb00/cohen.htm. Accessed February 29, 2000.

12. Department of Health and Human Services (DHHS) and the Henry J. Kaiser Family Foundation Panel on Clinical Practices for Treatment of HIV Infection. Guidelines for the Use of Antiretroviral Agents in HIV-Infected Adults and Adolescents. January 28, 2000. Available at: http://www.hivatis.org/trtgdlns.html#AdultAdolescent. Accessed February 29, 2000.

13. Carpenter CC, Cooper DA, Fischl MA, et al. Antiretroviral therapy in adults: updated recommendations of the International AIDS Society-USA Panel. JAMA 2000;283:381-390. Available at: http://jama.ama-assn.org /issues/v283n3/abs/jst90023.html. Accessed February 29, 2000.

14. British HIV Association. British HIV Association (BHIVA) guidelines for the treatment of HIV-infected adults with antiretroviral therapy. Available at: http://www.aidsmap.com/bhiva. Accessed February 29, 2000.

15. Pitcher CJ, Quittner C, Peterson DM, et al. HIV-1-specific CD4⁺ T cells are detectable in most individuals with active HIV-1 infection, but decline with prolonged viral suppression. Nat Med 1999;5:518-525.

16. Malhotra U, Berrey MM, Huang Y, et al. Effect of combination antiretro viral therapy on T-cell immunity in acute human immunodeficiency virus type 1 infection. J Infect Dis 2000;181:121-131. Available at: http://www.Journals.uchicago.edu/JID/Journal/issues/ v181n1/990699/990699.html. Accessed February 29, 2000.

17. Margolick JB, Muñoz A, Donnenberg AD, et al. Failure of T-cell homeostasis preceding AIDS in HIV-1 infection. Nat Med 1995;1:674-680.

18. Gange SJ, Muñoz A, Chmiel JS, et al. Individual T-cell homeostasis failure and progression to clinical AIDS. Presented at: 5th Conference on Retroviruses and Opportunistic Infections; February 1-5, 1998; Chicago. Abstract 78.

19. Lori F, Rosenberg E, Lieberman J, et al. Hydroxyura and didanosine long-term treatment prevents HIV breakthrough and normalizes immune parameters. AIDS Res Hum Retroviruses 1999;15:1333-1338.

20. Miller V, Rottmann C, Hertogs K, et al. Mega-HAART, resistance and drug holidays. Presented at: Second International Workshop on Salvage Therapy for HIV infection; May 19-21, 1999; Toronto. Abstract 30.

21. Stewart B. The style of Philip Glass. Available at: http://www.cyberhalides.com/curator/composer.html. Accessed February 29, 2000.

22. Philip Glass. Encyclopaedia Britannica. Available at: http://www.britannica.com/bcom/eb/article/5/0,5716,37725+1,00.html. Accessed February 29, 2000.

23. Gradually/we became aware/of a hum in the room . . . Sometimes it was/a murmur/Sometimes it was/a pulse/Sometimes it seemed/to disappear . . . Maybe it's the hum/of our parents' voices/long ago in a soft light . . . Maybe it's the hum of changing opinion. "Changing Opinion." Lyrics by Paul Simon, music by Philip Glass, from "Songs From Liquid Days. CBS Inc. 1986.

24. Altman LK. Promise and peril of new drugs for AIDS. New York Times February 8, 2000:F1,F6.

25. Kuritzkes D. Structured treatment interruptions. Medscape HIV/AIDS. Audio report available at: http://hiv.medscape.com/Medscape/ CNO/2000/retro/public/audio.html. Accessed February 29, 2000.

Return to Article Contents

Weaker Bones:Another HAART Side Effect

Pablo Tebas, MD, and coworkers at Washington University in St. Louis used DEXA scans to compare bone mass in 60 people taking a PI combination, 35 HIV-positive people who never took a PI, and 17 healthy seronegative volunteers [abstract 207]. They found significantly lower bone density scores in the PI group than in seropositive and seronegative controls (P = 0.02 and P = 0.04, depending on the type of density score). Age may have influenced the results. People taking PIs were significantly older than seronegative controls (P < 0.001).

Tebas also reported that bone mineral densities did not correlate with abnormal fat changes in the PI group, a finding indicating that the side effects have different mechanisms of action. He declined to speculate on a cause of ostepenia and noted that the analysis did not include data on the duration of HIV infection or PI therapy.

Jennifer Hoy, MD, from the Alfred Hospital in Melbourne, did have some duration data in her osteopenia study [abstract 208]. She found no link between DEXA-determined bone mineral loss and length of antiretroviral therapy, length of PI therapy, individual drugs, CD4⁺ count, viral load, or insulin level. Like Tebas, she could not correlate testosterone levels with osteopenia.

The overall prevalence of osteopenia in this group of 80 men with lipodystrophy was 28 percent; 10 percent had osteoporosis. Osteopenia correlated with older age, lower weight, and lower lean body mass.

Hoy's analysis begins to address the question of whether PIs, in particular, lead to osteopenia, because the men she studied were members of the Australian PI switch study (see abstract 205). Osteopenia did not improve within 24 weeks among people who swapped their PI for an NNRTI, abacavir, adefovir, and hydroxyurea. Nor did bone mineral density continue to wane among people who stayed with their PI.

Return to Article Contents

000310
IA000301

ÆGIS is made possible through unrestricted grants from Roxane Laboratories, the National Library of Medicine, and donations from users like you. Always watch for outdated information. This article first appeard in 2000. This material is designed to support, not replace, the relationship that exists between you and your doctor.

Copyright © 2000 - International Association of Physicians in AIDS Care. All rights reserved. http://www.iapac.org