No matter what your precise interest in this field -- antiretroviral pharmacology? ever-changeful epidemiology? devilish accessory genes? -- you will probably find at least one slide session on that topic and serried ranks of worthy posters.
And like the slippery retrovirus it seeks to delimit, CROI keeps evolving. Originally called the "National Conference on Human Retroviruses," CROI now hosts an international cast with a global (but strictly scientific) agenda. Especially in the last few years, the conference has offered a forum for research on HIV prevention and care in the epidemic's far-flung epicenters.
The 12th CROI, held February 22-25, 2005, in Boston, paid special attention to prevention of mother-to-child transmission with that embattled nonnucleoside reverse transcriptase inhibitor (NNRTI), nevirapine (NVP). This second part of the IAPAC Monthly's conference coverage also probes studies on resistance during chronic infection, antiretroviral side effects (the "vexations"), and the gay epidemic in North America and Europe (the "flirtations").
In a perspicacious and palliative talk, James McIntyre (Chris Hani Baragwanath Hospital, Johannesburg) observed that a solitary dose of NVP does not cause liver failure [abstract 8]. No significant clinical or laboratory toxicities emerged in 4,429 women or 4,875 infants exposed to single-dose NVP in six trials, in addition to the much-bruited HIVNET 012.
The tragedy is not that so many pregnant women are taking this drug, McIntyre counseled, but that so few do. He estimated that about 3 percent of HIV-infected pregnant women who need NVP or other short-course antiretrovirals get them.
But a sole dose of NVP does breed resistant virus -- probably in every infected woman who takes it. That judgment came from resistance savants Douglas Richman (University of California, San Diego) and François Clavel (Bichat-Claude Bernard Hospital, Paris). Slowly dwindling levels of this durable drug as HIV revs its replicating engine virtually guarantee that the virus will evolve the one or more mutations that can imperil responses to this drug -- or efavirenz (EFV). Conventional assays may not spot this resistant virus, but supersensitive assays can, as three CROI studies showed:
Johnson focused on 50 women in whom neither standard sequencing nor real-time PCR spotted K103N before they took one dose of NVP. After treatment the standard assay sighted K103N in 10 of 50 samples, and the souped-up CDC assay confirmed that finding. But in the 40 samples with no hint of K103N on routine testing, PCR descried the mutant in 16 (40 percent).
Next Johnson rolled out a real-time PCR probe aimed at the Y181C nonnuke mutation. With a population detection limit of 0.3 percent, the assay saw no Y181Cs before women took NVP. But in 46 women with no hint of Y181C on standard sequencing, real-time PCR picked it up in five (11 percent).
Searching plasma samples collected up to nine months after single-dose NVP, even Johnson's real-time PCR test couldn't see Y181C in six samples examined at three to six months or in five samples scoured at six to nine months. But K103N touched off the real-time PCR alarm in four of six samples surveyed at three to six months and in three of five at six to nine months.
Overall, Johnson figured that two thirds of women studied with real-time PCR had K103N or Y181C after single-dose NVP, a 62 percent jump from estimates with standard sequencing.
Reading an allele-specific PCR meter, Sarah Palmer (National Cancer Institute, Frederick, Maryland) confirmed that K103N and Y181C hang on for months after solo-dose NVP, even if routine sequencing can't spot them [abstract 101]. Palmer's analysis involved three groups of women enrolled in NVP prophylaxis trials -- eight in whom standard sequencing saw resistant virus six weeks and six months after single-dose NVP but not at 12 months, 11 in whom the standard test registered resistant virus at six weeks but not six or 12 months, and 10 in whom standard testing eyed no mutations in the six-week sample. With a population detection limit of 0.1 percent, the PCR test routinely tracked down K103N or Y181C in samples judged mutation-free by standard sequencing (Table 1).
In Palmer's study the overall frequency of mutant virus, 69 percent, equaled the rate figured by Johnson.
If even these supersnooping assays see resistant virus in about two thirds of samples from women who take single-dose NVP, why do Richman and Clavel think virtually everyone who swallows the drug once ends up with resistant virus? Because Johnson's and Palmer's PCR tests espy mutations only in plasma samples, but most virus -- resistant or otherwise -- takes cover in deeper lairs.
Shayne Loubser (National Institutes for Communicable Diseases, Johannesburg) used real-time PCR to trawl for mutations locked up in viral DNA inside peripheral blood mononuclear cells (PBMCs) [abstract 102]. The good news is that he graphed rapidly dwindling K103N frequencies in DNA over time, seeing almost none after 12 months. The bad news is that few virologists in the audience sounded encouraged. One observed that more resistant virus probably lurks in other vaults -- like memory cells in lymphoid tissue. Loubser agreed but maintained that his PBMC data confirm a marked overall dropoff with time. Retrovirus dean John Coffin (US National Cancer Institute, Frederick, Maryland) said other workers have had more success than Loubser in logging persistent mutations in PBMCs.
Another clinically crucial question is how well NVP will prevent HIV transmission when women try the single-dose regimen a second time. A case-control study by Neil Martinson (Johns Hopkins University, Baltimore) tried to find an answer, but his conclusion did not satisfy everyone [abstract 103].
This prospective study involved women taking NVP to avert transmission in a second pregnancy. The 75 "cases" had all tried single-dose NVP before, whereas the 132 controls had not. Martinson counted eight transmissions among cases (10.7 percent) versus five in controls (3.8 percent). In an analysis factoring in other transmission risks, those rates meant that women trying single-dose NVP a second time had a 2.3 times higher risk of giving their infant HIV. But because of the wide 95 percent confidence interval (CI) around that adjusted odds ratio (OR) (0.7 to 7.6), Martinson concluded that NVP works as well in a second try.
Not everyone agreed. Dissenters argued that the wide confidence interval may not mean repeat NVP works as well as first-time NVP, but that the study lacked statistical power to discern a clear difference.
Samples collected about two weeks after 3TC prophylaxis began showed the 3TC-evoked M184V mutation in nine of 13 children, the NNRTI K103N mutation in one, and the zidovudine (AZT)-related M41L in another. All of the M184Vs disappeared a median of five months after prophylaxis stopped, but M41L and K103N hung on in plasma samples. Ten of 13 children taking NVP prophylaxis had one or more NNRTI mutations soon after treatment began. All of those mutations endured for months after treatment stopped.
Giuliano urged wider use of rapid HIV assays after birth to avoid exposing infected infants to prophylactic monotherapy.
Robert Hogg (British Columbia Centre for Excellence in HIV/AIDS Vancouver) counted 238 deaths in the cohort for an all-cause mortality of 17.2 percent. Genotyping 3,120 viral samples -- a median of one per person -- he found resistance-linked mutations in 395 people (28.5 percent), including 271 isolates (68.1 percent) with resistance to 3TC, 138 (34.9 percent) with resistance to other nucleoside reverse transcriptase inhibitors (NRTIs), 199 (50.3 percent) with resistance to NNRTIs, and 106 (26.8 percent) with resistance to PIs. Nearly half of the resistant isolates -- 43.8 percent -- had double-class resistance, and 7.9 percent had triple-class resistance.
A multivariate model of baseline and time-dependent factors rounded up some familiar suspects in shortened survival -- older age, less physician experience with HIV, worse adherence, lower baseline CD4 count, and higher baseline viral load. Having resistant virus raised the risk of death 1.8 times in this model (95 percent CI 1.34 to 2.41). Another multivariate model retooled to weigh the mortal impact of specific mutation patterns found a heightened hazard ratio (HR) of death with NNRTI and non-3TC NRTI mutations, no clear pattern with 3TC-resistant virus, and apparent protection with PI-resistant virus:
Hogg and resistance maven Richard Harrigan did not speculate on why failure of NNRTIs -- but not PIs -- ups the odds of death. But in a just-published study by this team, probably involving mostly the same people, Harrigan and Hogg found that first-line NNRTI therapy independently raised the risk of multiclass failure 1.84 times (p = 0.001).2
To suspend treatment in this 94-person study, people needed a CD4 count above 450 cells/mm3 and a viral load under 50 copies/mL, and they had to restart if their T-cell tally fell under 350 cells/mm3 or if an opportunist arose. They could also resume treatment whenever they felt the urge. Thierry Prazuck (Regional Hospital Center, Orléans) reported that median months off treatment reflected lowest-ever CD4 counts:
Even nine months of freedom from antiretrovirals probably sounds like a good deal to lots of people. But besides the risk of retinitis and other ugly setbacks, everyone should know that resistance also poses a threat during CD4-steered STIs. That finding came from a randomized comparison of 87 people taking an STI and 30 continuing treatment [abstract 679]. Two lessons learned in earlier trials came home to roost in this one:
Looking for resistance mutations during and after drug breaks guided by CD4 counts or viral load, Lidia Ruiz (IrsiCaixa Foundation, Barcelona) found them in 32 of 87 break takers (37 percent). Among the 31 people with NRTI-linked mutations, 25 had sampled NRTI monotherapy or duo-therapy before taking a stronger combo, and most of them had thymidine analog mutations (TAMs) reflecting this earlier treatment.
Nine of 19 people (42 percent) who suspended an NNRTI regimen ended up with nonnuke mutations. In seven of these nine the mutations arose during the drug break, and in two after they resumed treatment. Genotyping of pre-NNRTI viral DNA did not reveal these mutations. All told, new mutations sprang up in four people (5 percent) who had not taken one- or two-drug therapy in earlier days.
Together these studies tell a cautionary tale: Despite mostly gladsome news on CD4-guided drug breaks, these holidays can have unhappy endings. In a review talk at CROI, Diane Havlir (University of California, San Francisco) -- no milquetoast when it comes to novel strategies -- warned that CD4-guided breaks still need rigorous study to sort out their risks and benefits [abstract 181].
Breaking isolates into six-month clusters based on 1998 to 2004 submission dates to Virco, mutation counters tallied more than 3,330 NNRTI mutation combinations not seen in earlier half-year periods, more than 17,200 NRTI mutation patterns making their debut, and more than 22,700 PI mutation arrays with no forebears in Virco's vaults. At the same time, 97 percent of PI mutation sets, 95 percent of NRTI mutation sets, and 87 percent of NNRTI mutation sets showed up in 10 or fewer isolates.
Exercising understatement, Rinehart concluded that "previously undetected combinations of mutations continued to be received at a steady rate."
These findings should stir more than idle curiosity; they reflect the necessarily gnarled dynamics of viral evolution, as HIV beads strings of amino acid substitutions aimed to escape drug pressure and keep itself fit. While Virco's Rinehart offered a satellite-high view of that breakneck evolution, Eoin Coakley (ViroLogic) snapped a submicroscopic close-up of a unique mutation tweak in one individual [abstract 716].
The man studied had a viral load of 6,547 copies/mL while taking NVP plus 3TC, abacavir (ABC), and tenofovir (TDF). Except for a quick taste of saquinavir (SQV) during a time when his viral load stayed under 50 copies/mL, he had tried no PIs. Keeping 3TC, ABC, and TDF, he traded NVP for atazanavir (ATV) and saw his viral load retreat to 118 copies/mL. Adding ritonavir (RTV) sent it below the 50-copy cutoff. But a few months later he endured a rebound to 7,535 copies/mL.
Genotyping after the rebound did not turn up I50L, the signature mutation seen with failure of unboosted ATV. But N88S, encountered in vitro but never apart from I50L in clinical isolates, did appear (along with some minor protease changes). Snooping through ViroLogic's viral database, Coakley pulled up all I50L or N88S isolates without other primary PI mutations. Tracking the prevalence of N88S from the third quarter of 2003 to January 2005, Coakley charted a climb from 1.04 to 2.86 percent. Over the same period, I50L prevalence vaulted from 0.15 to 3.65 percent. Every N88S in the database proved susceptible to lopinavir (LPV) and amprenavir (APV), while more than 95 percent laughed off ATV and nelfinavir (NFV). The patient's clinician swapped APV/RTV for LPV/RTV and his viral load sunk below 50 again.
This appears to be the first case of primary resistance to RTV-boosted ATV, Coakley noted. Whether N88S will routinely emerge as the culprit in such failures awaits further study. But this case study by ViroLogic neatly makes the point proposed in Virco's mass analysis: Expect the unexpected.
But don't expect top HIV clinicians to agree about genotypes, phenotypes, or both when picking rescue regimens. That conclusion emerged from a crafty exercise cooked up by Andrew Zolopa (Stanford University, California) with help from Stephen Becker (Pacific Horizon Medical Group, San Francisco), and some ViroLogic stalwarts [abstract 727]. They sweet-talked 109 clinicians attending regional advisory board meetings into picking new regimens based on actual genotypic (GT), phenotypic (PT), or GT-PT data -- but with no clues from case histories.
Zolopa dealt out GTs, PTs, and GT-PTs in random order and made sure all samples had decreased susceptibility (PT resistance) to at least one drug. Otherwise the samples could have no, minimal, moderate, or significant PT-GT discordance, resensitization or hypersusceptibility to certain drugs, quantifiable differences in PT susceptibility without PT-GT discordance, or PT-GT discordance attributed to mixtures in genotype. This is not a test you want to take after sampling single malts. Zolopa and collaborators powwowed on consensus picks of "acceptable" or "unacceptable" antiretroviral options.
Given these ground rules, the percentage of "acceptable" calls may rank as a minor miracle. The median percentage of regimens with at least three "acceptable" drugs was 65 percent (range 37 percent to 96 percent) for interpreting GTs, 86 percent (range 42 percent to 99 percent) for reading PTs, and 86 percent (47 percent to 97 percent) for PT plus GT. Demographics like professional degree, number of patients cared for, or self-assessed resistance acumen did not affect the results in a multivariate analysis.
Zolopa believes the results suggest that the resistance test used "may result in very different antiretroviral treatment regimens for some patients." Using PT or PT plus GT may yield better results than GT alone, he proposed. But he cautioned that this study cannot clinch that conclusion because it did not factor in actual outcomes in the people whose virus got analyzed.
K65R showed up in 1,881 samples (3.2 percent) and T215F/Y plus at least two other TAMs in 8,411 (14.2 percent). But only 24 samples -- a piddling 0.4 percent -- shared both K65R and T215F/Y plus two or more other TAMs. Some of those rare genomes also bore the infamous Q151M multi-NRTI resistance complex.
Parikh concluded that regimens including AZT and nucleosides or nucleotides that select K65R have a high genetic barrier to resistance and merit further study. Evidence of this high barrier came last year from Graeme Moyle (Chelsea and Westminster Hospital, London), who found that TDF plus Trizivir (AZT, 3TC, and ABC) shut down viral replication as well as Combivir (AZT/3TC) plus EFV after 48 weeks in previously antiretroviral-naive people.3 The pretreatment median viral load stood above 100,000 copies/mL in both treatment arms, and the median pretreatment CD4 count languished below 200 cells/mm3.
At the 12th CROI, two groups tested the triple nucleoside/tide regimen AZT/3TC/TDF without the extra oomph from ABC in treatment-naive people. French clinicians reported four virologic failures and five toxicity dropouts among 42 people who took the regimen for eight months [abstract 599]. As Parikh predicted, none of the failures came with K65R plus TAMs. These French study participants started AZT/3TC/TDF with a median 233 cells/mm3 and a median load of 4.88 logs (about 75,000 copies/mL).
A Ugandan trial of the same triple therapy in antiretroviral-naive people with much more advanced disease -- medians of 100 cells/mm3 and 333,000 copies/mL -- found a 54 percent sub-50-copy response rate at 24 weeks in a missing-data-equal-failure analysis [abstract 22]. (See "Another try with triple nukes" on page 77 of the March 2005 issue of the IAPAC Monthly.)
The K65R substitution may emerge more often than previously reported when 3TC/ABC/TDF fails, closer scrutiny of the ESS30009 trial suggested [abstract 720]. Clonal analysis spotted K65R in samples from four of four people in whom researcher originally blamed regimen failure on M184V/I alone [abstract 720]. Glaxo's Peter Gerondelis and colleagues believe these findings "suggest that the incidence of K65R at week 12 was higher than the 54 percent originally reported."
Emergence of K65R plus M184V is not the end of the world for previously treatment-naive people -- only the possible end of full susceptibility to TDF and all NRTIs except AZT.4 Indeed, Roland Landman (Bichat-Claude Bernard Hospital, Paris) reported that virus pocked with K65R and M184V looked hypersusceptible to EFV, NVP, and AZT by phenotypic fingerprinting [abstract 710]. Most of 22 people who picked up K65R with or without M184V returned to the realm of undetectability after 14 to 177 weeks of rescue therapy.
Fourteen of these people got K65R and other mutations in the French trial of once-daily 3TC/ABC/TDF5 and eight in the Gilead comparison of TDF with stavudine (d4T) plus 3TC and EFV.6 Fourteen people had K65R with M184V, five had those mutations plus NNRTI-induced changes, and three had K65R with NNRTI mutations. Sixteen people started a PI (LPV/RTV in 10), and six still naive to NNRTIs started EFV or NVP. Thirteen people started AZT, 11 started didanosine (ddI), nine kept TDF, and five kept 3TC.
After a median 48 weeks of follow-up, 19 people (88 percent) chiseled their viral load to a sub-50 sum. Among the three people who did not, two earned a "lost-to-follow-up" label and one had adherence problems. Two people in whom 3TC/ABC/TDF had failed kept taking those drugs and merely added EFV; both got below 50 copies/mL. Phenotyping in the whole group suggested that K65R/M184V mutants remained fully susceptible to either AZT or d4T and partly susceptible to TDF.
Does TAM-tainted virus respond to emtricitabine (FTC) as well as to 3TC? Early results from 3TC's maker, Glaxo, suggest these mutants may stay susceptible to 3TC more often than to FTC, though the differences in many cases were slim [abstract 713]. And as Glaxo's Lisa Ross observed, the results for FTC need clinical verification because she inferred the phenotypic ("fold-change") cutoff for FTC from 3TC data.
Ross figured median fold-change resistance to the two NRTIs in six mutant sets, defining TAMs as M41L, D67N, K70R, L210W, T215F/Y, and K219E/H/N/Q/R. Almost all K65R mutants readily shrugged off both 3TC and FTC, whereas virus with two or three TAMs at positions 67, 70, or 219 looked considerably more susceptible to 3TC (Table 2).
But if 3TC has an edge over FTC in slamming 67-70-219 TAMs, d4T may outpunch AZT on this so-called TAM-2 pathway, according to a 625-sample analysis by Alessandro Cozzi-Lepri (Royal Free and University College Medical School, London) [abstract 708]. The study also confirmed that TAM-1 and TAM-2 clustering does not happen by chance alone.
Cozzi-Lepri started with 625 EuroSIDA isolates trammeled by two or more TAMs. Two independent labs figured that 343 samples (54.9 percent) took the TAM-1 path, 114 (18.2 percent) preferred the TAM-2 path, and 168 (26.9 percent) maundered along other ill-defined routes. One factor independently predicted evolution of the TAM-2 path and two factors lowered the chance of a TAM-2 trail at the following adjusted OR and 95 percent CIs:
Even more intriguing was the suggestion that d4T-containing regimens may beat back TAM-2 virus, while AZT regimens roll over when faced with those mutations. Among people with a TAM-2 profile, the six-month viral load drop averaged 1.27 logs with d4T but only 0.19 log with AZT (p = 0.02). Response rates to TAM-1 virus were equivalent with d4T (-0.71 log) and AZT (-1.17 log). But Cozzi-Lepri adjured all to interpret those changes with caution: Because this is a cohort study -- not a trial randomizing people to d4T or AZT -- irksome "residual confounders" may sway the results.
On the lipodystrophy front, a randomized study saw benefits for TDF over ABC when swapping out a thymidine analog. Although a cohort study logged signs of kidney problems with TDF, the clinical significance of those hints (if any) remains obscure. A person's genetic make-up may sway the risk of that primeval NRTI problem, neuropathy.
For those who have been cryohibernating for the past few years, D:A:D is an 11-cohort European, Australian, and US assemblage of 23,441 people with HIV infection. Wafaa El-Sadr (Harlem Hospital, New York) reported that 18 percent remain innocent of antiretrovirals, 82 percent have tried NRTIs, 69 percent PIs, and 35 percent NNRTIs. The cohort is heavy in smokers, with a prevalence of 47 percent, and 42 percent have lopsided lipids. At last count the D:A:D team tallied 277 MIs for an overall incidence of 3.6 per 1,000 person-years.
A multivariate analysis singled out six factors that independently predicted a higher MI risk:
For the first time D:A:D also dug up a possible explanation for the higher heart attack risk with combination therapy, and it was no surprise -- bad lipid readings. Every time total cholesterol ticked up 1 mmol/L, the MI risk rose 12 percent (95 percent CI 1.04 to 1.21). Every time triglycerides doubled, the MI risk climbed 12 percent (95 percent CI 1.04 to 1.21). And every time high density-lipoprotein (HDL) cholesterol edged up 1 mmol/L, the MI risk hopped 15 percent (95 percent CI 1.06 to 1.24).
Sabin weighed five risk predictors -- current smoking; hypertension; high total cholesterol (>6.2 mmol/L) or total-to-HDL cholesterol ratio (>6.5) or low HDL cholesterol (<0.9 mmol/L); family history of cardiovascular disease; and age 45 or older in men and 55 or older in women. She defined high risk as having an earlier MI or stroke, having diabetes, or having two or more of the above risk factors.
The proportion of people with two or more heart disease harbingers rose from 34.3 percent in 2000, to 37.4 percent in 2001, 41.0 percent in 2002, and 41.5 percent in 2003, largely because people got older, got heart disease, or got diabetes (from 3.5 percent in 2000 to 5.4 percent in 2003). And the proportion of people who picked up a "high-risk" label over these four years bulged from 36.0 to 39.3, 43.1, and 43.7 percent.
But D:A:D doctors and cohort members did not sit on their hands as heart disease headlines blared. From 2000 to 2003 the percentage of people taking antilipid drugs climbed from 3.5 to 7.6 percent. The quotient of D:A:D nicotine addicts rose from 47.4 percent in 2000 to 51.9 percent in 2002, then slipped to 45.0 percent in 2003. Among men over 45 and women over 55, average total cholesterol and triglycerides dwindled marginally over the years, from 5.4 to 5.2 mmol/L and from 2.0 to 1.9 mmol/L.
Mixing these variables in one model, Sabin learned that the relative hazard (RH) of a heart attack dipped significantly after 2001:
Notably, the D:A:D team does not essay the arduous labor of sorting out MI risk by individual antiretroviral class or drug. So the precise effect of a worldwide turn to NNRTIs and to ATV in the past several years -- if any -- remains unknown in this cohort. But in the northern California cohort of Daniel Klein (Kaiser Permanente, Oakland), fervid adoption of ATV and swelling prescription of lipid-lowerers may be paying off [abstract 869].
Klein ranks among the premier tollers of the MI tocsin with his early report that men with HIV landed in the hospital for heart disease much more often than men without an HIV diagnosis.7 But that early work failed to forge a significant link between PI therapy and heart problems. And to this day Klein cannot make that link, though his case-control comparison now embraces 5,162 HIV-infected men tracked for 25,251 person-years and 42,531 controls followed for 295,005 person-years. Although the HIV group still has significantly higher coronary heart disease and MI rates, coronary "event" rates stratified by duration of PI therapy and by calendar period do not climb in "a clear linear fashion" over time.
Two reasons for this statistical shambling may be an Olympian leap in ATV prescriptions -- from none in the second quarter of 2003 to 27 percent of prescriptions in the fourth quarter of 2004 (p<0.0001) -- and a nearly equal leap in antilipid prescriptions among PI takers -- from 1 percent in the fourth quarter of 1997 to 22 percent in the fourth quarter of 2004 (p<0.0001). Klein did not report changes in NNRTI prescribing, but d4T fell from favor with a 48 percent prescription rate in 2001 to 17 percent in 2004 (p<0.0001).
Aquitaine cohort researchers gauged smoking rates and PI and antilipid drug use from 2000 to 2003 to estimate their effect on carotid artery intima media thickness (IMT), a sturdy predictor of heart disease [abstract 871]. At the study's starting point, reported Rodolphe Thiébaut (University of Bordeaux, France), 127 of 233 cohort members (60 percent) took a PI regimen, 138 (59 percent) smoked, five (2 percent) took fibrates, and none statins.
During the first year of follow-up, eight more people started fibrates, eight tried statins, 42 switched from their PI, and 10 quit smoking. Triglycerides, total cholesterol, and ominous low-density lipoprotein (LDL) cholesterol stayed stable in that period, but median IMT fattened from 0.55 to 0.57 mm (p<0.0001).
Over the next two years, 52 more people abandoned PIs, 30 more started antilipid drugs, and 14 more stopped smoking. At this point the 36-month prevalence of lipid-clipping therapy measured 17 percent, non-PI therapy 60 percent, and smoking 51 percent. In this two-year stanza total cholesterol and LDL cholesterol fell significantly (p<0.0001 and p = 0.05), as did median IMT (from 0.57 to 0.53 mm, p<0.0001).
Thiébaut traced a nearly significant link between the descending IMT slope and the no-smoking trend (p = 0.06), while changes in lipids and PI use did not correlate with IMT slope in this analysis. The statistical gravitas of snuffing cigarettes ebbed some with further adjustment for age, gender, body mass index, and total cholesterol (p = 0.11).
Lest anyone become too encouraged by these trends, Paige Williams (Harvard School of Public Health, Boston) and US AIDS Clinical Trials Group (ACTG) comrades reported an atherosclerotic cardiovascular disease incidence twice as high as D:A:D's (7.2 events versus 3.6 events per 1,000 person-years) [abstract 867]. And over half of the ACTG enrollees who succumbed to heart disease had less than a 10 percent baseline risk on the Framingham scale. But the ACTG cohort was much smaller than the D:A:D pack (643 people) and follow-up shorter (median 4.8 years). Also, an ACTG subset of 433 people with longer follow-up had a higher heart risk at baseline.
This analysis involved 256 people with risky lipid readings (triglycerides above 500 mg/dL, total cholesterol above 200 mg/dL, LDL cholesterol above 130 mg/dL) for at least the past three months. All switched, notably, to RTV-boosted ATV, and many also traded their NRTIs for ddI, 3TC, and/or TDF. Triglycerides, total cholesterol, and non-HDL cholesterol all dropped significantly after six months (-21 percent, n = 141; -13 percent, n = 143; -14 percent, n = 108) and 12 months (-43 percent, n = 32; -22 percent, n = 32; -24 percent, n = 28). The total-to-HDL cholesterol ratio ebbed from 5.7 at baseline to 4.9 at month 12 (p = 0.006).
Meanwhile, the switch to ATV and other regimen rejiggers boosted the proportion of people with a viral load under 500 copies/mL from 50 percent to 74 percent at month six (n = 141) and to 75 percent at one year (n = 32). CD4 readings rose by 45 cells/mm3 at month six and 48 cells/mm3 at month 12. Three people had to quit ATV because of jaundice and one because of spurting transaminases. All told, eight people stopped ATV because of side effects.
At the 11th CROI in 2004, a 16-week trial failed to find a triglyceride advantage for fish oil (omega-3 fatty acid) plus diet and exercise counseling versus counseling alone.8 That small study used a European product called Coromega at a dose of about 3 g daily. At the 2005 conference a much bigger double-blind, placebo-controlled comparison of another fish oil physic, Maxepa, significantly purged high triglycerides after eight weeks, and that drop held up through week 16 [abstract 39].
Pierre De Truchis (Raymond Poincaré Hospital, Garches, France) and coworkers recruited 146 people with triglycerides above 3 g/L while taking a stable antiretroviral combo. When dietary advice for four weeks failed to cut triglycerides to 2 g/L in 122 people, the researchers randomized them to take 2 g of Maxepa three times daily or placebo for eight weeks. Over the next eight weeks, everyone took Maxepa.
At week eight mean triglycerides plunged from 4.4 to 3.4 g/L with Maxepa while inching from 4.7 to 4.8 g/L with placebo (p = 0.0033). At that point 22.4 percent taking fish oil and 6.5 percent taking placebo notched a normal triglyceride level (p = 0.126). During the study's eight-week open-label phase, triglycerides stayed flat in the Maxepa group and sank to 3.4 g/L when the placebo group switched to fish tabs.
Side effect rates did not differ between groups during the eight-week blinded phase. Maxepa had no impact on viral load or CD4 count in this study. The fish tabs has no known interactions with antiretrovirals, although De Truchis did not measure drug levels. Maxepa's thrice-daily dosing may seem rigorous to people getting used to once-a-day antiretrovirals.
Everyone in this largely white, male study group had a viral load below 50 copies/mL for at least 24 weeks on the same regimen. For most people (63 percent assigned to TDF and 74 percent assigned to ABC), that regimen did not include a PI. The groups differed little in median age, nadir CD4 count, current CD4 count, or total limb fat (3.0 kg in the TDF arm and 2.9 kg in the ABC arm). The TDF group had a longer antiretroviral history (median 5.7 versus 4.9 years).
Three of 52 people (6 percent) randomized to TDF quit the study before week 48, compared with eight of 53 (15 percent) assigned to ABC. One person stopped TDF because of toxicity, and three dropped ABC for that reason -- all with a hypersensitivity reaction. A missing-data-equal-failure analysis of the 2005 conference (DEXA) arm-plus-leg fat at week 48 figured a 393-g gain with TDF and a 316-g rally with ABC, a nonsignificant difference (p = 0.97).
Subanalyses of 48-week DEXA results suggested that people switching from AZT did better with ABC (+210 g versus +66 g with TDF), while people taking d4T at baseline gained more with TDF (+529 g versus +375 g with ABC). Study participants taking a non-PI regimen also did somewhat better with TDF (+432 g versus +247 g). There weren't enough people in these subgroups to make these differences significant. But Moyle told IAPAC Monthly that the fat recovery in the group stopping AZT -- combined with data from ACTG 384 and ACTG 5125s [abstract 40] -- suggest that "AZT is a player in lipoatrophy and should be avoided in the long term."
People assigned to TDF did enjoy significantly greater (though small) improvements in certain lipids:
In follow-up questions Andrew Carr (St. Vincent's Hospital, Sydney) noted that more people assigned to TDF took d4T at baseline (77 percent versus 59 percent assigned to ABC). Adjusting the lipid analysis for that variable, he proposed, may make the significance of these differences disappear. Either way, TDF may be the safer short-term switch option, given the low but real risk of hypersensitivity with ABC.
A 24-week study at 15 ACTG sites found better peripheral fat gains with a non-NRTI regimen than with a simple switch of d4T or AZT to ABC [abstract 45LB]. Robert Murphy (Northwestern University, Chicago) and colleagues randomized 27 people with lipoatrophy to continue their current regimen for 24 weeks, 37 to trade d4T or AZT for ABC, and 40 to eschew NRTIs entirely and start LPV/RTV plus NVP.
After six months the LPV/RTV/NVP group gained CT-measured subcutaneous thigh fat (the study's primary endpoint) while the other groups continued shedding thigh fat (+8.4 percent versus -0.2 percent with ABC and -0.32 percent in the control arm, p = 0.06). The non-NRTI group added 8.0 CD4 cells/mm3, while the ABC group lost 4.8 cells/mm3 (p = 0.03) and the control group picked up an extra 2.4 cells/mm3. Both the non-NRTI contingent and the ABC troop gained subcutaneous adipose tissue (+16.6 percent and +9.2 percent), while the control group lost 8.8 percent. And both the non-NRTI and ABC assignees lost much more visceral adipose tissue (-15.0 percent and -15.3 percent) than the control arm.
Murphy did not report changes in lipids or other metabolic markers. A second ACTG study found that switching to another no-nuke regimen -- EFV plus LPV/RTV (533/133 mg twice daily) -- restored limb fat but boosted total cholesterol, LDL cholesterol, and triglycerides [abstract 40].
Other studies tracked bone, glucose, and kidney complications in people taking TDF with or without ddI. Worries over bone loss with TDF began when monkeys given high doses showed defects in bone building.13 Then a 144-week comparison of TDF with d4T (plus 3TC and EFV) charted lower spine density in the TDF group, though most bone mineral loss happened during weeks 24 to 48.6 Pursuing these leads, Nutrition for Healthy Living Cohort researchers measured total bone mineral density with DEXA scans over a median 2.9 years (IQR 1.5 to 4.1 years) in 283 men and 94 women [abstract 825]. The cohort included 100 people who took ddI at least part of the time during follow-up, 66 who took TDF, and 191 who took d4T.
Denise Jacobson (Tufts University, Boston) reported a median baseline CD4 count of 364 cells/mm3 and a median viral load below 1,000 copies/mL. The median change in bone density measured -0.17 percent yearly (IQR -0.8 to +0.43 percent, p = 0.0004). The loss came to 0.51 percent through three years of follow-up (IQR -2.5 to +1.3 percent).
In an analysis adjusted for age, gender, race, albumin, menopause, smoking, viral load and weight change, and time between visits, treatment with TDF (p = 0.0002), longer use of ddI (p = 0.001), and bilirubin above 2 mg/dL (p = 0.0003) independently predicted greater bone mineral loss. Longer treatment with d4T translated into less loss of bone density (p = 0.003).
Combining TDF with 400 mg of ddI daily raised the risk of hyperglycemia in a retrospective comparison of 78 people taking both drugs, 42 taking only TDF, and 57 taking only ddI [abstract 829]. Follow-up in this single-center study began before the recommendation to trim ddI doses with TDF, so the initial ddI dose was 400 mg in people weighing over 60 kg and 250 mg in those weighing less. During follow-up Theresa García-Benayas (Carlos III Hospital, Madrid) and coworkers cut the ddI dose to 250 mg for heavier people and to 200 mg for lightweights.
Fasting glucose averaged about 95 mg/dL in all study groups when follow-up began. After 12 months glucose rose to about 107 mg/dL in the ddI/TDF group while staying stable in the other groups (p<0.05). Multiple linear regression analysis considering ddI use, TDF use, PI use, weight, age, and viral load found that lower weight (p = 0.033) and ddI/TDF (p = 0.047) independently upped the risk of hyperglycemia. Eighteen of 22 people whose glucose soared, and six of seven diagnosed with diabetes, had taken 400 mg of ddI with TDF
No one, one hopes, gives 400 mg of ddI with TDF anymore. Fewer physicians probably combine these drugs at all. But long-term kidney and bone risks with TDF remain to be fully defined. Joel Gallant (Johns Hopkins University, Baltimore) offered a step toward that goal by comparing creatinine levels and clearance in 344 people starting TDF and 314 starting another NRTI [abstract 820]. Both groups were about three quarters male and three quarters African American, and about one third got HIV from injecting drugs. The groups also balanced well in viral load and CD4 count and diabetes rates (12 percent with TDF and 8 percent in controls). Gallant published his findings just after the conference.14
Median baseline creatinine levels measured 0.8 mg/dL in both groups and median creatinine clearance 117 mL/min in the TDF group and 118 mL/min in controls. After a median 303 days of TDF therapy and 336 days of other-NRTI therapy, median creatinine levels climbed 0.15 mg/dL with TDF and 0.10 mg/dL in the control group (p = 0.01). Creatinine clearance dipped 13.3 mL/min with TDF and 7.5 mL/min in controls (p = 0.005). The drop in clearance measured 10 percent with TDF and 6 percent with non-TDF nukes (p = 0.007).
Gallant found two independent predictors of percent change in creatinine clearance -- TDF use (p = 0.006) and CD4 count below 50 cells/mm3 (p<0.001). These correlations held true when baseline clearance exceeded 50 mL/min -- the signal to use full-dose TDF. Multivariate juggling also caught modest correlations between percent change in creatinine clearance and (1) baseline clearance below 50 mL/min and (2) diabetes (p = 0.10 for both). Age, gender, race, hypertension, hepatitis virus coinfection, and use of LPV/RTV or other specific PIs or NNRTIs did not correlate with a change in creatinine clearance.
"Although statistically significant," Gallant noted, "the clinical significance of these findings is unclear," partly because the drop in creatinine clearance did not force withdrawal of TDF. While 5.5 percent quit TDF at the time of maximal decline in renal function, 6.7 percent quit other NRTIs at that point.
Gallant reminded colleagues to test renal function before starting TDF and to follow advice on adjusting the dosing interval if clearance lies below 50 mL/min (see note 15).
Hulgan compared 147 people who had grade 1 or worse neuropathy with 362 no-neuropathy controls. Predictably, most people who suffered neuropathy, 73 percent, took ddI/d4T. The median age measured 39 years in cases and 35 years in controls. The analysis involved only non-Hispanic whites, in whom mitochondrial haplogroup H predominates. People who fell into haplogroup T had a 2.8 times higher risk of neuropathy in a univariate analysis (p = 0.019) and a 5.5 times higher risk if they took ddI/d4T (p = 0.006). Multivariate sifting gleaned three independent neuropathy predictors:
Hulgan stressed that these incriminating results must be confirmed in other cohorts. He did not determine mitochondrial DNA content or lactate levels in cases and controls.
Readers with keen interest in this case surely memorized the details long ago, and this review will not add to the nimiety. At the 12th CROI, Martin Markowitz (Aaron Diamond AIDS Research Center, New York) offered a richly detailed poster printed in agate type, and those sundries appeared a few weeks later in a free-access Lancet article.16 Nor will this article broach the battle royal over the propriety of announcing the case at a press conference (except to imagine the ugly fracas if the case leaked before formal disclosure).
The salient facts brought forward by this sad history are not controversial at all:
Across North America and Europe, the HIV epidemic has become a babel of viral subtypes, ethnic helter, and social skelter -- applied in a rich schmeer of risk behaviors. Whoever suggested the original acronym GRID -- for gay-related immune deficiency -- must be happy that tag didn't stick. Yet, although blacks in the United States and women worldwide shoulder disproportionate burdens of this plague, some of the most pungent recent trends involve the primal "risk group" -- gay men in rich countries.
Fraternité á grands frais
Despite a big population of HIV-infected people from sub-Saharan Africa, France may be looking at a resurgent gay epidemic. Using a recent-infection assay for the first time on a national level, Caroline Semaille (Institut de Veille Sanitaire, Saint-Maurice) learned that gay sex accounted for 51.2 percent of new infections from March 2003 to March 2004, followed distantly by heterosexual intercourse (25.7 percent), injecting drugs (16.4 percent), and "others/unknown" (29.3 percent, a high quotient hinting at undisclosed same-sexism or drug use) (p<0.0001).
These findings may mean that HIV has started spreading faster among French gays, Semaille concluded, that more gays got tested recently, or both. More testing, one could speculate, may reflect more trips to the sexually transmitted disease (STD) clinic for syphilis and lymphogranuloma venereum (LGV), the frightening emerging STD considered below.
Gay men in France are giving each other not only HIV, but also hard-to-treat genotype 4 hepatitis C virus (HCV), reported Marie-Laure Chaix (René Descartes University, Paris) [abstract 122]. In a cluster of 12 Parisians infected with the same HCV strain, two declined anti-HCV therapy. Nine of the remaining 10 did not respond to antiviral therapy in six months, and the tenth responded then relapsed.
As Bennett noted, though, the analysis may under-represent people diagnosed in private practice or nonpublic centers. So it may under-represent people with better access to antiretrovirals. But men who have sex with men (MSM) made up the largest part of the cohort by exposure group (61 percent versus 25 percent heterosexual and 14 percent injecting drug users). Because the six states are Colorado, Illinois, Maryland, Michigan, Virginia, and Washington, the survey missed the biggest gay urban centers.
Bennett and collaborators collected 787 samples from newly diagnosed (not necessarily recently infected), untreated people and genotyped them for reverse transcriptase and major protease mutations. The overall rate of resistant virus transmitted in 2003-2004 measured 14.5 percent, with NNRTI-resistant virus at the head of the class (8.4 percent). The survey found 24 people (3.1 percent) with virus resistant to two or more antiretroviral groups. So, although transmission of multidrug-resistant virus remains rare in the United States, it is hardly a freak phenomenon.
A substudy of 633 demographically grouped people revealed no differences in rates of infection with resistant virus by race or ethnicity, HIV transmission group, age, or gender (Table 3).
The overall transmission rate in the CDC survey -- 14.5 percent -- was lower than the 19.7 percent found in a North Carolina study and much lower than the 24.1 percent figured in a largely gay New York City population by Anita Shet (Aaron Diamond AIDS Research Center) [abstract 289]. And the rate in New York, Shet reported, is on the rise. Her study involved 112 people with acute or recent infection who came to the Aaron Diamond Center in 2003 and 2004. All but two were men, most of them gay. The group's age averaged 34.3 years and ranged from 19 to 56 years.
Twenty-seven people (24.1 percent) had virus bearing at least one resistance mutation, and 11 (9.8 percent) had mutations conferring resistance to more than one antiretroviral class -- trebling the rate in the CDC six-state study. Multiclass resistance rose significantly in this cohort since 2000 (p = 0.03). Three people (3.1 percent) had virus with mutations to the first three antiretroviral classes. If there is any good news here, it is that no one had mutations to the fusion inhibitor ENF and that people infected with resistant virus responded to treatment as well as those with wild-type virus.
As in the six-state CDC audit, transmission of NNRTI-resistant virus is climbing in New York City. Shet recorded NNRTI mutations in 7.1 percent in 2003-2004 versus 2.6 percent in 1995-1998 (p = 0.09). This trend toward swelling transmission of NNRTI mutations may reflect the relatively greater fitness of this virus compared with PI or NRTI mutants -- or even wild-type virus (see note 17).
The bustling trade in NNRTI mutant virus between sex partners "necessitates caution when considering initial empiric combination therapy with NNRTIs," the Aaron Diamond team counseled. They now start treatment of acute infection with a boosted PI regimen until they get genotyping results.
Charles Hicks (Duke University, Durham, North Carolina) looked for resistance mutations in two North Carolina groups of acutely or recently infected untreated people -- 110 referred to HIV clinics at Duke or the University of North Carolina (the "clinic" group) and 127 diagnosed at state-run HIV testing centers (the "state" group) [abstract 673].
The clinic and state groups did not differ much in age, with medians of 27 years (range 17 to 89 years) in the clinic and 30 years (range 15 to 59) in the state. More people in the clinic population were men (90 percent versus 68 percent) and gay (65 percent versus 39 percent), while the state program had a higher proportion of blacks (61 percent versus 43 percent).
The groups did not differ much, however, in rates of resistant virus. Overall, 25 (19.7 percent) had virus with at least one resistance mutation, and four (3.1 percent) had mutations rendering virus resistant to more than one class. Unlike the CDC and Aaron Diamond populations, the North Carolina groups had higher rates of resistance to NRTIs (7.1 percent) than to NNRTIs (6.3 percent). PI resistance prevalence measured 3.1 percent. Hicks told IAPAC Monthly that he could verify no risk difference by HIV transmission category. The largest proportion of those infected with resistant virus -- but not a majority -- got infected through gay sex.
Together these two studies suggest a sadly similar rate of resistance transmission in rural and urban North Carolina and on New York's mean streets.
Why does superinfection seem more likely in recent seroconverters? Grant suggested that this trend "may reflect a window period for susceptibility (as in nonhuman primate research), or seroconverters may harbor several viruses that appear sequentially due to viral escape for immunologic reasons." Superinfection may become rarer during chronic HIV infection because of (1) specific antiviral immune responses to partner viruses, (2) viral interference, or (3) nonspecific immune responses.
Whatever the mechanisms, superinfection is on the upswing in the San Francisco Bay Area. These four (and possibly seven) cases arose in the last 195 person-years of observation. Grant saw none in the preceding 56 person-years.
People who blithely risk HIV infection (and maybe even some clinicians who see them) may consider acute infection a nasty but self-limiting sweaty bout that does little immediate damage -- surely nothing that antiretrovirals can't fix five or six years down the road. Wrong. Reviewing his own recent work and that of others, Douek took the body count of HIV's early blitzkrieg:
All of this happens before seroconversion, while many just-infected people sit home from work, wondering how they got "the flu." The wasted CD4 troops never bounce back -- and this onslaught is visually awesome, as Douek demonstrated by contrasting a slide of normal intestinal lymphoid tissue and the blasted mucosa inside a recently infected person. Before an HIV antibody test even signals infection, virus-induced inflammation burns though the bowels, replacing healthy mucosal tissue with collagen. For those without medical training, collagen is the stuff that makes up crusty cartilage and bone -- not at all what you want in your guts.
"This doesn't look like a viral infection," Douek opined. "This looks like chemotherapy."
Magid Herida (Institut de Veille Sanitaire, Saint-Maurice, France) and colleagues retrospectively analyzed stored rectal specimens from January 2002 to April 2004 and prospectively rated samples from five Paris STD clinics and the national reference center in Bordeaux. They counted three LGV cases in 2002, 18 in 2003 (13 in the last quarter), and 102 in the first 11 months of 2004.
Collecting epidemiologic data from the first 14 people with rectal LGV, Herida found that all were gay men reporting unprotected anal sex with pick-up partners. None had visited an LGV-endemic area. Eight (57 percent) had HIV infection, nine (64 percent) had another STD in the last year, and all had anorectal signs or symptoms.
Symptom duration ranged from 11 to 120 days before diagnosis (median 50 days) and included rectal pain, discharge, and tenesmus (painful spasm of the urogenital diaphragm). Constipation, inguinal ulceration, and inguinal lymphadenopathy also characterize this disease. Three people with HIV had fever, and eight had "deep, extended rectal ulcerations." One man with a late diagnosis endured "a rectal tumor-like stricture." LGV is tough to diagnose, the French team observed, because it may mimic rectal carcinoma or Crohn's disease. Three HIV-infected people had cancer-like lesions.
From the Netherlands, Joke Spaargaren (Municipal Health Service of Amsterdam) and coworkers reported a new Chlamydia trachomatis strain causing LGV in gay men [abstract 896]. Because the earlier Dutch LGV outbreak involved gays in Rotterdam, he wondered if the STD made an earlier debut in Amsterdam. To find out, Spaargaren analyzed 82 rectal samples from 72 men who came to the health service STD clinic with proctitis in 2002 and 2003. Proctoscopic exam separated them into two groups -- 42 symptomatic men with pussy anal seepage or bloody ulcerative rectal growths, and 32 without symptoms or rectal mucosal inflammation.
With polymerase chain reaction (PCR)-based genotyping and gene sequencing, Spaargaren learned that 39 of 51 samples from symptomatic men (78 percent) bore a new L2b strain of C. trachomatis, as did 13 of 31 samples (42 percent) from asymptomatic men. Most of the men with LGV were gay. Spaargaren concluded that the Dutch LGV outbreak can be traced back at least as far as 2002 in Amsterdam.
In a retrospective cohort analysis of 3,455 HIV-infected people who visited the Owen Clinic at least once between January 2000 and December 2003, Mathews cataloged 94 cases of MRSA, 38 (40 percent) defined as community acquired and 56 (60 percent) as nosocomial (probably picked up in a hospital or extended-care facility, or from an indwelling catheter). He linked only one factor to nosocomial acquisition of MRSA: cotrimoxazole resistance raised the risk 12.4 times.
Diagnosis of community-acquired or nosocomial MRSA jumped more than six times over the four years of study, from under one case per 100 person-years before 2003 to about five per 100 person-years in the second half of 2003. In a multivariate analysis, heterosexual transmission of HIV lowered the risk of MRSA 90 percent (adjusted HR 0.1, 95 percent CI 0.0 to 0.7, p = 0.012). Mathews told IAPAC Monthly that 65 percent of those with MRSA were gay men who did not inject drugs and 12 percent were gays who did shoot up. Taking antiretrovirals in the past half year trimmed the MRSA risk 40 percent (p = 0.02), while a low CD4 count or a high viral load independently raised the risk:
Mathews cautioned that the climbing incidence may reflect heightened scrutiny as clinicians learn about MRSA, and unmeasured confounders like active substance abuse or short-term antibiotics could fudge the findings. Even so, the close correlation between poor control of HIV and MRSA "suggests a potential direct effect of HIV on anti-staphylococcal defenses."
How MRSA travels from one person with HIV to another remains unknown. Rieg and UCLA colleagues explored a nasal route, hypothesizing that gay men with HIV carry MRSA more often than MSSA in nasal secretions. They culled suspect secretions from 194 gay men attending their clinic and found S. aureus in 51 samples (26.3 percent). But 43 staph-laden samples (22.2 percent) carried MSSA compared with eight (4.1 percent) that bore MRSA.
Rieg rated seven of the eight MRSAs as community acquired, because one carrier had been in the hospital in the past half year. MRSA proved more common among African Americans (five of eight cases) and among people with a college education (seven of eight). Comparing the eight people with MRSA and the 186 without it, he tied three factors to the resistant bug:
Because few HIV-infected gays in this population carry MRSA in the nose, Rieg suggested that skin-skin or skin-fomite contact may explain transmission. (A fomite is any object that may carry a pathogen.) An alternate explanation may be that nasal MSSA picks up the Mcc/4 gene conferring methicillin resistance before inciting active infection.
But it's hard to answer Harold Jaffe's (University of Oxford, England) question about why gay men still risk their health almost 25 years after he helped us understand how HIV spreads. Speaking at CROI about public health worries stirred by the New York City case, Jaffe defused the fluster over using case reports to scare people into better behavior.
"The important thing is to put out the facts," he said. "If the facts are scary, then people will be scared."22
But given what we already know of the rampant nihilism among many sexually active gays, one wonders what it takes to scare them.
Mark Mascolini writes about HIV infection (firstname.lastname@example.org).