When to change a failing regimen: second-line and salvage strategies

Mark Mascolini for

Several studies show that it’s better for people with no good salvage options to continue a failing regimen rather than stop treatment altogether [1-3]. But how fast to abandon a faltering regimen when rescue options exist remains a contentious issue. Four EACS studies addressed this question.

Alessandro Cozzi-Lepri from London’s Royal Free Hospital and EuroSIDA cohort colleagues made several trenchant discoveries in their analysis of 110 people who continued a virologically failing three-or-more-drug regimen [4]. Viral loads inched up slowly as the faltering treatment continued, and CD4 counts drifted downward at a languid pace. Those dynamics in individual patients often encourage clinicians to stick with the same regimen rather than shop for options. But the EuroSIDA crew tracked a remorseless rise in resistance-conferring mutations as shaky regimens continued.

Cozzi-Lepri and colleagues studied 110 people who had two genotypes done while their viral load persisted above the 400-copy threshold on the same regimen. They figured a genotypic sensitivity score (GSS) for the failing regimen in which 0 means viral resistance to the regimen components, 0.5 means intermediate resistance, and 1 means sensitivity to the antiretrovirals used.

The median time between the cohort’s first and second genotype measured 6 months and ranged from 2 to 28 months. The median year of the first genotype was 1998 with a range from 1996 to 2001. Reflecting that era, almost half of the regimens rested on unboosted PIs, slightly more than 10% used a boosted PI, and slightly fewer than 10% used an NNRTI.

The group’s CD4 count averaged 280 cells/mm3 between the first and the second genotype, dropping about 10 cells/mm3 monthly. The average viral load inched up 0.14 log copies/mL monthly. But during these apparently tranquil months, 85 of 110 people (77%) picked up at least one resistance mutation, including:

  • One or more thymidine analog mutations in 27 (24.6%)
  • The 3TC mutation M184V in 7 (6.4%)
  • One or more nonnucleoside mutations in 13 (11.8%)
  • One or more major PI mutations in 40 (36.4%)
  • One or more minor PI mutations in 51 (46.4%)

At the first genotype, on average, the covertly floundering regimens included only 1.1 drugs to which HIV remained susceptible.

A multivariate analysis segregated a single factor that independently predicted evolution of more mutations as the failing regimen continued: People with fewer mutations on the first genotype—the very people with the most rescue regimen options—had the highest risk of gathering more mutations.

A different type of analysis in a different population reached the opposite conclusion-on average, few mutations evolve in people who continue a failing regimen with a modest viral load for 48 weeks [5]. But this finding is not too surprising when seen in the context of this Italian cohort. Firstly, many study participants had at least one NNRTI mutation (so the virus had no need to evolve new mutations to escape NNRTIs). Secondly, PI mutations tended to recede from detection (because most people were taking a non-PI regimen), and thirdly, nucleoside mutations did keep piling up.

The University of Brescia’s Paola Nasta ran this analysis as a substudy of the IMPROVE trial, which randomises lopinavir/ritonavir-naive people to continue an incompletely suppressive regimen or to start a new combination including lopinavir/ritonavir. Everyone had a viral load between 1000 and 20,000 copies/mL for at least 6 months while taking the same regimen, everyone had a CD4 count at or above 200 cells/mm3, and two or more regimens had failed in all study participants.

The 74 people randomised to stick with their antiretrovirals and the 76 assigned to try a new mix including lopinavir/ritonavir had similar baseline traits. The groups had taken antiretrovirals for a average 6.6 years, including (on average) 2.9 nucleosides, 0.9 NNRTIs, and 2 PIs. The last regimen before randomization consisted of two nucleosides and an NNRTI in 44.7%, two nucleosides and a PI in 37.3%, and three nucleosides in 17.9%. In the maintenance group 69% were taking a non-PI medley.

The baseline CD4 count averaged 422 cells/mm3 and the baseline viral load 3.6 log copies/mL. The study group had an average 3.39 nucleoside-related mutations, 3.16 PI mutations, and 0.93 NNRTI mutations (1.01 NNRTI mutations in the maintenance group).

After 48 weeks of follow-up, 11 of 74 people (15%) in the maintenance arm reached a study endpoint-either a viral load jump to 30,000 copies/mL or a CD4 drop below 200 cells/mm3. Only 3 people (4%) in the lopinavir/ritonavir switch arm reached one of those endpoints, a significant difference (P = 0.02). While 50 people (66%) in the switch group claimed a viral load below 50 copies/mL at week 48, no one in the maintenance group did. Similarly low proportions dropped out of the study because of side effects-3% in the maintenance group and 4% in the switch group.

Among people maintaining their baseline regimen, median total mutations numbered 7.7 at baseline and 7.7 after 48 weeks. But Nasta reported changes in specific mutation groups over the course of the study in people who stayed with the same drugs: While the median number of thymidine analog mutations stayed stable at 2.3, the median number of all nucleoside-induced mutations rose from 3.4 at baseline to 3.8 at week 48. Median numbers of NNRTI mutations hardly changed, from 1.01 at baseline to 1.06 at week 48. And median PI mutations dropped from 3.2 to 2.8 during the study, as these mutations tended to become undetectable in the 69% of maintainers not taking a PI.

Nasta also figured a genotypic sensitivity score (GSS) for the current regimen and a GSS for future options. On this GSS scale 0 represents the highest level of resistance for each drug and 3 indicates the lowest level for each drug. The final GSS represents the sum of scores for all drugs in a regimen or all drugs still available. While the GSS for the current regimen changed hardly at all, the GSS for future options dropped from 7.6 (+/- 3.9 standard deviation) at baseline to 6.9 (+/-4.8) at week 48. That ebb, though small, does not appear to support the researchers’ conclusion that “future treatment options … seem to be preserved” in the population they studied.

PIs resistant to resistance

Even after virologic failure of one, two, or three PI regimens, a hefty majority of viral isolates from Italian and US patients remained sensitive to lopinavir/ritonavir, fosamprenavir, or fosamprenavir/ritonavir in a study by Andrea De Luca from Rome’s Catholic University [6].

The analysis involved 643 people in the Catholic University cohort or the Stanford University database who had complete treatment histories on file, the entire protease sequence available for genotyping, and a history of at least one PI failure. Three or more PI regimens had flopped in 271 people (42%), two in 167 (26%), and one in 205 (32%). The most frequently used PIs were indinavir (28%), nelfinavir (22%), saquinavir/ritonavir (19%), lopinavir/ritonavir (8%), ritonavir (7%), and saquinavir (6%).

According to the Stanford genotypic resistance algorithm, which considers lopinavir/ritonavir but only unboosted fosamprenavir, most isolates retained either full sensitivity (34%) or partial sensitivity (20%) to lopinavir/ritonavir, while 46% of isolates were judged resistant. The Stanford algorithm called equivalent proportions fully sensitive (32%) or partially sensitive (18%) to unboosted fosamprenavir, while 49% were judged resistant.

According to the French national AIDS trial group (ANRS) algorithm, which considers boosted fosamprenavir and lopinavir, big majorities of isolates from these cohorts remained fully sensitive to fosamprenavir/ritonavir (91%) and lopinavir/ritonavir (83%). That difference in sensitivity was statistically significant (P < 0.001), but the ANRS formula reckoned that another 12% of isolates had intermediate sensitivity to lopinavir/ritonavir. Among isolates from people in whom three or more PI regimens failed, 87% remained fully susceptible to both boosted PIs, according to ANRS.

Still, a longer antiretroviral treatment history did trim the chances for full sensitivity to fosamprenavir/ritonavir and lopinavir/ritonavir according to ANRS rules. A multivariate analysis figured that each additional failed regimen lowered the chance of full susceptibility to fosamprenavir/ritonavir by 40% (odds ratio 0.60, 95% confidence interval 0.40 to 0.91, P = 0.017). A similar analysis determined that each added month of antiretroviral experience nipped the chance of full susceptibility to lopinavir/ritonavir by 2% (OR 0.98, 95% confidence interval 0.97 to 0.99, P = 0.006).

Earlier ritonavir use tended to lower odds of viral susceptibility to lopinavir/ritonavir after PI failure (OR 0.68, 95% confidence interval 0.42 to 1.08, P = 0.10), and earlier saquinavir use tended to cut the chance of viral susceptibility to fosamprenavir/ritonavir (OR 0.69, 95% confidence interval 0.37 to 1.25, P = 0.22).

Although longer treatment experience meant a higher risk of PI resistance in the Rome and Stanford cohorts, accumulating years of antiretroviral therapy did not translate into deteriorating virologic response in the Spedali Civili HIV clinic in Brescia, Italy [7]. In fact the percentage of people with an undetectable viral load climbed from just under 50% in 2001 to over 70% in 2005, reported Carlo Torti.

The cohort included 1406 people still naive to at least one antiretroviral class and 1341 who had tried the first three classes of antiretrovirals. Among 2449 people with antiretroviral experience, 911 (37.2%) had used more than seven antiretrovirals and 1130 (46.1%) had tried four to seven.

More treatment experience appeared not to curb chances of maintaining an undetectable load in this cohort. Torti could not measure viremia in 79.2% of people on their first regimen, 84.7% on their second or third regimen, and 74.6% trying their fourth or later combination. HIV RNA lay below detection limits in 78.5% of people treated for 6 to 12 months, 85.6% treated for 1 to 2 years, and 84.6% treated more than 2 years.

Torti concluded that in his cohort longer treatment with a wider array of antiretrovirals did not signal “treatment exhaustion”. Although regimen switching proved common, continued viral suppression appeared to indicate sustained activity of new and already-used antiretrovirals, and perhaps sharper therapeutic planning by Brescia’s HIV clinicians.



  1. Miller V, Phillips AN, Clotet B, et al. Association of virus load, CD4 cell count, and treatment with clinical progression in human immunodeficiency virus-infected patients with very low CD4 cell counts. J Infect Dis 2002;186:189-197.
  2. Ledergerber B, Lundgren JD, Walker AS, et al. Predictors of trend in CD4-positive T-cell count and mortality among HIV-1-infected individuals with virological failure to all three antiretroviral-drug classes. Lancet 2004;364:51-62.
  3. Kousignian I, Abgrall S, Graba S, et al. Effect of HAART, CD4 cell counts, and viral load on incidence of AIDS-defining events. 12th Conference on Retroviruses and Opportunistic Infections. February 22-25, 2005. Boston. Abstract 592.
  4. Cozzi-Lepri A, Phillips AN, Ruiz L, et al. Evolution of drug resistance in HIV infected patients remaining on a virologically failing HAART regimen. European AIDS Conference. November 17-20, 2005. Dublin. Abstract PE17.4/2.
  5. Nasta P, Matti A, Gatti F, et al. Are future treatment options reduced in heavily pre-treated HIV+ patients with sustained low level viremia randomised to defer HAART switch? European AIDS Conference. November 17-20, 2005. Dublin. Abstract PE3.4/8.
  6. De Luca A, Di Giambenedetto S, Colafigli S, et al. Prevalence of genotypic resistance to boosted fosamprenavir and lopinavir of clinical isolates from patients failing protease inhibitors: influence of genotypic resistance interpretation and number of failed regimens. 10th EACS, November 17-20, 2005. Dublin. Abstract PE3.4/11.
  7. Paraninfo G, Torti C, Quiros-Roldan E, et al. Antiretroviral drug experience and the treatment response at a population level: is treatment exhaustion really there? 10th EACS, November 17-20, 2005. Dublin. Abstract PE7.9/5.

Links to other websites are current at date of posting but not maintained.