HBV or HCV coinfection produced higher risk from treatment interruptions: drug holidays and hepatitis don’t mix

Mark Mascolini,

SMART, the trial comparing continuous antiretroviral therapy with CD4-count-guided drug breaks, started with the hypothesis that drug holidays would lower the risk of antiretroviral side effects without threatening progression of HIV infection [1]. But it ended with a trove of data showing that treatment lulls pose a substantial risk of HIV progression while making a host of non-AIDS complications more likely.

At the February 2007 Conference on Retroviruses and Opportunistic Infections, SMART statistician Andrew Phillips reported that CD4-guided treatment breaks heightened chances of heart disease, possibly because steady antiretroviral therapy exerts an overall positive effects on risk factors [2].

At the 4th IAS Conference on HIV Pathogenesis, Treatment, and Prevention, SMART investigators served up results of fresh analyses showing that:

  • Interrupting antiretrovirals proved “particularly unsafe” for people with hepatitis B or C virus (HBV or HCV) infection [3].
  • SMART enrollees coinfected with HBV had to restart antiretrovirals more often than those without HBV [4].

SMART signed up 5472 mostly treatment-experienced people and assigned 2752 to keep taking antiretrovirals, regardless of CD4 count and 2720 to defer treatment until their CD4s tumbled below 250, then to resume or start therapy, only to stop again when CD4s climbed back above 350 [1]. The trial came to an abrupt end in January 2006, after an average 16 months of follow-up, when investigators found a 2.6 times higher risk of opportunistic disease or death as well as higher risks of heart, liver, and kidney disease in the drug-holiday group.

Two SMART substudies showed that CD4-guided treatment breaks may be a particularly bad idea for people coinfected with HBV or HCV [3,4]. Analysis of 922 SMART enrollees coinfected with HBV and/or HCV found that they accounted for about half of all non-AIDS deaths during the study, even though coinfected people made up only 17% of the whole cohort [3].

Defining chronic HBV as a positive test for hepatitis B surface antigen for more than 6 months and chronic HCV as positive for HCV antibody, the SMART team counted 922 coinfected people in the whole cohort (16.8%) and analyzed 467 in the drug-break group and 446 in the steady-therapy group. Equivalent proportions in both study arms had HBV only (n = 110), HCV only (n = 798), or both HBV and HCV (n = 14). In the coinfected subgroups, nadir (lowest-ever) CD4 count and CD4s at study entry were also equivalent in the two groups (median nadir 257 in the treatment-interruption group and 250 in the steady-therapy group; median entry CD4s 598 in the interruption group and 567 in the steady group). About two thirds of coinfected SMART enrollees had a viral load under 400 copies and about one quarter had AIDS.

When SMART ended, risk of opportunistic disease or death proved nearly identical in treatment interrupters with and without HBV or HCV coinfection. Coinfected people who took drug holidays had a 2.58 times higher risk of opportunistic disease or death than people who stayed on therapy. Among SMART enrollees without HBV or HCV, that risk was 2.57 times higher in the drug-holiday group.

Risk of death from a nonopportunistic disease proved 3.9 times higher in coinfected break takers than in the HIV-only drug-break group, and 3.5 times higher in the coinfected steady-therapy group than in the HIV-only steady-treatment group. SMART statisticians figured that this higher risk of non-AIDS deaths in coinfected people entirely accounted for the overall 2-fold higher risk of opportunistic disease or death in coinfected versus noncoinfected enrollees. The overall risk of a non-AIDS death was more than 3.5 times higher in coinfected people than in people without hepatitis virus coinfection (Table 1).

Table 1. AIDS and non-AIDS death rates with and without hepatitis coinfection

Total AIDS deaths Rates per 100PY (95% CI) Total non-AIDS deaths Rates per 100PY (95% CI)
HBV and/or HCV coinfected 2 0.14 (0 to 0.33) 37 2.52 (1.71 to 3.33)
HBV and/or HCV uninfected 5 0.08 (0.01 to 0.15) 41 0.69 (0.48 to 0.90)

But hepatitis itself did not explain the higher non-AIDS death risk in HBV/HCV-coinfected people. In the coinfected group death rates per 100 person-years measured about 0.2 for liver disease, 0.3 for kidney disease, and 0.5 for non-AIDS cancers and substance abuse.

SMART investigators concluded that because the risk of non-AIDS deaths runs so much higher in HBV/HCV-coinfected people, “the strategy of antiretroviral therapy interruption is particularly unsafe in these patients.”

Using the same definitions of chronic HBV and HCV infection, another SMART team discovered that HBV coinfection by itself made restarting antiretrovirals more likely in treatment interrupters [4]. This analysis focused on 2669 study participants, all of them randomized to take CD4-guided drug breaks. Median baseline CD4 count measured 560 in 65 people coinfected with HBV, 608 in 402 people with HCV, and 595 in 2202 infected only with HIV. Respective nadir CD4 counts stood at 207, 265, and 250. Similar proportions in all three groups were taking tenofovir, emtricitabine (FTC), and/or 3TC.

In the average 16 months of follow-up, 63.1% of study participants in the HBV group had to resume therapy, compared with 45.5% in the HCV group and 39.2% in the group without HBV or HCV. Median CD4 counts when treatment resumed were similar in the three groups – 233 with HBV, 240 with HCV, and 232 with neither hepatitis virus. A multifactorial analysis pinpointed seven factors that independently made treatment resumption more or less likely, including HBV coinfection, which raised the risk by two thirds (Table 2). HCV coinfection had no impact on the need to restart therapy.

Hepatic flares or restarting antiretrovirals at higher CD4 counts did not explain the higher resumption rate in people with HBV. Rather, SMART statisticians reckoned that HBV-coinfected people had to resume treatment more because their CD4 counts plunged faster than those of other people when they took drug breaks. While 24.4% in the HBV group who resumed treatment did so because of a speedy CD4 drop, 21.6% without hepatitis and 16.4% with HCV restarted therapy for that reason.

Table 2. Independent predictors of need to restart therapy in SMART

Hazard ratio 95% confidence interval p-value
Raised the risk
HBV coinfection 1.67 1.22 to 2.29 0.0014
Prior AIDS diagnosis 1.42 1.23 to 1.61 <0.0001
Baseline viral load <400 copies/mL 1.19 1.04 to 1.37 0.023
Highest viral load 1.19 1.11 to 1.28 <0.0001
Age (per 10 years older) 1.14 1.07 to 1.21 0.0001
Lowered the risk
Nadir CD4 count (per 100 cells higher) 0.67 0.63 to 0.90 <0.0001
Baseline CD4 count (per 100 cells higher) 0.87 0.85 to 0.90 <0.0001

This article is part of a longer report on the SMART trial

See also:

HCV/HBV Coinfected at Greater Risk in SMART

Higher rate of HAART reinitiation among HIV-HBV coinfected patients in the episodic arm of the SMART study


  1. The Strategies for Management of Antiretroviral Therapy (SMART) Study Group. CD4+ count-guided interruption of antiretroviral treatment. N Engl J Med. 2006;355:2283-2296.
  2. Phillips A, Carr A, Neuhaus J, et al. Interruption of ART and risk of cardiovascular disease: findings from SMART. 14th Conference on Retroviruses and Opportunistic Infections. February 25-28, 2007. Los Angeles. Abstract 41.
  3. Tedaldi E, Puoti M, Neuhaus J, et al. Opportunistic disease and mortality in patients co-infected with hepatitis C virus (HCV) and/or hepatitis B virus (HBV) in the SMART (Strategic Management of Antiretroviral Therapy) study. 4th IAS Conference on HIV Pathogenesis, Treatment, and Prevention. July 22-25, 2007. Sydney. Abstract TUAB203.
  4. Dore G, Soriano V, Neuhaus J, et al. Higher rate of antiretroviral therapy reinitiation among HIV-HBV coinfected patients in the episodic therapy arm of the SMART study. 4th IAS Conference on HIV Pathogenesis, Treatment, and Prevention. July 22-25, 2007. Sydney. Abstract TUAB204.

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