Mortality According to CD4 Count at Start of Combination Antiretroviral Therapy Among HIV-infected Patients Followed for up to 15 Years After Start of Treatment: Collaborative Cohort Study

The strong association of CD4 count at start of combination therapy with subsequent survival in HIV-infected patients diminished during the first 5 years of treatment. After 5 years, lower baseline CD4 counts were not associated with higher mortality.

3-and 5-year survival were published in 2002 and 2007 [3,4] and showed CD4 count at start of ART (baseline) to be the strongest predictor of short-term mortality. However, baseline CD4 count becomes less prognostic when 6-month response to treatment is taken into account [5]. HIV-1 RNA load at 6 months and changes in CD4 count from 6 to 36 months after start of ART are prognostic for AIDS in patients who have survived 3 years of ART [6].
To date, short-term prognosis has been extrapolated to predict long-term disease-free survival [7,8]. Because mortality in the general population varies by age, sex, and country of residence, estimates of mortality rate ratios (MRRs) may be more informative when corrected for background mortality [9]. We investigated how long patients bear the burden of increased mortality due to starting ART with low CD4 counts, by studying the prognostic value of baseline CD4 count for mortality up to 15 years after start of ART, using both standard survival analysis models and models for relative survival.

Study Sample
The ART-CC, which is described in detail elsewhere (http://www. art-cohort-collaboration.org) [10], is an international collaboration of HIV cohort studies from North America and Europe that combines data on HIV-infected individuals aged ≥16 years who were antiretroviral-naive when they started ART with a combination of at least 3 drugs. Eligible regimens include at least 1 PI, NNRTI, or integrase inhibitor and 2 nucleoside-analogue reverse transcriptase inhibitors (NRTIs). The present study included data from 18 cohorts, which are listed in the Supplementary Appendix. All were approved by local ethics committees. The National Health Service Health Research Authority South West-Cornwall and Plymouth Research Ethics Committee, United Kingdom, approved the study (REC reference 12/SW/0253). The data used here included follow-up from 1996 until 31 July 2013. We included patients who started ART between 1996 and 2001 (so had at least 12 years of potential follow-up) and had at least 1 available baseline (within 3 months prior to starting ART) measurement of CD4 count and viral load.
We used unadjusted (Kaplan-Meier) and adjusted estimates of cumulative mortality to examine whether CD4 count at start of ART was prognostic for mortality. The adjusted estimates were for a typical patient: men who have sex with men (MSM) aged 30-39 years, who started ART between 1998 and 1999 in the French Hospital Database on HIV (FHDH) cohort without a diagnosis of AIDS, and with high viral load (HIV-1 RNA > 100 000 copies/µL). We used Poisson models to estimate crude mortality rates (MRs) with 95% confidence intervals (CIs), for all patients and by baseline CD4 count. We then estimated MRs separately according to duration of time on ART (<0.5, 0.5-0.9, 1-2.9, 3-4.9, 5-9.9, and ≥10 years after ART start), overall, and by baseline CD4 count. We used Cox models stratified by cohort to estimate crude and adjusted (for sex, age, transmission risk group, AIDS at baseline, baseline viral load, and year of starting ART) MRRs by baseline CD4 group with 200-349 cells/µL as the reference category, overall and separately according to time on ART. Relative survival models based on generalized linear models (modified Poisson models) were used to account for variation in background population mortality [11]. Expected death rates for each country split by sex and 5-year age group were obtained from the Human Mortality Database (www.mortality.org). We estimated adjusted (for the same set of covariates) relative MRR, overall and according to duration of time on ART. The prognostic value of baseline CD4 count according to time since start of ART was compared by estimating the adjusted MRR per CD4 count category (assuming a log-linear relationship with MR across the CD4 categories). All analyses were performed using Stata software version 13 (StataCorp, College Station, Texas). There were 6344 deaths during 359 219 years of follow-up. Baseline CD4 count was highly prognostic for unadjusted cumulative mortality in all patients ( Figure 1). Baseline CD4 count was also highly prognostic for adjusted cumulative mortality, which is illustrated for a typical patient group (MSM in the FHDH cohort, aged 30-39 years, who started ART between 1998 and 1999 without a diagnosis of AIDS, and with high viral load [HIV-1 RNA > 100 000 copies/µL]) in Figure 2. There is a clearly graded increase in mortality with decreasing baseline CD4 count: Cumulative 15-year mortality varied from 7.1% (95% CI, 2.9%-11.3%) for baseline CD4 count >500 cells/ µL to 10.7% (95% CI, 4.7%-16.7%) for baseline CD4 count <50 cells/µL. A table showing the corresponding numbers at risk for Figure 1 at different time-points split by CD4 count is available from the authors on request.

A total
Numbers of deaths and crude MRs per 1000 person-years, according to baseline CD4 count and time since start of ART,   are shown in Table 2. Overall mortality per 1000 person-years declined from 33 (95% CI, 30-36) during the first 6 months of ART to 14 (95% CI, 13-15) among those who survived 10 years from start of ART, despite the aging of the study population over this period. Patients who started ART with a CD4 count <50 cells/µL experienced the greatest declines in MRs over time on ART, from 88 (95% CI, 78-99) per 1000 personyears during the first 6 months to 15 (95% CI, 13-17) per 1000 person-years after 10 years from start of ART. Although MRs were strongly inversely associated with baseline CD4 counts at shorter times since start of ART, the magnitude of these associations diminished with increasing duration of ART. Beyond 10 years of ART, there was little evidence of differences in crude MRs according to CD4 count at start of ART. Table 3 shows associations of CD4 count at start of ART with mortality at different durations since start of ART and averaged over all follow-up time, estimated from crude, adjusted, and relative survival models. Estimates are presented as MRRs with 95% CIs, with CD4 count 200-349 cells/µL as the comparator group. As expected crude, adjusted, and relative MRRs averaged over all follow-up time decreased with increasing CD4 count at start of ART. After adjusting for other prognostic factors, the MRR comparing baseline CD4 count <50 cells/µL with 200-349 cells/µL during the first 6 months of ART was 2.81 (95% CI, 2.12-3.71): this declined to 1.59 (95% CI, 1.31-1.92) 3-4.9 years after start of ART. During the first 5 years of ART, rate ratios comparing patients whose baseline CD4 count was >350 cells/ µL with those whose count was 200-349 cells/µL were <1. However, there was little evidence that baseline CD4 count was prognostic for mortality after 5 years of ART as the adjusted MRRs per CD4 category were 0.97 (95% CI, .94-1.00; P = .054) and 1.02 (95% CI, .98-1.07; P = .32) among patients followed for 5-9.9 and ≥10 years, respectively. Adjusted relative MRRs were generally further from 1 than the corresponding adjusted MRRs, indicating that failing to account for expected mortality partially obscured the association of CD4 count with mortality risk. Differences between the 2 analyses diminished with increasing time on ART because the strength of the association of mortality with baseline CD4 count also decreased with time.

DISCUSSION
In our collaborative analysis of 37 496 HIV-infected patients from 18 cohorts starting triple-combination ART between 1996 and 2001, there was a strong inverse association of CD4 count with mortality during the first year of ART, which diminished over the next 4 years. From 5 years after start of ART, baseline CD4 count was of little prognostic value. Even patients who started ART with very low (<50 cells/µL) CD4 counts may experience convergence of their mortality risk to that of patients with intermediate (200-349 cells/µL) or high (≥500 cells/µL) baseline CD4 count, from 5 years after the start of treatment. As expected,   associations of baseline CD4 count with mortality were attenuated after adjusting for other prognostic factors at the time of starting ART. Associations moved away from the null after further adjusting for background mortality in relative survival models, but this did not impact our results substantially. The importance of CD4 count nadir as a prognostic factor for survival in HIV-infected individuals starting ART is well established [3]. However, the extent to which patients with low baseline CD4 count surviving the first years of treatment remain at an increased risk of death, compared to patients starting with higher baseline CD4 counts, has remained unclear. To our knowledge, our study is the first to demonstrate that patients who started treatment in later stages of HIV disease may expect their mortality risk to become similar to that of patients starting treatment with higher CD4 counts, if they survive the first 5 years of therapy. The majority of patients (85.7% [95% CI, 85.2%-86.3%]) who started ART with CD4 count <200 cells/ µL in this cohort collaboration do indeed survive 5 years. This is an important message for patients and physicians alike, as the moving from a higher-to a lower-risk group through long-term treatment adherence may be a powerful motivator. Our findings are consistent with previous studies which showed that current CD4 count is more important than baseline CD4 count [5,6].
For the first 5 years of ART, there was a graded inverse relationship between baseline CD4 count and mortality, which was consistent with our previous findings [12]. After 5 years of ART, >15% of the patients with a baseline CD4 count of <50 cells/µL had died. These results are consistent with the rate of immunological nonresponders to ART previously observed [13,14]. Patients with poor immunological response or multiple AIDS or non-AIDS morbidity continue to have a substantially increased mortality ratio after 3 years of treatment [14,15]. A low CD4 nadir is not only an established risk factor for death, but also for poor CD4 recovery [16][17][18]. This may reach a plateau after about 2 years of ART [18], although a very recent analysis indicates slow but steady long-term recovery of CD4 counts under continued treatment [15].
These findings raise the question of how to identify patients at high risk of mortality during the first 5 years of ART and what The relative survival model accounts for age-, sex-, and country-matched population mortality. Models were stratified by cohort. Adjusted and relative survival models were adjusted for sex, age, transmission risk group, AIDS at baseline, viral load, and year of starting ART.
strategies could reduce this risk. These could include screening programs and intensive adherence counseling. A previous study demonstrated that patients with a CD4 count<200 cells/µL at baseline are at a higher risk for AIDS in particular, but also for non-AIDS-related mortality [12]. In that study, malignancies and infections were the primary causes of death, and both occurred more frequently in patients with a CD4 count <200 cells/µL. The observed convergence of survival rates after 5 years may be partly explained by the early death of patients with low to very low baseline CD4 count, especially immunological nonresponders, whereas others gradually achieve immunological recovery. The suggestion of a lower risk among those who started ART with very low compared with intermediate CD4 count after 10 years may reflect a cohort of survivors who are adherent and respond well to therapy, but is also consistent with a chance finding. We analyzed large numbers of patients from high-quality clinical cohorts. We obtained more accurate comparisons by using relative survival models that adjust for both patient characteristics and the background age, sex, and country general population mortality. Adjusting associations of CD4 count with mortality for age and sex as covariates in a conventional analysis that only considers mortality of those with HIV infection may be inducing bias toward the null. Although additionally adjusting for background mortality increased associations of baseline CD4 count with both cumulative mortality and early mortality, it did not affect our conclusions about the convergence of mortality in those who survived 5 years from starting ART.
Our study has some limitations. We chose patients starting ART between 1996 and 2001 to allow ≥10 years of follow-up for each person included. Whereas only such patients have been treated with long-term ART, newer drugs that have been introduced since 2002 are associated with better immune recovery [19] and/or CD4 recovery in immunological nonresponders [19][20][21], compared with drugs on which the patients in our study started combination ART. More second-line treatment options and new formulations with improved convenience have become available, allowing better treatment adherence and successful treatment of drug-resistant infections. Therefore, patients across all baseline CD4 count strata are likely to experience lower mortality than that of the patients included in our study.
While we included a broad range of patients with different risk factors from Europe and North America, generalizability to patients treated in other settings may be limited. Our results may be biased by loss to follow-up. Although loss to follow-up rates were high, cohorts have good ascertainment of death procedures and it is therefore more likely that patients transferred care to another center not in ART-CC or dropped out of care, rather than died [22]. However, our results should be generalizable to individuals remaining in HIV care. The high death rate observed in patients with unknown risk transmission is linked to the high death rate in older patients because risk group was poorly recorded in cohorts that included more patients aged >60 years. Data on non-HIV-related factors (eg, socioeconomic disadvantage or lifestyle risk factors) were not available, but may have contributed to excess mortality during the first 5 years of treatment in patients with low baseline CD4 counts.
In conclusion, CD4 count at start of ART strongly predicts MRs during the first 5 years of ART. This finding reinforces the need for earlier diagnosis and treatment of people living with HIV. However, there is little evidence that CD4 count at start of ART predicts mortality after 5 years of ART. This is a positive message for patients: The burden of increased mortality associated with starting treatment late is alleviated for those who survive 5 years on ART.

Supplementary Data
Supplementary materials are available at http://cid.oxfordjournals.org. Consisting of data provided by the author to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the author, so questions or comments should be addressed to the author.

Notes
Basilea, Merck/MSD, Gilead, and Pfizer; personal fees from Basilea, Merck/ MSD, Gilead, and Pfizer; and nonfinancial support from Astellas, Merck/ MSD, and Gilead. M. C. reports grants from Gilead and ViiV. G. A. B. reports grants from Bristol-Myers Squibb, and research support to institution from Amgen and consultant fee from Definitcare, LLC. H. M. C. reports grants from NIH and Patient-Centered Outcomes Research Institute. T. R. S. reports grants from Wellcome Trust. J. M. M. reports research and academic grants, personal fees, and fees for lectures from AbbVie, BMS, Gilead, Merck, Novartis, Pfizer, and ViiV Healthcare. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.