CD4+ count and risk of non-AIDS diseases following initial treatment for HIV infection

Abstract

Background

Reductions in AIDS-related morbidity and mortality following the advent of combination antiretroviral therapy have coincided with relative increases in chronic non-AIDS end-organ diseases among HIV+ patients.

Objective

To examine the association of latest CD4+ counts with risk of non-AIDS diseases in a cohort of 1397 patients who initiate antiretroviral therapy.

Methods

CD4+ counts and HIV RNA levels along with fatal, and non-fatal, AIDS and non-AIDS diseases (liver, cardiovascular, renal, and cancer) were assessed over a median follow-up of 5 years. Cox proportional regression models were used to study risk associations.

Results

A total of 227 patients experienced an AIDS event and 80 patients developed a non-AIDS disease event. Both AIDS and non-AIDS diseases rates (events/100 person-years), respectively, declined with higher latest CD4+ counts: 13.8 and 2.1 with latest CD4+ counts less than 200 cells/μl; 2.0 and 1.7 for counts of 200–350 cells/μl; and 0.7 and 0.7 for counts greater than 350 cells/μl. After adjusting for baseline covariates and the latest HIV RNA level, risk of AIDS and non-AIDS diseases were lowered by 44% (95% confidence interval for hazard ratio 0.50–0.62, P <0.01) and 14% (95% confidence interval for hazard ratio 0.77–0.96, P =0.01), respectively, for each 100 cell/μl higher latest CD4+ count.

Conclusion

Higher CD4+ counts on antiretroviral therapy are associated with lower rates of non-AIDS diseases and AIDS. These findings expand our understanding of the implications of HIV-related immunodeficiency and motivate randomized studies to evaluate the effects of antiretroviral therapy on a broad set of clinical outcomes at CD4+ counts greater than 350 cells/μl.

Introduction

Combination antiretroviral therapy (ART) as treatment for HIV-1 infection generally leads to CD4+ cell recovery [1], and corresponding reductions in AIDS-defining illnesses and mortality [2]. As a consequence of prolonged survival [3,4], morbidity and mortality among HIV-positive patients are increasingly due to non-AIDS diseases, including liver [5,6], cardiovascular [6–8], and renal [9] diseases, and certain cancers [10–12]. In fact, non-AIDS mortality is now the major cause of death among HIV-infected patients receiving ART [6,8,13,14].

Reasons for the apparent increased importance of non-AIDS diseases among patients with HIV are likely due to multiple factors. The HIV-infected population is aging, and higher risk of these diseases in the presence of reduced risk of AIDS would be expected. Non-AIDS disease risk may be increased by coinfection with viruses such as human papillomavirus (HPV) and hepatitis B or C [15–17], and a large fraction of HIV patients smoke cigarettes or have other risk factors for non-AIDS diseases [7,12,18,19]. Non-AIDS disease risk may also be increased as a consequence of HIV and ART. Concerning the latter, duration of protease inhibitor treatment has been associated with an increased risk of myocardial infarction [20]. Recommendations to defer ART until CD4+ counts decline to less than 350 cells/μl are largely because of concern about adverse events that include non-AIDS diseases associated with ART. Recent findings from the Strategies for the Management of Antiretroviral Therapy (SMART) study, however, raise questions about the balance of risks and benefits of ART for non-AIDS diseases. In SMART, episodic use of ART guided by CD4+ cell count was compared with continuous ART: episodic ART resulted in an increased risk of liver, cardiovascular, renal and cancer events, both fatal and non-fatal [21,22]. Thus, even though ART may increase the risk of some non-AIDS diseases, this risk may be greater if HIV infection is not treated.

Motivated by findings from SMART, we examined whether higher CD4+ counts with ART use are associated with a reduced risk of non-AIDS diseases (liver, cardiovascular, renal, and cancer) among HIV-infected patients taking ART. Our aim was to quantify this association and to compare it to the association of follow-up CD4+ count with risk of AIDS. The Flexible Initial Retrovirus Suppressive Therapies (FIRST) trial, conducted by the Community Program for Clinical Research on AIDS (CPCRA), was well suited to study this relationship in that both fatal and severe nonfatal AIDS and non-AIDS events were collected in a standardized manner over several years following initiation of ART.

Methods

Design

We used follow-up data (median of 5 years) for patients enrolled in FIRST. The study design and primary results of FIRST have been reported [23,24]. Briefly, between 1999 and 2002, 1397 ART-naïve HIV-infected participants were randomized equally to one of three ART strategies [nucleoside reverse transcriptase inhibitors (NRTI) plus protease inhibitors (PI), NRTI plus a non-nucleoside reverse transcriptase inhibitor (NNRTI), or the use of all three ART classes]. Following randomization and initiation of ART, participants were seen at month 1, month 4, and every 4 months thereafter for protocol-required data collection. At these visits, standardized data collection forms were completed for clinical endpoints, CD4+ counts were measured locally and HIV RNA levels were determined centrally (Roche Amplicor 1.0, Nutley, New Jersey, USA).

Outcomes

All new or recurrent AIDS events, deaths from any cause, non-fatal cardiovascular, renal, and liver disease events, malignancies, and other serious non-AIDS clinical events were ascertained. Qualifying AIDS events included presumed and definitive cases defined by US Centers for Disease Control and Prevention (CDC) AIDS criteria (1993) [25], adapted to include additional conditions related to immunodeficiency (footnote in Table 1). Non-AIDS disease events included the following fatal and non-fatal conditions: liver (cirrhosis by histology, as cites, esophageal/gastric varicies, hepatic encephalopathy or death from liver failure); cardiovascular (myocardial infarction, stroke, coronary artery intervention, or death from chronic atherosclerotic cardiovascular disease); renal (end-stage renal disease or death from chronic kidney disease); non-AIDS cancers (all cancers excluding Kaposi sarcoma, lymphoma, and invasive cervical cancer). Analyses were also considered for all-cause mortality and an expanded non-AIDS disease outcome that combined non-AIDS diseases (as above) with additional non-AIDS events ascertained in FIRST. Theses included congestive heart failure, coronary artery disease requiring drug treatment, myocarditis, pericarditis, pancreatitis, and bacterial pneumonia (single episode, non-AIDS defining). Structured case report forms were completed for AIDS and non-AIDS disease events. A Clinical Events Committee (CEC) reviewed and adjudicated AIDS events using prespecified criteria (see Appendix online). Deaths associated with both an AIDS and non-AIDS event were categorized as AIDS-related.

Table 1.

Baseline characteristics of FIRST cohort and by selected outcomes.

Group	FIRST	Outcome
		AIDS	Non-AIDS	Liver	Cardiovascular	Renal	Cancer (non-AIDS)
Patients	1397	227	80	14	24	14	32
Deaths	188	89	27	8	5	4	10
Age (median)	38	38	41	41	44	40	41
Sex (% female)	21	23	16	7	21	14	16
Race/ethnicity
Black (%)	54	59	56	21	58	93	53
Latino (%)	17	20	9	21	4	0	9
White (%)	26	19	33	57	38	0	34
Other (%)	3	2	3	0	0	7	3
Intravenous drug use (%)	15	18	24	36	25	21	19
Hepatitis B or C (%)	25	32	40	79	29	50	31
Prior AIDS (%)	38	67	48	57	50	57	34
Baseline CD4 (cells/μl)	163	43	114	170	82	30	146
Median/IQR	36–332	14–132	24–257	37–288	19–248	11–251	45–266
Baseline RNA (log10 copies/ml)	5.2	5.5	5.3	5.1	5.4	5.2	5.3
Median/IQR	4.6–5.6	5.0–5.9	4.8–5.7	4.8–5.3	4.7–5.7	4.8–5.7	4.6–5.8

Non-AIDS: non-fatal and fatal events related to liver, cardiovascular, and renal diseases and non-AIDS defining cancers. AIDS: non-fatal and fatal events adapted from 1993 CDC AIDS criteria to include additional conditions: invasive aspergillosis, bartonellosis, Chaga’s disease (American trypanosomiasis) of the central nervous system, disseminated herpes zoster, visceral Leishmaniasis (Kala-Azar), Hodgkin lymphoma, non-Hodgkin’s lymphoma (all cell types), chronic intestinal microsporidiosis (>1 month), Nocardiosis, extrapulmonary Penicillium marneffii, extrapulmonary Pneumocystis jerovecii, and Rhodococcus equi disease. IQR, interquartile range.

Statistical analysis

Cox proportional hazards regression models were used to study the relationship of AIDS, non-AIDS disease, and the other outcomes of interest, with baseline covariates corresponding to age, sex, race, baseline CD4+ count and HIV RNA level (log₁₀ transformed), prior AIDS events, and coinfection with hepatitis B or C virus and time-dependent covariates corresponding to follow-up CD4+ count and HIV RNA levels (<50 versus ≥50 copies/ml). To examine a possible non-linear association with latest CD4+ count, we added a quadratic term and compared the fit of that model to the model with the linear term. Models were also fit using different CD4+ categories instead of continuous CD4+ count. Follow-up data were censored either when patients were lost to follow-up, on the closing date of the study (16 September 2005), or for death not attributable to the event considered. Patients who experienced both AIDS and non-AIDS diseases were counted for both outcomes when they were considered separately. Associations with latest CD4+ count did not vary significantly by randomized treatment group in FIRST, therefore analyses are pooled over the three treatment groups [26]. Laboratory measurements prior to events are referred to as ‘latest,’ and measurements prior to initiation of ART are referred to as ‘baseline’ (obtained at visits prior to randomization). Findings are summarized with hazard ratios (HR)and 95% confidence intervals (CI). HR for continuous latest CD4+ count corresponds to 100 cell/μl higher counts. Analyses were also carried out with log₂ transformed CD4+ count. HRs from these models associated with a doubling of CD4+ count are cited. Statistical analyses were performed using SAS (Version 8.2). All rates are per 100 person-years of follow-up and all reported P-values are two-sided.

Results

Patient characteristics

The 1397 patients in FIRST had a median baseline CD4+ count of 163 cells/μl [interquartile range (IQR): 36, 332 cells/μl]. Of those assigned to a PI-based regimen, 62% were prescribed nelfinavir with another 24% using ritonavir-boosted PI. Ritonavir-boosted PI use increased during follow-up and was used in over 50% of patients by the end of the trial. As part of the initial ART regimen efavirenz accounted for 58% of NNRTI use, and the combination of zidovudine and lamivudine accounted for 53% of NRTI use.

Patients were followed for a median of 60 months; all patients were to be followed at least 42 months. Approximately 70% of patients achieved an HIV RNA level less than 50 copies/ml during the first year of ART. The mean of CD4+ counts after 32 months was 444 cells/μl for all patients and for those with baseline levels of less than 200, 200–350, and greater than 350 cells/μl, it was 335 (n =660), 487 (n =260), and 666 cells/μl (n =276), respectively.

Baseline characteristics for all patients in FIRST and for those who developed an AIDS (227 patients) or non-AIDS disease event (80 patients) during follow-up are presented in Table 1. Non-AIDS events included seven cases of cirrhosis, eight myocardial infarctions, seven strokes, and 11 patients with end-stage renal disease. The most common malignancies were skin (six; includes two melanoma), lung (five), and anal cancer (five). Deaths not attributable to AIDS or the non-AIDS diseases considered in Table 1 numbered 72. Of these, 16 were a result of sepsis/shock, 17 from respiratory failure or pneumonia, 17 from other cardiovascular causes (arrhythmia, heart failure, pulmonary embolus, or aneurysm), and 20 were due to unknown causes.

Associations of baseline factors with AIDS and non-AIDS diseases

In univariate analyses, a 100 cells/μl higher CD4+ count and a 1 log₁₀ lower HIV RNA level at baseline were associated with a lower risk of AIDS (CD4+ HR 0.60, 95% CI 0.53–0.67; HIV RNA HR 0.50, 95% CI 0.41–0.61) and non-AIDS disease (CD4+ HR 0.84, 95% CI 0.74–0.96; HIV RNA HR 0.80, 95% CI 0.59–1.07). A prior AIDS event before study entry and coinfection with hepatitis B or C virus was associated with a higher risk of AIDS (prior AIDS HR 3.89, 95% CI 2.95–5.13; hepatitis B or C coinfection HR 1.49, 95% CI 1.13–1.97) and non-AIDS events (prior AIDS HR 1.66, 95% CI 1.07–2.59; hepatitis B or C HR 2.17, 95% CI 1.39–3.40).

Multivariate models including all baseline covariates resulted in significant associations between prior AIDS diagnosis and AIDS (HR 2.12, 95% CI 1.55–2.89), and between older age (per 10 years) and non-AIDS diseases (HR 1.70, 95% CI 1.35–2.13). Further, older age (by 10 years) was independently associated with a greater risk of cardiovascular (HR 1.89, 95% CI 1.28–2.80) and non-AIDS cancer (HR 2.29, 95% CI 1.64–3.21) events. Hepatitis B or C virus coinfections were associated with liver events (HR 13.04, 95% CI 3.50–48.49); 9 out of the 14 events occurred in those coinfected with hepatitis B or C virus. Finally, black ethnicity was associated with a higher risk of renal events (HR 11.60, 95% CI 1.50–89.50).

Associations of latest HIV RNA levels with AIDS and non-AIDS diseases

In univariate analyses, a latest HIV RNA level less than 50 copies/ml, compared with at least 50 copies/ml, was also associated with lower risk for AIDS (HR 0.24, 95% CI 0.17–0.35) and non-AIDS diseases (HR 0.48, 95% CI 0.30–0.78). After adjusting for latest CD4+ count and baseline covariates, latest HIV RNA less than 50 copies/ml remained associated with a lower risk of AIDS (HR 0.46, 95% CI 0.32–0.68) and non-AIDS diseases (HR 0.52, 95% CI 0.31–0.87).

AIDS and non-AIDS disease risk by latest CD4+ count

Event rates were compared across three latest CD4+ categories (less than 200, 200–350, greater than 350 cells/μl; Figs. 1 and 2). Rates (and 95% CI) of AIDS and non-AIDS diseases, respectively, were 13.8 (11.8–15.9) and 2.1 (1.4–2.9) with latest CD4+ counts less than 200 cells/μl, 2.0 (1.3–2.8) and 1.7 (1.0–2.4) for counts of 200–350 cells/μl, and 0.7 (0.4–1.0) and 0.7 (0.4–1.0) for counts greater than 350 cells/μl. Non-AIDS disease rates decreased with increasing latest CD4+ levels, though to a lesser degree than for AIDS events. The steeper decline in risk for AIDS, compared with non-AIDS diseases, is evident in the unadjusted HRs. Above latest CD4+ counts of 350 cells/μl, rates for non-AIDS disease are similar to those for AIDS (Fig. 1). This is further illustrated in Fig. 2, which presents the proportion of total events that were due to AIDS or non-AIDS events by latest CD4+ count strata. AIDS events occurring at CD4+ counts greater than 350 cells/μl consisted primarily of severe/chronic herpes simplex disease (n =3), esophageal candidiasis (n =3), recurrent (less than 1 year) bacterial pneumonia (n =4), and lymphoma (n =3); non-AIDS events at higher CD4+ counts were more serious (11 non-AIDS cancer, three cirrhosis, and three coronary artery intervention). As expected events occurring at CD4+ counts less than 200 cells/μl tended to be more serious for AIDS (32 esophageal candidiasis, 25 Pneumocystis jirovedi pneumonia, 19 Mycobacterium avian complex, 17 crytococcosis, and 10 cytomegalovirus) and for non-AIDS (nine non-AIDS cancer, six stroke, and four end-stage renal disease).

Fig. 1. AIDS and non-AIDS disease rates by latest CD4+ count.

Rates of AIDS (black) and non-AIDS diseases (gray) per 100 person-years are presented across three categories of latest CD4+ counts. Error bars represent upper 95% CI, and the univariate hazard ratios (HR) with 95% CI for each CD4+ category (reference: less than 200) are also presented. Rates for both AIDS and non-AIDS diseases decline with higher CD4+ counts, though the risk gradient is steeper for AIDS. Non-AIDS includes non-fatal and fatal events related to liver, cardiovascular, and renal diseases and non-AIDS defining cancers.

Fig. 2. Proportion of AIDS and non-AIDS events by latest CD4+ count.

The proportion of severe events attributable to AIDS (black) and non-AIDS diseases (gray) is presented across three latest CD4+ count categories. Numbers of events are indicated within the respective bar, and total person-years at each CD4+ category are also presented. The relative frequency of non-AIDS events exceeds that of AIDS events at higher CD4+ counts. Non-AIDS includes non-fatal and fatal events related to liver, cardiovascular, and renal diseases and non-AIDS defining cancers.

Univariate and multivariate HR for AIDS and non-AIDS disease are shown in Fig. 3a. In multivariate analyses, a 100 cell/μl higher CD4+ count was associated with a 44% (95% CI for HR 0.50–0.62, P <0.01) lower risk for AIDS and a 14% (95% CI for HR 0.77–0.96, P =0.01) lower risk for non-AIDS diseases. A regression model with a quadratic term for CD4+ count fit better for AIDS (P <0.0001) but not for non-AIDS events (P =0.15). For AIDS, a 100 cell/μl higher CD4+ count was associated with a greater reduction in risk at lower CD4+ levels, for example, 54% for 200 versus 100 cells/μl, than higher levels, for example, 44% for 400 versus 300 cells/μl. This greater effect at lower levels is also evident from the HR cited in Fig. 1. We also considered the association with latest CD4+ after log₂ transformation. A CD4+ count doubling was associated with a 38% (P <0.01) lower risk of AIDS and a 20% (P <0.01) lower risk of non-AIDS diseases.

Fig. 3. Risk of event by latest CD4+ count.

Univariate (black) and multivariate (gray) hazard ratio (HR) estimates of risk for events are plotted per 100 cells/μl higher CD4+ count (a). ‘Non-AIDS’ includes nonfatal and fatal events related to liver, cardiovascular, and renal diseases and non-AIDS defining cancers. ‘All-cause mortality’ is death from any cause: ‘AIDS, non-AIDS, or Death’ is any non-fatal AIDS or non-AIDS event plus death from any cause. Multivariate HR are adjusted for latest HIV RNA and baseline covariates: age, sex, race/ethnicity, prior AIDS, and hepatitis B or C virus coinfection. Risk for individual non-AIDS diseases per 100 cells/μl higher CD4+ count are also presented (b). Point estimates to the left of 1 suggest a reduced risk at higher CD4+ levels.

A 100 cell higher CD4+ count was associated with a 35% (95% CI for HR 0.59–0.72) lower risk of death from any cause and a 30% (95% CI for HR 0.65–0.75, P <0.01) lower risk for the composite outcome of AIDS, non-AIDS, or death (Fig. 3a). We also examined two additional composite outcomes for non-AIDS diseases. The multivariate HR was 0.90 (95% CI 0.83–0.97) for a composite that included non-AIDS diseases plus any death not attributable to AIDS (145 patients). Using an expanded non-AIDS event composite combining non-AIDS diseases with other nonfatal cardiovascular events, pancreatitis, bacterial infection, and any death not attributable to AIDS (253 patients), the adjusted HR was 0.91 (95% CI 0.85–0.96).

Because of the smaller number of events, CIs for HR estimates of individual non-AIDS diseases were wider (Fig. 3b). With the exception of liver disease, multivariate HRs for separate events were less than one.

We tested for an interaction between latest CD4+ levels and latest HIV RNA level. There was evidence of a significant interaction for AIDS (P =0.01) but not for non-AIDS events (P =0.06). When the latest HIV RNA level was less than 50 copies/ml, the HR associated with a 100 cell higher CD4+ count was 0.69 (95% CI: 0.58–0.83) for AIDS and 0.74 (95% CI: 0.61–0.91) for non-AIDS. When the HIV RNA was at least 50 copies/ml, the corresponding HRs for AIDS and non-AIDS events were 0.51 (95% CI: 0.45–0.58) and 0.93 (95% CI: 0.81–1.05).

Discussion

Serious non-AIDS diseases, such as liver, cardiovascular, renal, and non-AIDS cancers, have contributed significantly to morbidity and mortality among HIV-infected patients because of the introduction of potent combination ART. It is unclear to what extent this is due to chronic immunosuppression, complications of ART, coinfection, or other established risk factors [7,19,27]. We examined the relationship between HIV-related immune suppression (latest CD4+ count) and key serious non-AIDS diseases during a 5-year median follow-up after initiation of ART in a cohort of 1397 HIV-infected individuals. Our findings indicate that: rates of non-AIDS diseases decrease with increasing CD4+ counts, though to a lesser degree than for AIDS events, whereas AIDS events dominate at CD4+ counts less than 200 cells/μl, non-AIDS disease are as common and likely correspond to greater morbidity than AIDS events among those who achieve CD4+ counts greater than 200 cells/μl with ART; and latest CD4+ count is associated with risk of non-AIDS diseases even after adjusting for additional risk factors.

Decreases in mortality rates among HIV-infected patients have been driven primarily by reductions in AIDS events [2,8,14]. Future reductions in HIV-related morbidity and mortality will require a better understanding of the risk factors associated with common non-AIDS diseases. Our investigation was motivated by findings from the SMART study that found intermittent use of ART, compared to continuous ART, was associated with a higher risk of AIDS and non-AIDS diseases (liver, renal, cardiovascular, and cancer) [21,22]. Our findings, among patients followed after initiating ART, are consistent with SMART in that a given CD4+ count difference predicts a greater difference in AIDS than non-AIDS disease. Our findings also suggest the incidence of non-AIDS diseases are similar to the incidence of AIDS at higher CD4+ counts (greater than 350 cells/μl). These estimates are likely conservative, because of our use of unambiguous endpoints as evidence of chronic end-organ disease. When our definition of non-AIDS diseases was expanded, the association with latest CD4+ count did not change appreciably. Although proportional risk reductions associated with a given difference in CD4+ count are lower for non-AIDS disease than AIDS, absolute risk reductions may be greater at higher CD4+ counts. Thus, if the potential reduction in non-AIDS risk as well as AIDS risk could be realized through earlier initiation of ART (at counts greater than 350 cell/μl), thereby maximizing CD4+ cell recovery and minimizing time spent at lower CD4+ levels, the public health benefit would be substantial.

Our findings are also consistent with studies that have investigated non-AIDS causes of death with latest CD4+ count. In the DAD (Data Collection on Adverse Events of Anti-HIV Drugs) study, non-AIDS cancer death now exceeds mortality from AIDS defining cancers, and the risk of death from liver disease or non-AIDS cancer was increased for those with lower latest CD4+ count [27,28]. Similarly, in the CASCADE (Concerted Action on SeroConversion to AIDS and Death in Europe) collaboration, a 100 cell/μl higher CD4+ count was associated with a 14% lower risk of death from non-AIDS malignancy, a 10% lower risk of death from liver disease, and an 11% lower risk of death from cardiovascular disease [6]. It is possible that CD4+ counts may decline just prior to death, as a consequence of certain non-AIDS diseases. This appears to be the case for liver disease [29]. Accounting for disease incidence and deaths, as in the present analysis, minimizes but does not eliminate this potential problem.

Future studies should focus on mechanisms underlying the influence of HIV-related immune suppression on non-AIDS disease risk. Although CD4+ counts and HIV RNA levels are established prognostic markers for AIDS [30] in untreated HIV, there is little information on these relationships with non-AIDS diseases. Though the primary purpose of this study was to examine the influence of CD4+ counts, non-AIDS disease risk was lower with latest HIV RNA levels less than 50 copies/ml. We also found that the risk association for CD4+ counts and non-AIDS diseases was similar to AIDS when latest HIV RNA was less than 50 copies/ml, whereas risk for AIDS at a given latest CD4+ level was relatively greater when HIV RNA was at least 50 copies/ml. These findings indicate that both CD4+ count and HIV RNA level are important determinants of risk of AIDS and non-AIDS diseases. The interrelationship of CD4+ count and HIV RNA in untreated cohorts, in which the range in HIV RNA levels will be greater, requires further investigation. Immune activation is an important driver of both HIV viral replication and CD4+ depletion [31,32]. Inflammatory biomarkers elevated during HIV infection [33] are also associated with progression of renal disease [34] and risk for cardiovascular events [35] and certain cancers [36]. The influence of chronic inflammation on the pathogenesis of atherosclerosis and cancer has been an area of intense investigation for a number of years. Depletion of CD4+ and CD8+ cells during HIV infection also impairs viral-specific immunity leading to progression of associated comorbid diseases. Viral hepatitis-associated liver disease is a clear example of this, as progression of hepatitis B or C related liver disease is accelerated by the degree and duration of CD4+ depletion [27,37].

Coinfection with pro-oncogenic viruses and impaired immune surveillance due to sustained CD4+ depletion may work together to increase risk of certain cancers during HIV infection.

Cancers associated with Epstein–Barr virus (primary central nervous system lymphoma, Burkitt’s lymphoma, primary effusion lymphoma, and Hodgkin’s Disease) and human papillomavirus (anogenital and cervical cancer) have known associations with HIV infection [38–40]. In support of this mechanism, cancer incidence was recently examined in a cohort of 28 855 patients before and after kidney transplantation, and rates of several types of cancer, many with viral causes, increased two fold after transplantation [41]. Observational studies suggest that both degree of CD4+ depletion and duration of HIV infection may increase risk for anal cancer [39] and other colorectal neoplasms [11]. Further elucidation of these risk associations may come from recent research suggesting that immune surveillance within the gastrointestinal tract may be particularly impaired due to the preferential depletion, and lack of recovery with ART, of mucosal CD4+ cells during HIV infection [42,43].

Strengths of this study include the use of a well-defined cohort followed in a consistent manner after initiation of ART, and the ascertainment of both non-fatal and fatal non-AIDS events. Sample size limited accurate estimates of risks for individual types of non-AIDS disease. In addition, detailed descriptions of non-AIDS disease risk by HIV RNA level were limited in that most patients had HIV RNA levels less than 50 copies/ml during follow-up.

Although ritonavir-boosted PI-based regimens were used as initial therapy in a minority of FIRST participants, CD4+ recovery was similar across randomized treatment strategies. In addition, treatment strategy did not alter non-AIDS disease risk by CD4+ count, but the risk assessment of specific drugs was limited as participants often changed the components of their ART regimen during follow-up. Finally, it should be emphasized that our finding of an association between latest CD4+ count and non-AIDS risk does not imply that use of ART to raise CD4+ count will reduce this risk. To establish that will require a randomized clinical trial.

In summary, we have established an association between latest CD4+ levels and risk for end-organ diseases not attributable to AIDS following initiation of ART. Further research is needed to establish whether HIV-related immune depletion truly leads to more frequent non-AIDS diseases, and to examine the underlying mechanisms. Ultimately, these findings support the need for a randomized clinical trial to determine the risk–benefit balance of starting ART at higher CD4+ counts.