In this large collaboration of HIV seroconverter cohorts, non-AIDS defining causes contribute to more than half of deaths especially serious non-AIDS-defining infections, malignancies, end-stage liver, cardiovascular and respiratory disease. We provide evidence for an association between these causes of death and several markers of immunodeficiency, with the exception of cardiovascular deaths which were, however, associated with higher levels of viral replication.
Our analysis confirms that latest CD4 cell count is associated with a higher mortality from conditions not traditionally defined as opportunistic, i.e liver diseases, non AIDS-infections, respiratory diseases, as well as non AIDS-malignancies, even when CD4 count measured at least six months before death was considered. In addition, we found a consistent and gradually increasing association between earlier markers of immunodeficiency, such as the nadir CD4 count, or duration of exposure to immunodeficiency, such as time spent with CD4 < 350/mm3
and these specific causes of non-AIDS defining deaths. Our analysis does not indicate a particular advantage of either CD4 cell marker compared to the latest CD4 cell count for non-AIDS related outcomes. However, the consistent finding of a raised risk of non-AIDS defining death and cumulative time spent below CD4 < 350/mm3
reveals important clinically-relevant information and provides a strong argument that HIV-infected patients may benefit from early initiation of antiretroviral treatment to reduce the risk of AIDS and non AIDS-related causes of death [24
In this study, high HIV RNA, but not CD4 cell markers, was associated with a higher risk of cardiovascular disease death. High HIV RNA level may be considered as a surrogate marker of HIV-related endothelial inflammatory activation, such as Interleukin -6 or ultra-sensitive C-reactive protein, that may be promoted through various mechanisms [25
Our study shows in particular a specific association of immune depletion and mortality from liver and cancer which may be additionally related to the result of ageing [26
], HCV and hepatitis B virus (HBV) infections, tobacco, alcohol, and injecting drug use [27
] as well as to toxicities from antiretrovirals [30
Our study also shows that liver disease is a frequent cause of death in the cART era, now that patients are surviving to experience this competing cause of death. Indeed, immunosuppression is associated with more rapid progression to liver fibrosis and cirrhosis [13
]. Mortality rates associated with end-stage liver disease may rise even further in the future due to the ageing of the HIV population, co-infections with viral hepatitis B and C, and long-term cART-related hepatotoxicities [32
A recent meta-analysis comparing the incidence of 28 types of cancer between HIV-infected patients, organ transplant recipients, and the general population, suggested that immunodeficiency was the main cause of cancer development [33
]. Although HIV is usually not considered as an oncogenic virus, HIV infection may contribute to carcinogenesis by persistent inflammation mediated by cytokines especially as immunosuppression provides the immunologic background that favours the development of non-AIDS-defining cancer [16
] and leads to a reduced ability to the control of oncogenic viruses [35
Studies exploring the relationship between immunodeficiency and non-AIDS-defining causes of death, however, have provided limited or biased evidence because they were based on small cohorts [36
] or used models based on CD4 cell measurements immediately preceding death. This study confirms findings from prevalent cohorts reporting that the latest CD4 cell count and time since seroconversion were associated with deaths from both pre-AIDS [37
] and non-AIDS causes. In addition, results from the SMART trial clearly implicate immunodeficiency with the occurrence of several severe non-AIDS events, including major cardiovascular events, end-stage liver disease, malignancies and kidney complications [39
Seroconverter cohorts, such as CASCADE, are better suited than prevalent HIV populations for studying the determinants of survival. As seroprevalent cohorts are recruited at various times after seroconversion with unknown HIV infection duration, early deaths might be missed leading to an underestimation of death rates and biased estimates of prognostic factors. In our analyses, the long duration of follow-up acquired for this population of seroconverters allowed us to assess four different approaches to model ‘exposure’ to immunosuppression, which ensured that immunosuppression had preceded death, rather than simply reflected the effects of terminal disease. Nevertheless, as all patients were under follow-up shortly after seroconversion, and based on inclusion criteria they may represent a selected population with earlier HIV diagnosis, closer follow-up, optimal HIV management and favourable outcome, as compared to a recent publication from the same collaboration [1
]. In addition, patients started cART with a median CD4+ above 320/mm3, substantially higher than the level seen in many seroprevalent cohorts [41
], reflecting the fact that patients were managed in an optimal setting in the context of early HIV diagnosis. However, although the overall mortality and the cumulative incidence of specific causes of death might under-estimate that in the wider HIV-infected population [13
], it is unlikely that the association between immunodeficiency and deaths from different causes would be greatly affected by this selection. Finally, the patients selected for this analysis might better reflect patients that are currently managed and followed-up by practitioners, as attested by a larger proportion of HIV acquisition in the more recent periods.
The quality of our database depended on the accuracy of recording of cause of death in each cohort, with 82% of causes that could be classified. Deaths were standardized using the Coding of Causes of Death classification (www.cphiv.dk/CoDe
], which ensured homogeneous coding for the reason of death across cohorts. Nevertheless, we did not apply the central adjudication process as in controlled trials, due to the fact our categories of deaths were large and therefore misclassification unlikely. However, each cohort uses its own methods to collect source data and the process for selecting and validate an underlying cause of death may differ across cohorts, explaining that 18% of causes of death remained unknown. Anyway, in such a situation, the proportion of deaths attributed to AIDS have likely been overestimated as patients with different causes of death might have been more likely to be classified as AIDS-defining when their current CD4 cell count was low. As a result, this misclassification may lead to an under-estimation of the relationship between non AIDS-defining causes of death and immunodeficiency. Thus, we do not believe that this bias could explain the strong associations with immunodeficiency that we observed.
Treatment adherence was not available in the CASCADE dataset, however any clinically relevant effect would be captured by our primary exposure variables (HIV RNA and CD4 cell count). Besides, traditional risk factors for specific non-AIDS-defining causes of death, such as tobacco consumption, could not be included as they are currently not pooled in CASCADE. It is difficult, however, to see how many of these factors would confound any relationship with immunodeficiency.
In conclusion, our results add to the growing body of evidence on the association of immunodepletion and important non-AIDS related morbidity [47
] and mortality and underline the need for continued large scale collaborative HIV cohort collaborations. Our results also plead for the use of non-AIDS defining death and morbidities as endpoints in clinical trials evaluating antiretrovirals. As HIV-infected patients in industrialized countries have higher life style related risks factors (e.g. alcohol and tobacco consumption) compared to the general population [29
], programmes for aggressive management of life style related cardiovascular risk factors are paramount [48
]. The burden of infections points to the need for harm reduction programs especially for injecting drug users who are at higher risk of dying from non-AIDS-defining causes [50
] in addition to improved management strategies for the treatment of chronic HCV and HBV infections. Most importantly, our data provide additional arguments for HIV screening programs of at-risk populations for the early detection and treatment of HIV infection.