| |
Cancers Now in HIV..."cancer mortality much higher in HIV+" - Cancer-Attributable Mortality Among People With Treated HIV in North America
|
| |
| |
Download the PDF here
Cancer-Attributable Mortality Among People With Treated Human Immunodeficiency Virus Infection in North America
CID July 3 2017 - Eric A. Engel et al
1National Cancer Institute, Bethesda and 2Information Management Services, Inc., Rockville, Maryland
Among PWHIV in North America treated with ART during 1995-2009, we estimate that 9.8% of all deaths were attributable to cancer. A larger proportion of deaths was attributable to NADCs (7.1%) than ADCs (2.6%). More than half of cancer- attributable deaths were attributable to NHL, lung cancer, and liver cancer; KS and anal cancer also contributed measurably.
Although we lacked data to assess cancer treatment in this study, PWHIV often have complex social and medical problems that prevent access to appropriate care [28].
In conclusion, we found that approximately 10% of deaths in PWHIV prescribed ART during 1995-2009 were attributable to cancer, but this fraction increased over time. A large proportion of these deaths were attributable to NHL, lung cancer, and liver cancer. Deaths due to NADCs will likely grow in importance as AIDS mortality declines further and PWHIV age.
⇒Cancer mortality among PWHIV estimated in our study (327 per 100000 person-years) was much higher than in the US general population during 2014 (186 per 100000 person-years; National Center for Health Statistics, http://www.cdc.gov/nchs/fastats/deaths.htm).
⇒Our results highlight important opportunities to reduce cancer mortality for PWHIV receiving ART. First, clinical and public health efforts should focus on reducing cancer incidence, by facilitating adherence to ART, enabling smoking cessation, and treating HBV and HCV infections [32] (see also https://aidsinfo.nih.gov). Second, screening for lung, liver, and anal cancers in appropriate high-risk groups may identify cancers when they are more amenable to treatment [33, 34]. Third, HIV-infected cancer patients must be provided access to timely and effective cancer treatment that is coordinated by experienced HIV and cancer specialists [28].
------------------------
Abstract
Background.
Cancer remains an important cause of morbidity and mortality in people with human immunodeficiency virus (PWHIV) on effective antiretroviral therapy (ART). Estimates of cancer-attributable mortality can inform public health efforts.
Methods.
We evaluated 46956 PWHIV receiving ART in North American HIV cohorts (1995-2009). Using information on incident cancers and deaths, we calculated population-attributable fractions (PAFs), estimating the proportion of deaths due to cancer. Calculations were based on proportional hazards models adjusted for age, sex, race, HIV risk group, calendar year, cohort, CD4 count, and viral load.
Results.
There were 1997 incident cancers and 8956 deaths during 267145 person-years of follow-up, and 11.9% of decedents had a prior cancer. An estimated 9.8% of deaths were attributable to cancer (cancer-attributable mortality rate 327 per 100000 person-years). PAFs were 2.6% for AIDS-defining cancers (ADCs, including non-Hodgkin lymphoma, 2.0% of deaths) and 7.1% for non-AIDS-defining cancers (NADCs: lung cancer, 2.3%; liver cancer, 0.9%). PAFs for NADCs were higher in males and increased strongly with age, reaching 12.5% in PWHIV aged 55+ years. Mortality rates attributable to ADCs and NADCs were highest for PWHIV with CD4 counts <100 cells/mm3. PAFs for NADCs increased during 1995-2009, reaching 10.1% in 2006-2009.
Conclusions.
Approximately 10% of deaths in PWHIV prescribed ART during 1995-2009 were attributable to cancer, but this fraction increased over time. A large proportion of cancer-attributable deaths were associated with non-Hodgkin lymphoma, lung cancer, and liver cancer. Deaths due to NADCs will likely grow in importance as AIDS mortality declines and PWHIV age.
-------------------------
DISCUSSION
Among PWHIV in North America treated with ART during 1995-2009, we estimate that 9.8% of all deaths were attributable to cancer. A larger proportion of deaths was attributable to NADCs (7.1%) than ADCs (2.6%). More than half of cancer- attributable deaths were attributable to NHL, lung cancer, and liver cancer; KS and anal cancer also contributed measurably.
The PAFs for these cancers are a manifestation of both the elevated incidence of new cases in PWHIV and the high subsequent mortality associated with a cancer diagnosis (manifested in the strong HRs). With widespread ART use, KS and NHL incidence have declined over time [7]. However, NHL remains one of the most common cancers in PWHIV in the United States, along with lung and anal cancers [27]. In our PWHIV population, liver cancer was less common than anal cancer but it was more common among decedents, and the HR associated with mortality was much larger for liver cancer (Table 2). As a result, the PAF was more than twice as high for liver cancer (0.9%) than for anal cancer (0.4%). Mortality following cancer diagnosis is affected by the inherent aggressiveness of tumors and whether patients receive effective treatment. Although we lacked data to assess cancer treatment in this study, PWHIV often have complex social and medical problems that prevent access to appropriate care [28].
Cancer-attributable mortality differed noticeably across population subgroups (Table 3), which to some extent reflects variation in cancer incidence and trends in overall mortality. Mortality attributable to NADCs was much higher in males than females and increased strongly with age. Mortality rates attributable to ADCs and NADCs were highest at low CD4 counts, although the corresponding PAF for NADCs was not especially high, because of high overall mortality in severely immunosuppressed individuals. Likewise, a growing proportion of all deaths during 1995-2009 period were attributed to NADCs, reaching 10.1% of all deaths after 31 December 2005. This pattern largely reflects the strong decline in overall mortality among PWHIV along with increasing mortality attributable to NADCs.
The PAF calculations assume that the association between cancer and mortality reflects a causal relationship, so that the corresponding deaths would be prevented if cancer could be eliminated or its treatment improved. A key assumption is thus that there is no unmeasured confounding of the association between cancer and mortality. We adjusted the Cox models for various demographic and clinical characteristics, and while the HRs changed noticeably in some instances, the unadjusted and adjusted PAFs were very similar. This similarity arises because the factor (HR − 1)/HR used in the PAF calculation is close to 1.00 whenever the HR is large, so that PAF≈Pd. However, if there were substantial residual confounding, our PAF estimates would not accurately capture the mortality caused by cancer.
Our overall PAF estimates are noticeably lower than those for cancer-related mortality presented in studies based on ascertained causes of death, that is, 7%-15% deaths reported from ADCs and 12%-27% from NADCs [2, 11-13, 18, 19]. Nonetheless, there are several reasons why we do not believe the PAF calculations substantially underestimated mortality due to cancer in our population. Notably, the PAF must always be less than Pd, because there cannot be more deaths attributable to cancer than there are deaths preceded by cancer. Thus, the maximum possible cancer-attributable mortality for our HIV population would be Pd = 11.9%, which is not far from our PAF estimate of 9.8%. It is unlikely that cancers were greatly underascertained in NA-ACCORD, because participating cohorts made substantial efforts to identify cancer cases, and standardized incidence ratios (comparing cancer risk in NA-ACCORD to the general population) are similar to those from other published estimates [8, 29-31]. Future analyses of NA-ACCORD data on reported causes of death, which are currently being compiled, will complement our results and provide information on deaths related to other important conditions.
Our main analysis assessed PWHIV receiving ART who, at the start of follow-up, had no prior history of cancer. Because it takes time for such a cohort to accumulate incident cancers and subsequent deaths, this restriction may have led us to underestimate PAFs, especially for earlier calendar intervals. In a sensitivity analysis, we therefore also included people with preexisting cancer, which increased the proportion of PWHIV in the cohort who died with a prior cancer, but the overall PAF increased only to 12.7%. Thus, our exclusion of PWHIV with prevalent cancer probably cannot entirely explain the above-mentioned differences with other studies. Our results could differ from those from prior studies if some deaths were misclassified in those studies, due to lack of detailed clinical information or the presence of multiple illnesses. Alternatively, our generally lower estimates could reflect differences in the HIV populations in terms of demographics, comorbid medical conditions, receipt of ART, or access to cancer care.
Important strengths of our study include the availability of a large representative population of PWHIV and validated data on cancers and deaths. We briefly note that Simard et al. previously estimated cancer-attributable mortality among people with AIDS in the United States during 1996-2006 [23] using a different PAF formula that required assigning cancer status at the start of discrete 6-month intervals. Because cancers may have occurred within the intervals, the Simard et al approach likely missed some cancer-associated deaths. In contrast, the current PAF method evaluated the association between cancer and death over a continuous timescale, so that deaths that occurred shortly after cancer diagnosis were appropriately assessed.
A limitation of our study is that we did not have validated data on incident cancers after 2009, so we could not estimate cancer-attributable mortality in more recent years. Two additional points should be discussed. First, our results resemble an "intent-to-treat" analysis of cancer-attributed mortality among PWHIV prescribed ART, that is, individuals were considered to be receiving ART continuously after initiation. This approach was used because patients are supposed to continue ART after initiation, and complete and accurate adherence information is difficult to obtain. However, we expect that not all patients were fully adherent, and some patients had documented periods of nonuse. Second, we examined only first cancers, so deaths were attributed to these rather than any subsequent malignancies. Subsequent cancers were somewhat uncommon (only 96 of 1997 PWHIV [4.8%] with a first cancer were later diagnosed with another). Also, to the extent that the subsequent cancers were caused by treatment of the first cancer (e.g., effects of radiation or chemotherapy-induced immunosuppression), it would be appropriate to attribute any deaths to the first cancer.
Cancer mortality among PWHIV estimated in our study (327 per 100000 person-years) was much higher than in the US general population during 2014 (186 per 100000 person-years; National Center for Health Statistics, http://www.cdc.gov/nchs/fastats/deaths.htm). Our results highlight important opportunities to reduce cancer mortality for PWHIV receiving ART. First, clinical and public health efforts should focus on reducing cancer incidence, by facilitating adherence to ART, enabling smoking cessation, and treating HBV and HCV infections [32] (see also https://aidsinfo.nih.gov). Second, screening for lung, liver, and anal cancers in appropriate high-risk groups may identify cancers when they are more amenable to treatment [33, 34]. Third, HIV-infected cancer patients must be provided access to timely and effective cancer treatment that is coordinated by experienced HIV and cancer specialists [28].
In conclusion, we found that approximately 10% of deaths in PWHIV prescribed ART during 1995-2009 were attributable to cancer, but this fraction increased over time. A large proportion of these deaths were attributable to NHL, lung cancer, and liver cancer. Deaths due to NADCs will likely grow in importance as AIDS mortality declines further and PWHIV age.
-----------------------
Introduction
Since the advent of effective antiretroviral therapy (ART) in 1996, mortality of individuals with human immunodeficiency virus (HIV) infection has declined substantially [1-3]. However, immune reconstitution may be incomplete with ART, and people with HIV (PWHIV) experience effects of chronic inflammation and immune activation [4-6]. Moreover, PWHIV continue to have elevated risk for some cancers [7, 8] due to immune abnormalities, coinfection with oncogenic viruses (eg, human papillomavirus, hepatitis B and C viruses [HBV, HCV]), and tobacco use [9, 10]. Cancers in PWHIV are typically divided into AIDS-defining cancers (ADCs: Kaposi sarcoma [KS], non-Hodgkin lymphoma [NHL], and cervical cancer), which can mark the onset of advanced immunosuppression, and non-AIDS-defining cancers (NADCs: all other cancers).
The decrease in overall mortality among PWHIV is largely explained by declining mortality from AIDS, although AIDS (including ADCs) still accounts for a sizeable proportion of deaths [2, 11-13]. As PWHIV live longer and age, the incidence of some NADCs has risen, and the overall burden of cancer has increased [7, 14]. Furthermore, HIV-infected cancer patients have worse survival from their cancer than HIV-uninfected cancer patients [15-17]. Recent studies of PWHIV in Europe, North America, and Australia, which have classified deaths based on clinical information or death certificate diagnoses, have assigned 7%-15% of deaths as due to ADCs and 12%-27% to NADCs [2, 11-13, 18, 19]. One challenge in interpreting these results is that it is difficult to assign a single underlying cause of death [20, 21], especially in people with multiple comorbidities. Furthermore, PWHIV may die outside hospitals, and low autopsy rates limit the information available to assign cause of death [12, 19].
Difficulties in determining cause of death for cancer patients have led to an emphasis on relative mortality methods to measure cancer-related mortality in other contexts [21]. These methods assess whether excess mortality is present in cancer patients compared with the general population and attribute any excess deaths to the cancer. Similarly, one can estimate the proportion of deaths in a population due to cancer premised on a statistical calculation of cancer-attributable mortality, that is, a population-attributable fraction (PAF) [22, 23]. This approach uses only information on the incidence of cancer and overall mortality in people with and without cancer, and it does not require clinical determination of causes of death.
Assessment of cancer mortality can help in the evaluation of the overall impact of cancer in PWHIV and target research, prevention, and treatment for maximum public health benefit.
In the present study, we analyzed data from a large North American consortium of HIV cohorts to estimate the cancer-attributable mortality among PWHIV treated with ART during 1995-2009.
|
|
| |
| |
|
|
|
