Prevalence and distribution of non-AIDS causes of death among HIV-infected individuals receiving antiretroviral therapy: a systematic review and meta-analysis

Abstract

The advent of antiretroviral therapy has significantly improved AIDS-related morbidity and mortality. Yet, among people living with HIV, deaths due to non-AIDS-defining illnesses have been on the rise. The objective of this study was to provide information about the global prevalence and distribution of non-AIDS causes of death in the last ten years among people living with HIV receiving antiretroviral therapy, by income levels of countries. We used broad search terms in Google Scholar, PubMed, and EMBASE to identify all studies that investigated the cause of death among people living with HIV receiving antiretroviral therapy, published after January 1, 2005. References were also identified from review articles and reference lists. Inclusion criteria were English language, the study’s end date was after 2005, all patients were HIV-positive, at least two-thirds of the patients were receiving antiretroviral therapy, at least one patient died of non-AIDS causes of death. Titles, abstracts, and articles were reviewed by at least two independent readers. Of 2951 titles identified in our original search, 151 articles were selected for further screening. We identified 19 studies meeting our full criteria, with patients from 55 different nations. Pooled non-AIDS causes of death prevalence estimates in high-income countries were 53.0% (95% confidence interval, 43.6–62.3), in developing countries 34.0% (95% confidence interval, 20.3–49.1), and in sub-Saharan countries 18.5% (95% confidence interval, 13.8–23.7). Statistically significant variation was noted within and between categories. Our findings show that a significant number of people living with HIV across the world die from cardiovascular disease, non-AIDS malignancies, and liver disease. There is a global need for further scrutiny in all regions to improve preventive measures and early detection according to distinct causes of death patterns.

Keywords

Non-AIDS HIV cause of death meta-analysis

Introduction

Expansion of access to antiretroviral therapy (ART) has changed the HIV epidemic trajectory by reducing AIDS-related mortality and morbidity among people living with HIV (PLWH).^1,2 The overall life expectancy among PLWH receiving ART is approaching that of the general population (life expectancy of PLWH in many developed countries is now near [or equal] that of the general population, however, there still remains quite a gap in sub-Saharan Africa).³ Some studies suggest, however, that incidence rates of diseases associated with advancing age in the general population are higher among PLWH.^4,5 With increasing life expectancy, non-AIDS related morbidities among these individuals are becoming all the more important. Cohort studies, either single cohorts or collaboration of cohorts, examining causes of death (COD) have reported increasing incidence of non-AIDS-related morbidity and mortality among PLWH.^6–8

The patterns of the increasing incidence of non-AIDS COD among PLWH in different settings are not necessarily the same. In countries where ART is more accessible, PLWH live longer, resulting in a shift in the cause of death. These COD include, but are not limited to, non-AIDS malignancies (NAM), cardiovascular disease (CVD), hypertension, diabetes mellitus, osteopenia/osteoporosis, and kidney and liver diseases.^6,7

Monitoring death among PLWH to identify specific causes can help public health officials to come up with targeted interventions. Therefore, many studies focused on surveillance of the patients exposed to these drugs for an extended period.^9,10 Some of them retrospectively reviewed medical records,¹¹ while others followed multiple cohorts for more than a decade to identify ART-related toxicities causing death.^6,7,10 This study aims to provide information about the global prevalence and distribution of non-AIDS COD in the last ten years among PLWH receiving ART, by income levels of countries.

Material and methods

Data sources

We searched Google Scholar, PubMed, and EMBASE for reports on non-AIDS COD among HIV patients receiving ART in English language journals published after 2005 using the following keywords: ‘non-HIV’ OR ‘non-AIDS’, ‘cause of death’ and ‘ART’ OR ‘HAART’. Studies were eligible to be included if the following criteria were met: 1) all patients were HIV-positive; 2) at least two-thirds of the patients were receiving ART; 3) the study’s end date was after 2005; and 4) at least one patient died of non-AIDS COD.

The search came up with 2480 results in Google Scholar, 70 results in PubMed, and 98 in EMBASE. Additionally, studies that cited or were cited by other pertinent studies were browsed for additional relevant references. To include studies from sub-Saharan Africa, we had to broaden the search by revising the keywords to ‘cause of death’, ‘HIV’ and ‘Africa’, which resulted in 16,900 results in Google Scholar, 215 in PubMed, and 568 in EMBASE. For each search query, Google Scholar can only show the first 1000 results. To ensure no studies were missed the search was repeated with different variations of the keywords. Two authors (AF and HM) assessed each study’s eligibility. We excluded duplicates of studies that appeared in the different databases. While reviewing the title and abstract, studies with topics outside of our scope were excluded, for example those that looked at the cost-effectiveness of ART, only focused on morbidity, or were studying risk factors associated with death. We also eliminated any study that specifically focused on a subgroup of people, such as children, pregnant women, intravenous drug abusers, and smokers. Studies that only examined one type of non-AIDS COD, ones that did not include specific CODs, or only included patients with co-infections (e.g. TB, hepatitis) were excluded. Only papers written in English were included. Of the abstracts reviewed, a total of 151 were selected for further screening. Disagreement was resolved by discussion with a separate author (MF). Our meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) statement.

Studies were further examined after the initial review to exclude the papers that combined the COD of patients with and without HIV. As for the publications by the ART Cohort Collaboration (ART-CC), Adverse Events of anti-HIV Drugs (D:A:D) and the Mortalité surveys in France that used an evolving database, the most recent and comprehensive paper was chosen, and the others were excluded. The ART-CC and D:A:D multi-cohort studies involved five of the same cohorts; however, since the majority of their data came from unique cohorts they were both included in our analysis. Papers were excluded that examined a single cohort involved in one of these collaborations. Authors whose data had missing or unclear information were emailed, five authors supplied the necessary data and those papers were included. This resulted in a total of 19 papers for the final analysis.

In studies which separated patients based on treatment, only patients receiving ART were included in the analysis. The total population size of the studies included participants from all years, but when possible only mortality data collected in the later years of the studies were extracted. From Chkhartishvili et al.¹² only data from 2005 to 2012, from Grinsztejn et al.¹³ only data from 2004 to 2009, from Schwarcz et al.⁹ only data from 2006 to 2011, and from Smith et al.⁷ only data from 2005 to 2011 were used. Since Falster et al.¹⁰ included cohorts from countries with varying income, data extracted were categorized into high-income (HIC) and developing (DC) country groups.

We divided the total COD into three categories: AIDS-related, non-AIDS-related, and unknown CODs. Subsequently, non-AIDS CODs were further differentiated by CVD, liver disease, NAM, renal, pulmonary, diabetes mellitus, infectious disease, ART drug-toxicity (we considered only the AIDS-related conditions [opportunistic infections and cancers] as AIDS-related, not the complications of HIV treatment), suicide, trauma, overdose, miscellaneous, and other (Table 2). In three papers, Karstaedt,¹⁴ MacPherson et al.,¹⁵ and Ogoina et al.,¹⁶ deaths were not categorized into AIDS and non-AIDS. For these papers, AIDS-related deaths were defined according to the CDC criteria, with the addition of non-specific infections, sepsis, and diarrheal diseases. When possible we kept all cancers together in the NAM category, however, three of the papers, Ingle et al.,⁶ Lee et al.,¹⁷and Morlat et al.,¹¹ defined hepatocellular carcinoma (HCC) as a liver disease. Since they did not specify the number of HCC cases from other liver diseases, for these papers we kept HCC in the liver disease category. Hepatitis was defined as a liver disease, and stroke was included in the CVD category. Adverse side-effects from ART were either labeled as such in the original study or as lactic acidosis. The category of trauma was defined as accidents and homicides. Three studies, Chkhartishvili et al.,¹² Ingle et al.,⁶ and Schwarcz et al.,⁹ categorized death due to substance abuse with overdoses. The COD labeled as other were defined as non-AIDS-related but gave no other additional information. The miscellaneous category includes COD defined as non-specified organ failure, epilepsy, pancreatitis, hematological disease, anemia, obstetric problems, CNS diseases, and gastrointestinal illnesses, as these were only present in a few studies.

Using the World Bank classification on the basis of income or region, the studies were categorized into three groups, namely high-income countries (HIC), developing countries (DC), and sub-Saharan countries (SSC).¹⁸

Statistical analysis

Prevalence of non-AIDS death among people receiving ART was computed by dividing the number of individuals dying from each non-AIDS-defining condition by the total number of deaths in the cohort; and for each category, number of deaths in that category divided by the total number of non-AIDS-related deaths in that study. We used metaprop command (Stata version 14.1) to pool prevalence and relative contributions.¹⁹ There are several advantages for using metaprop for this analysis. First, it allows us to include studies with proportions equal to zero or 100, while avoiding out-of-range confidence intervals (CIs). Second, because the random-effects model has been used to compute the pooled estimate, metaprop made it possible to perform a test of heterogeneity between groups in the sub-group analysis.¹⁹ It allows computation of 95% CIs using the score statistic and the exact binomial method and incorporates the Freeman-Tukey double arcsine transformation of proportions.²⁰ The program also allows the within-study variability be modelled using the binomial distribution. We used DerSimonian-Laird’s method²¹ to calculate the random-effects proportion of COD in each category. Study-specific and pooled CIs were computed using Wilson score interval.²² To evaluate heterogeneity between studies, we used I² statistics which describes the percentage of total variation across trial due to heterogeneity.²³

Results

We identified a total of 2951 abstracts from PubMed, Google Scholar, and EMBASE. After removing the duplicates, the abstracts were assessed for eligibility. There were 151 articles selected for further assessment, with 19 of included in the final analysis (Figure 1). There was a total 229,300 HIV patients from 55 countries evaluated in these studies.

Figure 1.

Flow diagram of search results.

Table 1 summarizes the setting, methodology, and characteristics of each study. There was a total of 11,373 patients (8699 from HIC, 1287 DC, and 1387 from SSC) in these 19 studies whose CODs were identified. Of the articles included in the analysis, seven studies were prospective and 12 studies retrospective. In prospective studies, five studies used either International Statistical Classification of Diseases and Related Health Problems tenth revision (ICD-10), Coding of Death in HIV (CoDe) for coding the COD among HIV patients, or the US Centers for Disease Control and Prevention (CDC) Classification System for HIV Infections (1993). Two studies used only CoDe,^7,13 one study used only ICD-10,²⁴ and two studies a combination of them.^6,10 Two prospective studies did not report their coding system.^26,30 Smith et al.⁷ was the largest prospective cohort with near 50,000 patients followed up from 1999 to 2011; and the smallest prospective cohort was in South Africa with 395 patients followed from 2004 to 2010.³⁰ To establish the COD, three retrospective studies used ICD-10,^9,11,27 one study CoDe,¹² one study CDC,²⁹ and two studies^1,25 a combination of classifications. The largest retrospective study involved a survey of 90 centers with approximately 82,000 HIV-infected patients at follow-up in France;¹¹ and the smallest one reviewed the death certificates of only 55 patients (we did not include the results related to 152 patients who had not received ART in this study) receiving ART in a major hospital in Nigeria.¹⁶ Three of the six SSC studies and one of the DC studies used verbal autopsy to establish COD; all were physician-coded rather than using automated methods.

Table 1.

Summary of studies.

									Deaths
Study	Years data collected	Location	Total follow-up times (PY)	Median baseline CD4 cell count	COD determination	Source of data	% on ART	Population	Total	AIDS- related	Non- AIDS- related
High-income countries
Falster et al.^10a,b	1999–2007	Asia-Pacific & Australia		290 (IQR 175,437)^c	CoDe^d, CDC	MR	100	2695	159	67	82
Leone et al.²⁴	2008–2009	Italy		244 (SD190)	ICD-10	CRF	100	1236	116	48	63
Stewart et al.^{25 e}	2006–2008	Canada		150 (IQR -,-)	ICD-10, CDC	DC, MR	>66^f		48	23	25
Lee et al.¹⁷	1998–2006	South Korea	6725	145 (IQR 44,277)		DB, MR	100	327	68	38	25
Ingle et al.⁶	1996–2009	Europe, North America	327,535	217 (IQR 96–340)	ICD-10, CoDe	DB, CRF	100	65,121	4237	1496	2078
Morlat et al.¹¹	2010	France		243 (IQR -,-)	ICD-10	MR	100	82,000	728	182	513
Schwarcz et al.^9b,e	1996–2011^g	United States			ICD-10	DB, DC, MR	89.1		1340	711	629
Smith et al.^{7 b}	1999–2011^h	Australia, Europe, United States	308,719	400 (IQR242–590)	CoDe	CRF	75.3	49,731	2003	508	1324
Developing countries
Falster et al.^10a,b	1999–2007	Asia-Pacific & Australia		223 (IQR 106,358)	CoDe^d, CDC	MR	100	1557	56	22	15
Pacheco et al.¹	1997–2006	Brazil	7037	250 (IQR -,338)	CoDe, CDC	AR, DC, MR	68	1538	226	111	98
Cuong et al.²⁶	2007–2010	Vietnam		41 (IQR 17,104)		MR, VA	100	640	60	49	11
Grinsztejn et al.¹³	1986–2009ⁱ	Brazil		264.5 (IQR 89.5,430.2)^c	CoDe	DB, MR	>66^f	3530	274	162	105
Chkhartishvili et al.^{12 b}	1989–2012^j	Georgia	13,572	238 (IQR 106,408)	CoDe	DB, MR	>66^f	3554	645	340	237
Bhattacharjya et al.²⁷	2006–2012	India		87(IQR -,-)^c	ICD- 10	MR	100	5612	370	247	36
Sub-Saharan countries
Mzileni et al.²⁸	2004–2006	South Africa				MR	100	2605	204	180	21
MacPherson et al.¹⁵	2005–2007	South Africa		93 (IQR 37,148)		MR, VA	100	153	106	86	20
Karstaedt^{14 b}	2004–2009	South Africa		26 (IQR-,-)		MR, VA	100	2943	305 ^k	228	69
Kouanda et al.²⁹	2003–2008	Burkina Faso		124 (IQR 60,183)	CDC	MR	100	5608	690	270	150
Ogoina et al.¹⁶	2006–2009	Nigeria		136 (IQR 56,199)		DC, MR	100	55	19	16	3
Otwombe et al.³⁰	2004–2010	South Africa		24 (IQR 10–35)		MR, VA	100	395	63	31	16

AR: autopsy reports; CA: clinical assessment; CDC: AIDS-related death as defined by the CDC; CRF: clinical research form; DB: database; DC: death certificate; MR: medical records; VA: verbal autopsy.

Data separated by country income.

Additional data provided by author.

Multiple categories reported.

Only used 2005–2007.

Study limited to deceased patients.

Universal access to ART.

Only included 2006–2011.

Only included 2005–2011 in COD data.

Only included 2004–2009 in COD data.

Only included 2005–2012 in COD data.

Multiple COD in 14 patients.

Table 2 reports the number of underlying non-AIDS-related CODs in the studies. The proportion of deaths in each study attributed to each cause is depicted in Figure 2. CVD was a common COD across the board. There were only two studies^25,27 with no reports of CVD. Malignancies and liver disease seemed to be more prevalent in HIC than in DC or SSC. Suicide and deaths from overdose accounted for a substantial portion of non-AIDS deaths in two studies^6,25 in the US. Deaths from ARV-related toxicity accounted for a significant number of deaths in SSC.^14,15,28,29

Table 2.

Non-AIDS causes of death by category.

Author	CVD^a	Liver	NAM	Pulmonary	Renal	DM	ID	DT	Suicide	Trauma	OD	Misc^b	Other^c	Unknown
High-income countries
Falster et al.^{10 ,d}	14	14	24	0	1	0	11	0	4	3	7	4	0	10
Leone et al.^24,e	8	19	17	0	0	0	9	0	0	0	0	0	10	5
Stewart et al.²⁵	0	3	13	2	0	0	0	0	0	0	0	3	4
Lee et al.¹⁷	5	5f	3	0	0	0	4	0	5	1	0	2	0	5
Ingle et al.⁶	291	302^f	461	61	37	5	314	11	97	61	191^g	65	182	663
Morlat et al.¹¹	73	45^f	197	8	6	0	68	0	34	0	6	26	54	33
Schwarcz et al.⁹	119	40	156	22	9		19	0	45	0	127^g	156	92
Smith et al.⁷	217	249	364	0	19	6	116	4	93^h	42	58	4	152	171
Developing countries
Falster et al.^10,d	4	1	2	0	0	1	4	0	1	0	0	2	0	29
Pacheco et al.¹	9	8	8	0	0	0	19	0	0	11	0	0	43	17
Cuong et al.²⁶	1	0	1	0	0	1	0	0	2	1	5	0	0
Grinsztejn et al.¹³	21	8	18	3	5	2	25	4	0	8	2	6	3	7
Chkhartishvili et al.¹²	36	106	39	6	6	0	10	0	4	5	14 ^g	11	0	68
Bhattacharjya et al.²⁷	0	0	3	0	0	0	0	0	6	0	0	9	18	87
Sub-Saharan countries
Mzileni et al.²⁸	6	6	0	0	0	0	0	12	0	0	0	0	0	3
MacPherson et al.¹⁵	2	3	3	0	1	1	6	2	0	0	0	2	1
Karstaedt¹⁴	13	0	7	16	3	0	0	10	1	2	0	14	3	22
Kouanda et al.²⁹	18	6	8	16	0	13	33	16	3	0	0	24	13	270
Ogoina et al.¹⁶	1	1	0	0	1	0	0	0	0	0	0	0	0
Otwombe et al.³⁰	3	0	0	0	0	0	12	0	0	0	0	4	0	16

CVD: cardiovascular disease; NAM: non-AIDS malignancies; DM: diabetes mellitus; ID: infectious disease; DT: ART drug-toxicity; OD: overdose.

Includes stroke.

Includes non-specified organ failure, epilepsy, pancreatitis, hematological disease, anemia, obstetric problems, CNS diseases, and gastrointestinal illnesses.

Non-AIDS undefined.

Data separated by country income.

Some numbers calculated from percentages.

Includes hepatocellular carcinomas.

Also includes death due to substance abuse.

Includes 13 cases of euthanasia.

Figure 2.

Proportion of non-AIDS causes of death among HIV patients.

Pooled non-AIDS COD prevalence estimates in HIC was 54.1% (95% CI, 46.1–62.0), in DC 28.1% (95% CI, 16.4–41.5), and in SSC 18.5 (95% CI, 13.8–23.7) (Figure 3). As noted, Q statistics and I² indicated heterogeneity of results across the categories (Q₂ = 53.6, p < .001, I²= 98.6). Statistically significant variation was noted within categories. The highest was in HIC (Q₂ = 283.4, p < .001, I²= 97.5), followed by DC (Q₂ = 144.1, p < .001, I²= 96.5) and SSC (Q₂ = 19.42, p < .001, I²= 74.2). There was a significant variation among studies in each category. In HIC category, prevalence of non-AIDS COD among PLWH in France¹¹ was at 70.4% (95% CI 67.0–73.8), while Lee et al.¹⁷ reported 36.8% (95% CI 25.4–49.3) in Korea. A similar pattern was in the studies conducted in the DC, with Bhattacharjya et al.²⁷ at 9.7% (95% CI 6.9–13.2) in India to 43.2% (95% CI 37.0–50.1) in Brazil;¹ in SSC Karstaedt¹⁴ reported 22.6 (95% 18.0–27.7) to 10.3 (95% 6.5–15.3).²⁹ Because of these high I² values, we performed several subgroups analyses (for example, by type of coding or type of the data) to address this issue (results not reported). However, because of the multifactorial nature of the heterogeneity and small number of studies in each subgroup, the subgroups still had high I² values, which indicate the complexity of the issue. For example Karastaedt¹⁴ and Mzileni et al.,²⁸ both from South Africa, reviewed medical records during similar time periods and reported significantly different proportions of non-AIDS COD in their studies (neither of them reported their coding system).

Figure 3.

Proportion of patients died of non-AIDS causes.

In the analysis by cause and category, there were no significant differences in the prevalence of death due to cardiovascular diseases (Figure 4a), with the pooled prevalence in the HIC (at 14.9% of all non-AIDS COD [95% CI, 12.6–17.3]), more than that in SSC (at 13.8% [95% CI, 8.7–19.8]) and in DC (at 11.1% [95% CI, 5.0–19.0]). In contrast, prevalence of malignancies as the COD was significantly higher in HIC (28.1% [95% CI, 23.0–33.3]) as compared to DC (7.1% [95% CI, 0.3–19.2]) or SSC (3.4% [95%CI, 0.4–8.1]) (Figure 4b). A similar pattern was seen for deaths due to liver diseases (Figure 4c), with the highest in HIC (14.3% [95% CI, 10.0–19.1]) followed by DC (8.4% [95% CI, 0.0–27.3]) and SSA (5.1% [95% CI, 0.0–16.3]). A high proportion, and wide variation, of deaths from liver disease in DC is mainly because of the study from Georgia, which proportionally has a much higher number of deaths from liver diseases. The highest prevalence of infectious diseases as the COD was noted in SSC (14.3% [95% CI, 0.0–40.9]), as compared to HIC (9.5% [95% CI, 5.6–14.3]) and DC (7.3% [95% CI, 0.0–23.9]) (Figure 4d). Death due to renal disease was found to have a low prevalence across the board (Figure 4e).

Figure 4.

(a) Proportion of non-AIDS deaths due to CVD. (b) Proportion of non-AIDS deaths due to malignancies. (c) Proportion of non-AIDS deaths due to liver diseases. (d) Proportion of non-AIDS deaths due to infectious diseases. (e) Proportion of non-AIDS deaths due to renal diseases.

Discussion

We identified evidence of a substantial proportion of HIV-infected individuals who have died of non-AIDS-defining conditions in both resource-rich and resource-limited settings. Although non-AIDS-defining conditions have been known to cause death among HIV-infected individuals, their anatomical and geographical distribution had not been systematically documented to this date.

There was a significant variation between and within categories, with the highest prevalence of non-AIDS-related COD in HIC and lowest in SSC. While the proportion of the patients died of CVD, liver and infectious diseases significantly varied among studies in each income category; the HIC studies were more homogenous in that regard than those from DC or SSC. As for non-AIDS-related malignancies, the biggest variation was in DC and the lowest was in SSC.

We found a significant portion of non-AIDS deaths due to CVD. Nonetheless, the explanation for cause in each income category could differ. In DC and SSC where patients are more likely to start ART when the disease is at an advanced stage, HIV-associated inflammation could potentially mediate the risk of CVD.^31,32 On the other hand, a systematic review shows that the relative risk of CVD among those on ART is twice that of HIV-negative people.³² As a result, those on ART are also at risk of CVD.

Higher numbers of liver disease in Georgia and HIC can be explained by the higher prevalence of risk factors such as pro-oncogenic viruses such as hepatitis B or C virus as compared to other studies in DC and SSC. Chronic diseases such as hepatitis C interacting with HIV or with alcohol use can lead to more rapid cirrhosis and more rapid development of hepatocellular carcinoma.³³ For example, in Georgia many IV drug abusers infected with HIV die of complications of chronic viral hepatitis.¹² Providing treatment for hepatitis B and C could help reduce liver-related mortality.

Although HIV can lead to several renal pathologies³⁴ and ARV drugs like tenofovir can damage kidney functions,³⁵ we could not find much evidence that renal disease causes any significant number of deaths among the HIV-infected population.

In HIV care, like other health-related issues, comparison of individuals living in high- vs. poor-resource settings should be done more carefully. The differences in COD could be a reflection of the disparity in the patients’ lifestyle and the kind of care they receive. In high-income settings, as the longevity of PLWH increases, their COD becomes more similar to that of the general population, where CVD, NAM, and chronic liver disease are the main COD.^6,7,9,11 On the other hand, in DC and SSC, despite the recent expansion of ART programs, early mortality rate is still high, mainly due to advanced immunosuppression at the time of ART initiation.^15,16,29,30 Patients’ outcomes depend on the type, duration, and quality of care they receive, which may significantly vary in different settings. HIV-infected people in HIC, in general, are more likely to receive better clinical management of HIV infection, including earlier ART initiation, as compared to DC and SSC. As a result, the patient in HIC can reach the age at which malignancies and CVD incidence increase.

Socio-behavioral risk factors, which could also account for the differences in COD, vary across categories.^9,12 In the HIC category, male patients, who comprise the majority of the studies’ population, are more likely to be involved in risk behaviors. Moreover, people aging with HIV infection are more likely to continue substance use (tobacco, alcohol, opioids, and other psychoactive substances),³⁶ which can cause CVD, NAM, or liver disease.

It should be noted that part of the heterogeneity seen in the results stems from substantial differences among studies in terms of time (both duration and year of data collection), magnitude, methodology, and outcome. In the HIC category, study populations are larger compared to those in SSC, and have a significantly longer follow-up. The ART Cohort collaboration (ART-CC), a collaboration of cohort studies of PLWH from Europe and North America, and the Data collection on Adverse events of anti-HIV Drugs (D:A:D) were the largest and longest cohorts, with a combined total of almost 100,000 patients being followed for more than a decade (four cohorts are shared by these two studies).

Limitations

There are some important limitations to this study that must be considered. Death is generally caused by a variety of factors, both underlying and contributing causes, rather than one single cause.³⁷ In many circumstances, particularly among PLWH, there is more than one cause that leads to a person’s death.¹⁴ Identifying the leading cause, especially with limited information, can be very challenging and in some cases even impossible.^38,39

Combination of the results of studies with various methodologies (retrospective and prospective studies) and quality (standard classification vs. unidentified classification method) have increased the statistical power of our analysis and may provide more information on the distribution of non-AIDS COD among PLWH; however, it can also introduce bias as some of these studies may suffer from misclassification. There was only one study¹⁷ in HIC and one study in DC^8,26 that did not mention classification methods to establish COD, while in SSC only one study²⁹ mentioned using CDC criteria. Selection of the single underlying COD, even using the set of rules in the ICD-10 instruction manual could be a daunting task. Not using standard methods, such as ICD-10 or CoDe, can significantly influence the level of accuracy of the findings of these studies. As a result, the studies that did not use ICD-10 or CoDe are more prone to misclassification.

Additionally, studies, particularly the retrospective ones, using death certificate to establish COD are also more likely to misclassify the COD. Reports of AIDS-related death could contain serious errors or be incomplete, whether it is in SSC³⁸ or in HIC,⁴⁰ which in turn can adversely affect identification of the underlying COD. Death certificates could be filled in erroneously if the mechanism of death rather than the underlying cause is listed.

The process for coding and selecting the underlying COD is complicated. This complex process requires a full electronic database of medical certification, an automated coding system, which may not be available in resource-limited settings. Moreover, accurate coding requires fully trained staff with detailed work and data flow. Without such, timely attendance of a skilled physician may not occur, possibly resulting in the death certificate containing incorrect coding.

The discrepancy in categorization in studies due to the approaches used when determining the COD (i.e. clinical vs. autopsy diagnosis) could also have introduced some bias due to misclassification errors.⁴¹

The majority of the SSC studies came from South Africa, which limits the generalizability of the findings. Secondly, in this region, many deaths occur outside healthcare settings and a cause is not recorded, so the deaths included in this study are likely to be atypical, again reducing their representativeness. Moreover, because of the stigma, some doctors may be reluctant to code a HIV-related cause of death, which may result in higher proportion of non-AIDS deaths.

Moreover, our study was limited to studies published in English, which could potentially introduce further bias, due to publication bias and exclusion of non-English reports. The findings of this study might also show an inflated rate of non-AIDS causes. Lost-to-follow-up is a major problem in many prospective studies. It has been shown that many of these patients who are lost to follow-up actually died, which results in the underestimation of overall mortality.⁴² The wide variation in lost-to-follow-up in different cohorts could explain part of the variation in these studies.

Conclusion

Our findings contribute to the growing number of studies demonstrating increasing proportions of deaths due to non-AIDS CODs and add to the literature by describing the distribution of these non-AIDS CODs across the world. Considering the fact that older age and a longer period of ART contribute to development of these underlying conditions or diseases, public health policies should be revised to accommodate the needs of a HIV-infected aging population according to regions. The elevated risk of death from CVD, non-AIDS malignancies, and liver diseases among PLWH in each region needs also further scrutiny to improve preventive measures and early detection according to distinct COD patterns.

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Pacheco

Tuboi

May

. Temporal changes in causes of death among HIV-infected patients in the HAART era in Rio de Janeiro, Brazil. J Acquir Immune Def Syndr 2009; 51: 624–630.

Farahani

Price

El-Halabi

. Trends and determinants of survival for over 200, 000 patients on antiretroviral treatment in the Botswana National Program: 2002–2013. AIDS 2016; 30: 477–485.

Van Sighem

Gras

Reiss

. Life expectancy of recently diagnosed asymptomatic HIV-infected patients approaches that of uninfected individuals. AIDS 2010; 24: 1527–1535.

Guaraldi

Orlando

Zona

. Premature age-related comorbidities among HIV-infected persons compared with the general population. Clin Infect Dis 2011; 53: 1120–1126.

Lang

Boccara

Mary-Krause

. Epidemiology of coronary heart disease in HIV-infected versus uninfected individuals in developed countries. Arch Cardiovasc Dis 2015; 108: 206–215.

Ingle

May

Gill

. Impact of risk factors for specific causes of death in the first and subsequent years of antiretroviral therapy among HIV-infected patients. Clin Infect Dis 2014; 59: 287–297.

Smith

Ryom

Weber

. Trends in underlying causes of death in people with HIV from 1999 to 2011 (D: A: D): a multicohort collaboration. Lancet 2014; 384: 241–248.

Tsertsvadze

Bolokadze

Gochitashvili

. Experience of antiretroviral treatment in Georgia. Central Eur J Public Health 2009; 17: 25–30.

Schwarcz

Hsu

. Changes in causes of death among persons with AIDS: San Francisco, California, 1996–2011. AIDS Patient Care STDs 2014; 28: 517–523.

10.

Falster

Choi

Donovan

. AIDS and non-AIDS related mortality in the Asia Pacific region in the era of combination antiretroviral treatment. AIDS 2009; 23: 2323–2323.

11.

Morlat

Roussillon

Henard

. Causes of death among HIV-infected patients in France in 2010 (national survey): trends since 2000. AIDS 2014; 28: 1181–1191.

12.

Chkhartishvili

Sharvadze

Chokoshvili

. Mortality and causes of death among HIV-infected individuals in the country of Georgia: 1989–2012. AIDS Res Hum Retroviruses 2014; 30: 560–566.

13.

Grinsztejn

Luz

Pacheco

. Changing mortality profile among HIV-infected patients in Rio de Janeiro, Brazil: shifting from AIDS to non-AIDS related conditions in the HAART era. PLoS ONE 2013; 8: e59768–e59768.

14.

Karstaedt

. Profile of cause of death assigned to adults on antiretroviral therapy in Soweto. SAMJ 2012; 102: 680–682.

15.

MacPherson

Moshabela

Martinson

. Mortality and loss to follow-up among HAART initiators in rural South Africa. Trans R Soc Trop Med Hyg 2009; 103: 588–593.

16.

Ogoina

Obiako

Muktar

. Morbidity and mortality patterns of hospitalised adult HIV/AIDS patients in the era of highly active antiretroviral therapy: a 4-year retrospective review from Zaria, Northern Nigeria. AIDS Res Treat 2012; 2012: 940580–940580.

17.

Lee

Kim

K-H

Lee

. Causes of death and risk factors for mortality among HIV-infected patients receiving antiretroviral therapy in Korea. J Korean Med Sci 2013; 28: 990–997.

18.

World bank. Country classification: classification of economies, Washington, DC: World Bank, 2015.

19.

Nyaga

Arbyn

Aerts

. Metaprop: a Stata command to perform meta-analysis of binomial data. Arch Public Health 2014; 72: 39–39.

20.

Freeman

Tukey

. Transformations related to the angular and the square root. Ann Math Stat 1950, pp. 607–611.

21.

DerSimonian

Laird

. Meta-analysis in clinical trials. Contr Clin Trials 1986; 7: 177–188.

22.

Newcombe

. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med 1998; 17: 857–872.

23.

Higgins

Thompson

Deeks

. Measuring inconsistency in meta-analyses. BMJ 2003; 327: 557–557.

24.

Leone

Gregis

Quinzan

. Causes of death and risk factors among HIV-infected persons in the HAART era: analysis of a large urban cohort. Infection 2011; 39: 13–20.

25.

Stewart

Chan Carusone

. Causes of death in HIV patients and the evolution of an AIDS hospice: 1988-2008. AIDS Res Treat 2012; 2012: 7–7.

26.

Cuong

Thorson

Sönnerborg

. Survival and causes of death among HIV-infected patients starting antiretroviral therapy in north-eastern Vietnam. Scand J Infect Dis 2012; 44: 201–208.

27.

Bhattacharjya

Sahu

Patel

. Causes of death among HIV-infected adults registered in selected anti-retroviral therapy centers in north-eastern India. World J AIDS 2015; 5: 90–99.

28.

Mzileni

Longo-Mbenza

Chephe

. Mortality and causes of death in HIV-positive patients receiving antiretroviral therapy at Tshepang Clinic in Doctor George Mukhari Hospital. Pol Arch Med Wewn 2008; 118: 548–554.

29.

Kouanda

Meda

Nikiema

. Determinants and causes of mortality in HIV-infected patients receiving antiretroviral therapy in Burkina Faso: a five-year retrospective cohort study. AIDS Care 2012; 24: 478–490.

30.

Otwombe

Laher

Tutu-Gxashe

. The effect of a maturing antiretroviral program on early mortality for patients with advanced immune-suppression in Soweto, South Africa. PLoS ONE 2013; 8: e81538–e81538.

31.

Hsue

Deeks

Hunt

. Immunologic basis of cardiovascular disease in HIV-infected adults. J Infect Dis 2012; 205: S375–S382.

32.

Islam

Jansson

. Relative risk of cardiovascular disease among people living with HIV: a systematic review and meta-analysis. HIV Med 2012; 13: 453–468.

33.

Thio

Seaberg

Skolasky

. HIV-1, hepatitis B virus, and risk of liver-related mortality in the Multicenter Cohort Study (MACS). Lancet 2002; 360: 1921–1926.

34.

Gupta

Smurzynski

Franceschini

. The effects of HIV-1 viral suppression and non-viral factors on quantitative proteinuria in the HAART era. Antiviral Ther 2009; 14: 543–543.

35.

Scherzer

Estrella

Yongmei

. Association of tenofovir exposure with kidney disease risk in HIV infection. AIDS 2012; 26: 867–867.

36.

Green

Kershaw

Lin

. Patterns of drug use and abuse among aging adults with and without HIV: a latent class analysis of a US Veteran cohort. Drug Alcohol Depend 2010; 110: 208–220.

37.

Kowalska

Friis-Møller

Kirk

. The coding causes of death in HIV (CoDe) project: initial results and evaluation of methodology. Epidemiology 2011; 22: 516–523.

38.

Nojilana

Groenewald

Bradshaw

. Quality of cause of death certification at an academic hospital in Cape Town, South Africa. SAMJ 2009; 99: 648–652.

39.

Burger

Groenewald

Bradshaw

. Validation study of cause of death statistics in Cape Town, South Africa, found poor agreement. J Clin Epidemiol 2012; 65: 309–316.

40.

Barchielli

Buiatti

Galanti

. Completeness of AIDS reporting and quality of AIDS death certification in Tuscany (Italy): a linkage study between surveillance system of cases and death certificates. Eur J Epidemiol 1995; 11: 513–517.

41.

Roulson

Benbow

Hasleton

. Discrepancies between clinical and autopsy diagnosis and the value of post mortem histology; a meta-analysis and review. Histopathology 2005; 47: 551–559.

42.

Bisson

Gaolathe

Gross

. Overestimates of Survival after HAART: Implications for Global Scale-Up Efforts. PLoS ONE 2008; 3: e1725–e1725.