Secondary Logo

Journal Logo

Original articles

The empirical evidence for the impact of HIV on adult mortality in the developing world: data from serological studies

Porter, Kholouda; Zaba, Basiab

Author Information



Over the past 25 years the impact of HIV on communities in the developing world in terms of years of life lost and social and economic decline has been immense. Although national level population data on deaths and life expectancy collected at different timepoints provide information on the scale of change, it falls to studies with information on the HIV serostatus of study subjects to provide the direct evidence that the change is caused by HIV, and to quantify the contribution of HIV to it.

This paper reviews the evidence on the impact of HIV on adult mortality from studies conducted in the community and in other settings. Findings are reported on the basis of comparisons of individuals without HIV infection with those who are infected. The latter are characterized as ‘sero-incident’ or ‘seroconverters', if the time of HIV seroconversion is reasonably well estimated, or ‘prevalent’ if it is not known. We review the findings that have accumulated since the last report [1], compare these findings and discuss factors likely to influence differences reported between studies. We explore the limitations of measures used to date and discuss a number of approaches to addressing issues of the comparability of findings from different populations. We also briefly discuss the use of data on HIV serostatus of the deceased obtained in mortuary studies, and the role and value of verbal autopsy data [2].

A wealth of data has accumulated and continues to inform and improve our understanding of the clinical course of HIV disease in sub-Saharan Africa, which bears the largest share of the world's HIV burden. However, this overview is not restricted to work undertaken in African settings, and includes findings reported from Asia and the Caribbean.

Review of evidence

Survival after HIV seroconversion

Besides being an important tool for epidemic modellers and healthcare planners, estimates of survival after HIV seroconversion are crucial to counsel infected individuals and to help inform treatment decisions. These estimates, and factors associated with them, also provide a useful insight into the pathogenesis of HIV. Four developing country cohorts have reported on estimated survival after HIV seroconversion in their study population: the Masaka cohort in Uganda (n = 191, 1156 person-years (PY) of follow-up); the Kisesa cohort in Tanzania (n = 209, 584 PY); the Royal Military Cohort in Thailand (n = 235); and the GHESKIO hospital-based study in Haiti (n = 42) [3–6]. Survival estimates have been derived from seroconverters with an HIV antibody-negative test within an interval of 2–3 years within their positive test, thus allowing reasonable estimates of the time of seroconversion to HIV to be made. Estimates of survival 5 years post-infection are similar for the two African cohorts at 79% (Kisesa) and 77% (Masaka) and higher at 82 and 88% for the Thai and Haitian cohorts, respectively, although confidence intervals are wide. Median survival from the Masaka cohort is estimated to be 8.6 years [95% confidence interval (CI) 5.6–12] in broad agreement with the Haitian GHESKIO study of 7.4 years (95% CI 6.2–10.2; Fig. 1).

Fig. 1.
Fig. 1.:
Estimated survival after HIV seroconversion. –– ♦ –– Masaka; –– ▴ –– Kisesa; –– □ –– Haiti hospital; –– ○ –– Thai military; – – × – – Manila commercial sex workers; – – + – – Nairobi commercial sex workers.

Earlier survival estimates from studies of commercial sex workers (CSW) with estimated dates of seroconversion were less favourable, with 52.1% cumulative probability of death at 6.5 years for Filipino CSW (n = 20) [7], and median time to AIDS of 4.4 years for Nairobi CSW [8]. In any case, even current estimates suggest that the survival of infected individuals in developing countries is somewhat shorter than the 10.9 years (95% CI 10.6–11.3 for 25–34 years at seroconversion) observed in industrialized countries before the introduction of potent antiretroviral therapy [9]. There are a number of reasons that may explain the differences between reported estimates, including the factors that are known to influence HIV survival. First, background mortality rates are different in the communities in which these cohort studies have been set up. The survival statistics quoted above have been calculated on the basis of all deaths occurring to infected individuals; no attempt has been made to account for mortality from causes other than HIV. All-cause adult mortality varies widely across Africa, and is much higher than in industrialized countries because of factors such as poor nutrition, vector-borne diseases, and inadequate health services. In the late 1980s, before the AIDS epidemic had had a major impact on adult mortality, the cumulative probability of dying between the age of 15 and 45 years in Africa ranged from 21 to 56%, higher overall and with a wider range than in Latin America and the Caribbean (11–20%), or in industrialized countries (8–13%) [10].

Second, age is known to be an important risk factor in the progression of HIV disease, with older age at seroconversion being associated with a shorter survival time [9,11–14]. The variations between these estimates may well reflect the age structure of each cohort, and direct comparisons between them may not be appropriate. Although the median age at seroconversion for the whole cohort is not always reported, it is of interest, however, that the Nairobi CSW, with the lowest survival probability, are older (mean age 31 years) than the women in the other cohorts (modal age group for both Kisesa and Masaka cohorts 20–24 years, median age in Haitian and Filipino CSW cohort 26.5 and 24 years, respectively).

Third, survival differences may be caused by the HIV sub-type of the subjects in each cohort [15–18]. This issue remains unresolved, as some investigators have found variations in the progression rate [15,16] by viral subtype, whereas others have not [17,18]. Such differences, if they exist, may be of little or no clinical consequence, or they may affect survival comparisons, not only between cohorts in developing and industrialized countries, but also within developing country populations.

Finally, it is known that a number of sources of bias in the recruitment and follow-up of cohort participants can lead to variations in estimated survival times [19]. One of the most important problems is unobserved mortality and other losses to follow-up that may occur between actual seroconversion, the time that the fact of seroconversion can be established through routine or voluntary testing, and enrolment into a cohort study. If an individual needs to survive a certain minimum time after seroconversion before they can enrol into a cohort study, survival estimates from that cohort will be longer than actual survival. The investigators of the Masaka cohort reported that the mortality rate of seroconverters who did not enrol in their cohort was higher than that for those enrolled. Such bias in estimates can be minimized by the use of statistical techniques in survival analyses, such as allowing for left censoring of subjects according to their entry into observation from their time of seroconversion. Individuals lost to follow-up may not have the same risk of death as those remaining under surveillance and right censoring could, therefore, be informative. If those lost to follow-up are more likely to have died, survival expectations will be overestimated. Conversely, if they are less likely to have died or if investigators are more likely to become aware of a fact of death than that an individual remains alive (but lost to follow-up), survival expectations will be underestimated. Apparent survival differences between cohorts may, therefore, be artefacts of the study design.

One further potential source of differences between developing and industrialized countries is the mode of infection. The majority of individuals in developing country cohorts are likely to have been infected through sex between men and women, and many through the receipt of contaminated blood products. In industrialized countries, however, cohorts tend to be dominated by those infected through sex between men, injecting drug users, and now to a lesser extent, the recipients of contaminated clotting factors. A large re-analysis of data from 38 seroconverter cohorts in industrialized countries, including data on 933 individuals infected through sex between men and women, found no evidence to suggest that the mode of infection is prognostic of survival, once age at seroconversion has been adjusted for [9]. It is unlikely, therefore, that differences between developing and industrialized country cohorts are the result of the mode of infection for seroconverters included in the cohort.

These factors account for the differences between survival estimates derived from the various cohorts in developing countries, and between those studies and results observed in industrialized countries. It is likely that differences in the mortality rates for individuals without HIV infection, so-called background mortality, make the greatest overall contribution to differences between estimates derived from developing and industrialized country cohorts.

Mortality rates and rate ratios

Irrespective of HIV infection, mortality rates in developing countries are higher in each age group than those in industrialized countries, largely because of other communicable diseases and poor nutrition. Recent direct data on these rates have become available from four community-based studies in: Kisesa, Tanzania (1994–1998) [4,20]; Rakai, Uganda (1994–1997) [21]; Karonga district, Malawi (1998–2000) [22]; and Masaka, Uganda (1989–2002) [14,23], as well as the earlier data from Mwanza, Tanzania (1991–1993) [24]. Mortality rates for uninfected individuals from these studies vary between two and five deaths per 1000 PY for those in their teens and twenties, increasing to between five and 17 deaths per 1000 PY for individuals in their thirties and forties (Fig. 2). Clearly, there is a great deal of variability between these findings, and no doubt other factors, such as violence, war and famine will also influence rates in other settings and at other timepoints. A further study based in a factory in Tanzania provided much lower estimates of mortality of the infected than those reported from the community-based studies, with one death per 1000 PY for individuals in their twenties rising to three to four deaths for those in their thirties and forties [25]. These very low estimates are possibly caused by a ‘healthy worker’ effect, and are low even in comparison with rates in developed country studies. In any case, the rates from the community studies are much higher than death rates for 15–29 and 30–49 year olds in industrialized countries, such as the United Kingdom, where less than one death per 1000 PY would be expected for 15–29 year olds, rising to two to four deaths per 1000 PY for 45–54 year olds [26].

Fig. 2. Mortality rates of HIV-infected and uninfected individuals in developed and developing countries.
Fig. 2. Mortality rates of HIV-infected and uninfected individuals in developed and developing countries.:
HIV positive —⋄— Rakai; —□— Kisesa; —X— Karonga; —ж— Mwanza; —○— Masaka; — CASCADE 10–15 Years; –– CASCADE ≤5 years. HIV negative –– ⋄ –– Rakai; –– □ –– Kisesa; –– × –– Karonga; –– ж –– Mwanza; –– ○ –– Masaka; –– UK, 1995.

Against this background of fairly high mortality, rates for infected individuals in community-based studies are higher still and vary considerably. Those reported to date vary between 25 and 45 deaths per 1000 PY for individuals in their teens, rising to 70–120, 90–150, and 90–200 for those in their twenties, thirties and forties, respectively (Fig. 2). The increase in mortality rates with increasing age is partly caused by the increase in background mortality as a result of ageing and is partly due to the fact that the duration of HIV infection is likely to be longer in older individuals, and because the immune system of infected individuals declines more rapidly at older ages. Estimates of HIV prevalence for the adult populations in those studies were 5, 7, 8–11, 10–13 and 16% for Mwanza, Kisesa, Masaka, Karonga and Rakai studies, respectively. In so far as the higher prevalence rates above represent epidemics of longer duration, we would expect higher mortality rates among HIV infected, to correspond to higher prevalence, especially at older ages, and this is, more or less, borne out by the data. The two Tanzanian studies (Mwanza and Kisesa), where the epidemic is more recent, have lower mortality rates in older HIV-infected individuals than the Ugandan studies (Masaka and Rakai) or the Karonga study from Malawi. However, as the Tanzanian studies also report lower mortality rates in the uninfected, it is difficult to judge the relative contributions these factors make to driving down mortality rates for the infected.

As with the mortality rates for the uninfected, estimates in HIV-infected factory workers in the study in Tanzania are much lower, at a rate of 13 deaths per 1000 PY for individuals in their twenties, rising to 29 and 76 deaths for individuals in their thirties and forties, respectively [25]. This low rate is likely to reflect mortality among individuals with a relatively short HIV infection duration and lower than average mortality from non-HIV causes, a ‘healthy worker’ effect. Mortality rates from one study in the Gambia are surprisingly high at 135 deaths per 1000 PY for teenagers, rising steeply to a rate of 570 deaths per 1000 PY for individuals in their forties [27]. These high mortality rates may reflect the fact that the study was hospital based and included a high proportion of individuals presenting with advanced disease (46% with CD4 cell counts < 200 cells/mm3 at presentation).

Age-specific mortality rates for infected individuals are slightly higher for men than women, probably a reflection of the background mortality advantage of women over men. The Gambia study reports HIV mortality rates for men aged 15–24 years that are 1.9 times that for women of the same age group, increasing to a threefold difference for 35–44 year olds [27]. This extreme sex differential in mortality rates is not, however, observed in other cohorts, suggesting that the men in the Gambian study may have been infected for much longer than women of the same age group. This finding is unexpected given that, in general, men are infected at older ages than women; it may reflect the fact that the HIV epidemic in the Gambia is at a very early stage, and these patterns may change or become reversed over time. As this is a hospital-based study, the distribution of patients by duration since infection may not represent the distribution in the community.

Apart from age, the most important factor to influence mortality rate is the duration of HIV infection, as mortality rate is known to increase with the duration of HIV infection. Estimates from the Collaborative Group on AIDS Incubation and HIV Survival in industrialized country cohorts before the introduction of potent anti-HIV therapies showed that in this setting, mortality rate increases from three deaths per 1000 PY (95% CI 2–5) in the first year after seroconversion to 170 deaths per 1000 PY (141–205) by 13 years after seroconversion for those infected between 1977 and 1996 at 25–29 years of age [9]. The shape of the underlying hazard function suggests that the risk of death is low in the first 3 years, rising steadily thereafter and reaching a plateau at approximately 7–8 years after seroconversion.

Fig. 2 also includes, for comparative purposes, age-specific mortality rates derived from the CASCADE Collaboration in industrialized countries before the advent of potent therapies [28]. These are shown at two different durations: less than 5 years since infection, and 10–15 years since infection. Age-specific mortality rates for uninfected individuals in the developed countries are represented by those for the general population of the UK. This figure clearly shows the mortality disadvantage of uninfected adults in developing countries; their mortality rates are between five and 10 times those of the UK population, although in all cases lower than the mortality of adults infected less than 5 years in the CASCADE cohorts. Infected adults in the developing country studies have mortality rates as high as those of the CASCADE cohorts after 10–15 years of infection, even though we would expect the average duration since infection in these cohorts to be considerably lower, as these are for the most part growing epidemics and the median survival time is probably less than 9 years.

The mortality rate ratio (MRR) of infected to uninfected individuals will depend on background mortality, the duration of HIV infection, and age. As the duration of infection and background mortality are both strongly associated with age, MRR comparisons are only meaningful if they are presented separately by age group, or standardized for age (Fig. 3). The general pattern of MRR by age group is of a relatively low ratio at 15–20 years of age, rising to a peak at 25–30 years of age and falling again with increasing age. The low MRR in early adult years is the result of relatively low mortality rates in infected individuals who will have been infected with HIV relatively recently. The lower ratios in later life are caused by other competing causes of death, which raise mortality rates in uninfected individuals.

Fig. 3. Mortality rate ratios of infected to uninfected individuals by age in sub-Saharan Africa studies.
Fig. 3. Mortality rate ratios of infected to uninfected individuals by age in sub-Saharan Africa studies.:
–– □–– Masaka; –– ⋄ — Rakai; – – ▵ – – Kisesa; –– • –– Mwanza; –– * – – Karonga.

Because women tend to be infected at younger ages than men, in a given age group infected women have, on average, a longer duration of HIV infection compared with infected men. In addition, age-specific mortality rates in uninfected individuals are higher for men than for women so that MRR are higher for women. The implications are that HIV contributes to the mortality of women in developing countries more than it does to that of men. Age-standardized MRR have been computed by Zaba [29], on the basis of published data from various studies [14,20–24] and are shown in Table 1. They suggest that, on average, infected men experience nine to 16 times the mortality rates of uninfected men, and infected women have 15–25 times higher mortality rates than uninfected women.

Table 1
Table 1:
Age standardized mortality rate ratios of HIV-infected to uninfected individuals.

Population-attributable fraction

The population-attributable fraction (PAF) of mortality caused by HIV is defined as the proportionate increase in mortality rate in the population as a whole, over and above the rate that would be experienced in the absence of HIV infection. In general, we would expect PAF to be higher in populations with a higher prevalence of HIV, in populations with a younger incidence pattern, and in more mature epidemics in which the mortality of infected individuals was higher because there would be a higher proportion of infected individuals with a long duration of infection. PAF statistics are commonly used for advocacy, and can be calculated in a variety of ways, exploiting data from different sources.

It can be calculated using the formula shown in equation 1, which only requires mortality to be measured in the uninfected population and in the population as a whole. The mortality rate for infected individuals, who are likely to be a minority, is not required, and neither is an estimate of HIV prevalence in the population.

A frequently encountered alternative formula, mathematically equivalent to the first, which requires a measure of the mortality rate for infected individuals as well as HIV prevalence in the study population is shown in equation 2. The denominator is simply the mortality in the population as a whole, expressed as the weighted sum of the mortality of infected and uninfected individuals. This formula is particularly useful for estimating the expected change in PAF as the prevalence changes. It can also be used to estimate PAF using data from case–control studies in which mortality rates of infected and uninfected individuals are known, but prevalence has to be estimated from another source.

Another less well known variant PAF formula is shown in equation 3, which is suitable for use with data pertaining to the HIV status of deceased individuals. This requires an estimate of the MRR, which in theory limits its usefulness because, by definition, population denominators are required to estimate MRR, and these cannot be obtained from mortuary studies. However, when the MRR is large, as is the case for HIV, particularly for the ages of 25–30 years, the formula is relatively insensitive to the exact value of MRR, and an ‘average’ value of age-specific or age-standardized MRR, based on other studies, can be used. This approximation allows the use of a relatively inexpensive data source to estimate PAF, on the basis of the proportion infected of all deaths.

Mortuary-based studies have the added advantage of not requiring any follow-up, although in Africa they would be limited to major urban areas in the foreseeable future.

A study in the Congo [30] reported that, of 145 corpses tested for HIV in two of Brazzaville's morgues, 26.2% (n = 38) were HIV antibody positive. When that study was conducted in 1996, HIV prevalence was estimated to be 5.2%, on the basis of the surveillance of pregnant women. In 2001, with an estimated prevalence of 6.3% in Pointe Noire, Congo, the investigators reported that 44.2% of the 1623 corpses tested were found to be HIV infected [31]. In the two studies in the Congo, the prevalence of HIV among those who died was five to seven times higher than the prevalence among the living. This compares well with the model-based predictions of the cohort relationship between the prevalence of HIV and the lifetime risk of dying from AIDS made by Blacker and Zaba [32,33].

Table 2 compares prevalence levels and HIV-attributable mortality for the five community-based studies featured in Table 1 and the two morgue studies discussed above. In the four prospective cohorts [14,20,21,23,24], prevalence has been measured directly from the study population and HIV-attributable mortality can be calculated directly using equation 1. For the retrospective cohort study in Karonga [22], we use equation 2, with alternative upper and lower estimates for the average prevalence (7–9%) covering the years in which the mortality data were collected. Finally, equation 3 is used to convert the proportion of HIV-infected corpses observed in the morgue studies [30,31] to an estimate of attributable mortality, using the age-standardized MRR for both sexes from the cohort studies, which range between 11 and 20 (Table 1). The prevalence estimates shown for the morgue studies are based on sentinel surveillance in antenatal clinics. In general, HIV-attributable mortality increases with HIV prevalence, suggesting that prevalence is the most important determinant of attributable mortality, although varying age structure, age at infection and epidemic maturity introduce some of the irregularities in the sequence.

Table 2
Table 2:
Population-attributable fraction of HIV mortality.

Verbal autopsies

Clinical endpoints in resource-poor settings are generally not easily ascertainable because they require costly laboratory diagnoses, which may have low prognostic value. Furthermore, it is often difficult to ascertain whether a death is related to HIV infection or whether HIV was incidental to it. This makes AIDS a difficult endpoint to use in clinical studies for comparing findings between studies or even over time within the same study.

Studies with information on the HIV serological status of participants have evaluated the prognostic value of using verbal autopsies to assign a probability to the deceased having been infected and HIV being the underlying cause of death [34]. In theory, this enables investigators to use verbal autopsies as a tool to estimate the contribution of HIV to mortality in the absence of information on HIV serostatus in the study population. However, the Kisesa cohort study has found that verbal autopsy techniques underestimate the proportion of deaths associated with HIV/AIDS [20]. In that study, among deceased individuals whose HIV status was known to those interpreting the diagnostic data, over half of the deaths were classified as HIV associated, but among those for whom the HIV status was not available, only 28% were diagnosed as HIV related on the basis of verbal autopsy alone. The Rakai cohort study also reported that over a third of the deaths occurring to HIV-infected individuals did not meet the clinical case definition for AIDS, so it is not surprising that verbal autopsy approaches fail to identify possible HIV-related deaths [21]. This suggests that the proportion of all deaths identified as HIV related based on verbal autopsy evidence alone should not be used to estimate HIV attributable mortality.

Net mortality from HIV-related causes

In order to develop a schedule of mortality caused by HIV that could be used to predict mortality levels during the epidemic in populations with different background mortality levels, we need to estimate patterns of ‘net’ mortality: the theoretical mortality patterns that would be experienced by infected individuals in the absence of mortality risk from causes unrelated to HIV. This is a somewhat artificial construct, given that other causes, particularly infections, may act synergistically with HIV to increase the risk of death to levels higher than may be expected simply from the probabilistic combination of risks for the HIV infected and the uninfected. Any ‘net’ mortality calculated may, therefore, be higher than the actual mortality caused by HIV infection alone.

The investigators of the Masaka cohort in Uganda have attempted to account for both background mortality and the duration of HIV infection [3]. The analyses utilized data from seroconverters, and estimated their survival expectations at different times after estimated seroconversion. Seroconverters were matched by age, in a 1 : 1 ratio, to uninfected individuals and both infected and uninfected were followed up. In order also to account for the risk of death, which is expected regardless of HIV infection, and over time as the subjects in the cohort age, a ‘net’ mortality schedule for HIV was calculated by subtracting the mortality hazards among the uninfected from the hazard experienced by the infected during each single year of follow-up. Results indicated that the contribution of HIV to mortality is low in the first 1–2 years after seroconversion, increasing over time in all age groups, most dramatically in the oldest age group. The median survival time for the ‘net’ schedule calculated for individuals aged under 40 years at seroconversion was approximately 9.5 years, but for those over 40 years at seroconversion it was much lower, approximately 5 years. This mirrors the results from the overall survival patterns discussed earlier, and suggests that the shorter survival times for individuals infected at older ages are not caused only by higher background mortality in the elderly.

A similar analysis was performed using data from the Kisesa cohort [4], although the shorter follow-up times precluded an analysis broken down by age at infection. The Kisesa results suggested that if HIV-related conditions were the only cause of death, the median survival time after seroconversion would be 8 years, with a 25% estimated probability of death within 5.7 years of infection. A comparison with the survival curve for the seroconverters in the cohort, in which 25% are estimated to die within 5.1 years of infection, suggests that non-HIV-related causes account for approximately 10% of the deaths of seroconverters.

In conclusion, results from community-based cohorts in developing countries with sufficient numbers of seroconverters and follow-up time suggest that the median survival is between 8 and 9 years for individuals infected at 20–29 years of age, but is considerably shorter for those infected at older ages. This is some 20% shorter than survival times in industrialized countries before the introduction of highly active antiretroviral therapy, but not as dramatically different as was originally postulated on the basis of early data from CSW.

In order to make meaningful comparisons of mortality rates of HIV-infected individuals in different study settings, these will need to be adjusted for age, and HIV infection duration because these factors are the major determinants of HIV disease progression. Furthermore, investigators need to account for background mortality, which may also affect survival post-infection. Most community-based studies will have data available to make adjustments for age and background mortality, but adjustment for the duration of HIV infection is problematical as this will be unknown for the vast majority of infected individuals. We do not yet have a large enough collection of age and duration-specific mortality rates for HIV-infected individuals to begin to construct standard patterns.

The variability in mortality rates among uninfected individuals in different cohort settings makes it difficult to compare mortality rates among the infected. One proposed way forward is to compare duration-specific HIV ‘net’ mortality rates and their associated survival curves. These are not easy to construct for studies with a relatively short follow-up time, because of the preponderance of infected individuals with unknown infection durations. The Kisesa cohort investigators have used the age and survival patterns of the seroconverters in their cohort to impute seroconversion dates for prevalent cases to try to overcome this difficulty, but these methods have not yet been tested elsewhere. The long-term aim for generating more robust age-specific ‘net’ mortality estimates should be to repeat analyses such as those undertaken by the Masaka cohort investigators, which account for background mortality, natural ageing and HIV infection duration in other community-based studies.

The impact of HIV on adult mortality in developing countries has been substantial for both men and women at all age groups. Its impact has been greatest, however, in individuals in their twenties and thirties, and proportionately larger in women than men. The variability reported across the studies on the impact of HIV on populations in developing countries is a function not only of the prevalence of HIV infection, but also reflects the variation in mortality rates in uninfected individuals and the mean duration of infection in individuals living with HIV/AIDS. Studies that measure the time from HIV infection for individuals are uniquely able to examine HIV disease progression throughout the whole period of infection, but such data are rare in the developing world. There is an urgent need to combine these available data and to obtain a clearer picture of the mortality patterns of HIV-infected individuals. This could inform policy on the delivery of potent anti-HIV therapy to infected populations in resource-poor countries, and act as the baseline against which the impact of therapy at the population level can be assessed. The potential for collecting long-term mortality data on populations unaffected by antiretroviral therapy will diminish rapidly in the next few years.


1. Boerma JT, Nunn AJ, Whitworth JAG. Mortality impact of the AIDS epidemic: evidence from community studies in less developed countries.AIDS 1998; 12 (Suppl. 1):S3–S14.
2. Okongo M, Morgan D, Mayanja B, Ross A, Whitworth J. Causes of death in a rural, population-based human immunodeficiency virus type 1 (HIV-1) natural history cohort in Uganda.Int J Epidemiol 1998; 4:698–702.
3. Whitworth J, Shafer LA, Mahe C, van der Paal L. Survival since infection and its relation to background mortality in the Masaka general population cohort.Paper presented at meeting Empirical Evidence for the Demographic and Socioeconomic Impact of AIDS. Durban, March 2003.
4. Urassa M, Isingo R, Mwaluko G, Boerma T, Zaba B. Mortality of HIV prevalent and incident cases in the Kisesa cohort study.Paper presented at meeting Empirical Evidence for the Demographic and Socioeconomic Impact of AIDS. Durban, March 2003.
5. Rangsin R, Chiu J, Sirisopana N, Eiumtrakul S, Khamboonruang C, Robb M, et al. Natural history of HIV subtype E infections in Thai men 5–7 years after infection. In: 8th Conference on Retroviruses and Opportunistic Infections. Chicago, 2001 [Abstract 207].
6. Deschamps MM, Fitzgerald DW, Pape JW, Johnson WD Jr. HIV infection in Haiti: natural history and disease progression.AIDS 2000; 14:2515–2521.
7. Manaloto CR, Perrault JG, Caringal LT, Santiago EG, Wignall FS, Gonzales VL, et al. Natural history of HIV infection in Filipino female commercial sex workers.J Acquired Immune Defic Syndr 1994; 7:1157–1168.
8. Anzala OA, Nagelkerke NJ, Bwayo JJ, Holton D, Moses S, Ngugi EN, et al. Rapid progression to disease in African sex workers with human immunodeficiency virus type 1 infection.J Infect Dis 1995; 171:686–689.
9. Collaborative Group on AIDS Incubation and HIV Survival. Time from HIV-1 seroconversion to AIDS and death before widespread use of highly-active antiretroviral therapy: a collaborative re-analysis.Lancet 2000; 355:1131–1137.
10. Murray CJL, Yang G, Qiao X. Adult mortality: levels patterns and causes, chapter 2. In: Feachem R, Kjellstrom T, Murray C, Over M, Phillips M (editors). The health of adults in the developing world. World Bank, Washington DC, USA; 1992. pp. 43–47.
11. Phillips AN, Lee CA, Elford J, Webster A, Janossy G, Timms A, et al. More rapid progression to AIDS in older HIV infected people: the role of CD4+ T-cell counts.J Acquired Immune Defic Syndr 1991; 4:970–975.
12. Rosenberg PS, Goedert JJ, Biggar RJ, for the Multicenter Hemophilia Cohort Study and the International Registry of Seroconverters. Effect of age at seroconversion on the natural AIDS incubation distribution.AIDS 1994; 8:803–810.
13. Pezzotti P, Phillips AN, Dorrucci M, Cozzi Lepri A, Galai N, Vlahov D, Rezza G. Category of exposure to HIV and age in the progression to AIDS: longitudinal study of 1199 people with known dates of seroconversion.BMJ 1996; 313:583–586.
14. Morgan D, Mahe C, Mayanja B, Okongo JM, Lugeba R, Whitworth JAG. HIV-1 infection in rural Africa: is there a difference in median time to AIDS and survival compared with that in industrialised countries?AIDS 2002; 16:597–603.
15. Kanki PJ, Hamel DJ, Sankale JL, Hsieh C, Thior I, Barin F, et al. Human immunodeficiency virus type 1 subtypes differ in disease progression.J Infect Dis 1999; 179:68–73.
16. Kaleebu P, French N, Mahe C, Yirrell D, Watera C, Lyagoba F, et al. Effect of human immunodeficiency virus (HIV) type 1 envelope subtypes A and D on disease progression in a large cohort of HIV-1-positive persons in Uganda.J Infect Dis 2002; 185:1244–1250.
17. Alaeus A, Lidman K, Bjorkman A, Giesecke J, Albert J. Similar rate of disease progression among individuals infected with HIV-1 genetic subtypes A–D.AIDS 1999; 13:901–907.
18. Laurent C, Bourgeois A, Faye MA, Mougnutou R, Seydi M, Gueye M, et al. No difference in clinical progression between patients infected with the predominant human immunodeficiency virus type 1 circulating recombinant form (CRF) 02_AG strain and patients not infected with CRF02_AG, in Western and West-Central Africa: a four-year prospective multicenter study.J Infect Dis 2002; 186:486–492.
19. Porter K, Johnson AM, Phillips AN, Darbyshire JH. The practical significance of potential biases in estimates of the AIDS incubation period distribution in the UK Register of HIV Seroconverters.AIDS 1999; 13:1943–1951.
20. Urassa M, Boerma JT, Isingo R, Ngalula J, Ng'weshemi J, Mwaluko, Zaba B. The impact of HIV/AIDS on mortality and household mobility in rural Tanzania.AIDS 2001; 15: 2017–2023.
21. Sewankambo NK, Gray RH, Ahmad S, Serwadda D, Wabwire-Mangen F, Nalugoda F, et al. Mortality associated with HIV infection in rural Rakai District, Uganda.AIDS 2000; 14: 2391–2400.
22. Crampin AC, Floyd S, Glynn JR, Sibande F, Mulawa D, Nyondo A, et al. Long-term follow-up of HIV-positive and HIV-negative individuals in rural Malawi.AIDS 2002; 16: 1545–1550.
23. Morgan D, Muwonge R, Mahe CJW. Natural history of HIV infections in rural Uganda – results after 10 years. In: XIIIth International AIDS Conference. Durban, 2000 [Abstract TuOrC428].
24. Todd J, Balira R, Grosskurth H, Mayaud P, Mosha F, ka-Gina G, et al. HIV-associated adult mortality in a rural Tanzanian population.AIDS 1997; 11:801–807.
25. Senkoro KP, Boerma JT, Klokke AH, Ng'weshemi JZ, Muro AS, Gabone R, Borgdorff MW. HIV incidence and HIV-associated mortality in a cohort of factory workers and their spouses in Tanzania, 1991 through 1996.J Acquired Immune Defic Syndr 2000; 23:194–202.
26. Office for National Statistics; General Register Office for Scotland; Northern Ireland Statistics and Research Agency. Age specific death rates: by sex regional trends 37 – Table 3.14. HMSO, London, 2000.
27. Schim van der Loeff MF, Jaffar S, Aveika AA, Sabally S, Corrah T, Harding E, et al. Mortality of HIV-1, HIV-2 and HIV-1/HIV-2 dually infected patients in a clinic-based cohort in The Gambia.AIDS 2002; 16:1775–1783.
28. Babiker AG, Peto T, Porter K, Walker AS, Darbyshire JH. Age as a determinant of survival in HIV infection.J Clin Epidemiol 2001; 54 (Suppl 1):S16–S21.
29. Zaba B. Direct assessment methods for measuring HIV/AIDS mortality. In: XIVth International AIDS Conference. Barcelona, July 2002 [Abstract MoOrC121].
30. Pictet G, Le Coeur S, M'Pele P, Brouard N, Lallemant M. Contribution of AIDS to the general mortality in Central Africa: evidence from a morgue-based study in Brazzaville, Congo.AIDS 1998; 12:2217–2223.
31. Le Coeur S. HIV status and AIDS mortality in Pointe-Noire, Congo.Abstract presented at meeting Empirical Evidence for the Demographic and Socioeconomic Impact of AIDS. Durban, March 2003.
32. Blacker J, Zaba B. HIV prevalence and the life-time risk of dying from AIDS.Health Trans Rev 1997; 7(Suppl. 2):45–62.
33. UNAIDS. Report on the global HIV/AIDS epidemic, 2000.Released at the XIIIth IAS Conference. Durban, July 2000.
34. Kamali A, Wagner HU, Nakiyingi J, Sabiiti I, Kengeya-Kayondo JF, Mulder DW. Verbal autopsy as a tool for diagnosing HIV-related adult deaths in Uganda.Int J Epidemiol 1996; 25: 679–684.

adult mortality; rate ratio; survival; seroconversion; developing countries

© 2004 Lippincott Williams & Wilkins, Inc.