Introduction
The HIV pandemic is the major public health challenge of the present decade [1]. In sub-Saharan Africa, HIV infection has mostly affected young adults in the prime of life, and has been spread predominantly by heterosexual transmission [2]. An HIV prevalence in excess of 20% have been recorded in the general population of some African cities.
The natural history of HIV infection in developed countries has been well-documented, with a median incubation period from seroconversion to AIDS of around 10 years [3-5]. By contrast, we are not aware of any published studies in which the incubation period has been accurately recorded in general population cohorts in Africa. A study of female sex workers in Nairobi suggested much more rapid progression to AIDS in this high-risk population, with an estimated median incubation period of only 4.4 years [6]. Studies of prevalent cases of HIV infection in rural Africa also suggest more rapid progression to disease and death than in developed countries [7-9].
Reports from hospitals in cities severely affected by the epidemic have shown that HIV is already the leading cause of death among young adults between 20 and 40 years of age [10,11]. Model projections have suggested that HIV may soon become the leading cause of adult mortality in sub-Saharan Africa [12].
HIV prevalence in rural areas is generally lower than in urban centres [13-15], and its impact on adult mortality may therefore be less than in the cities. Two studies have reported the effects of HIV on mortality in rural population cohorts in Uganda [7-9]. In Masaka district, HIV prevalence was 8% in adults aged 13 years and over, and 50% of all adult deaths and 89% of deaths in those aged 25-34 years were attributed to HIV infection [8]. In Rakai district, in a population with an HIV prevalence of 13% in adults aged 15 years and over, 52% of adult deaths were attributed to HIV [9]. In Rakai, the proportion of HIV-attributable deaths was considerably higher in trading settlements with high HIV prevalence rates than in rural villages.
Data on the clinical features of HIV disease in Africa are also limited. Studies in hospitals in major cities have recorded clinical manifestations of HIV differing from those in developed countries [16,17], and some tropical diseases are known to influence the course of HIV infection [18]. Little is known about the clinical manifestations of HIV disease and mortality in the rural areas where the majority of the African population lives. The World Health Organization (WHO) has proposed a clinical case definition for AIDS in developing countries [19], but concerns have been expressed about its usefulness [20].
This study reports the mortality rates observed during 2 years of follow-up of a cohort of 12 000 adults in a rural area of Tanzania with an adult HIV prevalence below 5%, and the signs and symptoms associated with these deaths.
Methods
The demographic characteristics of Mwanza region in north-west Tanzania have been described elsewhere [15,21]. Briefly, the region lies on the southern shores of Lake Victoria and the inhabitants are mainly subsistence farmers of the Sukuma tribe. Between 1991 and 1995, a randomized trial was conducted to measure the impact of improved treatment services for sexually transmitted diseases (STD) on HIV incidence; the design and implementation of this trial have been reported previously [21,22]. Impact was measured in a randomly selected cohort of approximately 1000 adults aged 15-54 years from each of 12 rural communities.
Individuals were enrolled in the cohort over a 12-month period starting in November 1991, after giving informed consent. Each subject was interviewed concerning demographic characteristics and aspects of sexual health. A venous blood sample was taken and tested for HIV antibodies using two independent enzyme-linked immunosorbent assay (ELISA) methods (Vironostika HIV MIXT Microelisa, Organon Teknika, Boxtel, The Netherlands and Wellcozyme HIV1+2 GACELISA, Murex Diagnostics, Dartford, Kent, UK). Sera giving indeterminate results, or for which the two ELISA results were discrepant, were tested by Western blot (HIV-1 Western blot, Epitope, Beaverton, Oregon, USA).
Cohort members were resurveyed approximately 2 years after enrolment, using similar data collection methods. The main outcome of the trial was HIV seroconversion, and this has been reported elsewhere [23]. Reasons for non-attendance at follow-up were recorded, and categorized as follows: died, moved permanently away from the community, travelled away from the community for a short time, and other reasons. Other reasons included refusals, individuals who could not be visited for logistical reasons (e.g, impassable roads), and individuals who were unknown to the village leaders and could not be located. Attendance rates and reasons for non-attendance were compared between those who were HIV-negative and HIV-positive at baseline.
Mortality rates (per 1000 person-years) were computed according to age, sex and baseline HIV status. Individuals who attended at follow-up contributed 2 person-years of observation, and those not seen (including those who died) were assumed to contribute 1 person-year. Rate ratios for mortality in HIV-positive compared with HIV-negative individuals were computed within each age-group, and the Mantel-Haenszel method was used to calculate age-adjusted rate ratios for men, women and both sexes. The proportion of deaths attributable to HIV infection [the population attributable fraction (PAF)] was computed separately for each age-sex-group, and overall PAF for men, women and all subjects were computed by accumulating attributable deaths over age-groups.
To obtain information on the clinical features of deaths among cohort members, the homes of all those reported to have died were revisited between 3 and 8 months after the follow-up survey. An appropriate respondent was identified, and this was usually the household member who had cared for the deceased during the final illness. This respondent was questioned about the signs and symptoms preceding death using a verbal autopsy questionnaire developed to investigate adult deaths in rural Africa (D. Chandramohn and G. Maude, personal communication, 1993). The interview was conducted by one of six carefully trained non-medical field workers, who normally worked as teachers or cultural extension officers. Respondents were first asked to describe the final illness in their own words and to state their opinion of the cause of death. Precoded questions were then asked concerning specific signs and symptoms. The interviewers were blind to the HIV status of the deceased. Reported signs and symptoms were compared in HIV-negative and HIV-positive deaths. The pre-coded questions were used to classify deaths as 'AIDS' or 'non-AIDS' using the WHO clinical case definition [19]. A simplified definition was also used, which considered deaths as AIDS-related if they were associated with at least two of the following four symptoms: fever for more than 30 days, cough for more than 30 days, severe loss of weight, and diarrhoea for more than 30 days. During the interview, respondents had three opportunities to mention HIV or AIDS as a cause of death: in response to a question on the reasons for a visit to a health centre, during the unstructured account of the final illness, or in response to a final direct question on the possibility of the death being HIV/AIDS-related. The association of the HIV status of the deceased with each of these definitions of 'AIDS' was examined.
Results
An unequivocal HIV test result was obtained in the baseline survey for 12 501 adults (5859 men and 6642 women) aged 15-54 years. Five hundred and five (4.0%) individuals were HIV-positive: 217 (3.7%) men and 288 (4.3%) women.
Two years later, 8836 (71%) of the cohort were seen at the follow-up survey, 4233 (72%) men and 4603 (69%) women. The remaining 3665 were not seen and the reasons for non-attendance are shown in Table 1. Death was given as the reason for the non-attendance of 199 subjects, but three of these were later found by the verbal autopsy team to be living. Of the 196 confirmed deaths (111 men and 85 women), 123 individuals were HIV-negative at baseline, and 73 (37%) were HIV-positive. Rates of non-attendance for reasons apart from death were similar in HIV-negative and HIV-positive persons (Table 1), although HIV-positive individuals were slightly more likely to have moved or travelled.
The overall mortality rate of this adult cohort was 9.2 per 1000 person-years. In HIV-negative individuals, the mortality rate increased with age, from 3.6 in those aged 15-19 years to 10.7 in those aged 45 years and over (Table 2). In HIV-positive individuals, the overall mortality rate was 93.5 per 1000 person-years, with similar rates for those aged over 20 years. Excess mortality was observed in HIV-positive persons of both sexes and all age-groups. Age-adjusted rate ratios were 18.93 [95% confidence interval (CI), 11.19-28.79] for women, 12.75 (95% CI, 8.32-20.26) for men, and 15.68 (95% CI, 11.18-21.03) overall.
The proportion of deaths attributable to HIV infection (PAF) was 35% overall, and 53% amongst those aged 20-29 years. The PAF was higher in women (41%) than in men (30%).
A verbal autopsy was administered for 178 (91%) of the 196 confirmed deaths. In the remaining 18 cases, the family of the deceased had moved out of the community. In 40 (22%) cases, the respondent was a spouse of the deceased, in 15 (8%) a child of the deceased, and in 27 (15%) a parent of the deceased (Table 3). Other respondents included brothers (n = 23), sisters (n = 14), other blood relatives (n = 23), and marriage in-laws (n = 19). For the remaining 17 deaths (9%), a close relative who attended the deceased could not be found and the respondent was a neighbour, village leader or traditional healer who had been present during the period of death. The respondent was less likely to be the spouse or child of the deceased for HIV-positive deaths, and this was partly because only 37 (58%) HIV-positive persons who died were married at the time of the baseline survey (19 men and 18 women), compared with 84 (74%) HIV-negative persons. Even amongst those who were married, however, it was less likely that spouses of children of HIV-positive individuals could be traced as respondents than for HIV-negative persons (Mantel-Haenszel χ2 adjusted for sex = 7.75; P < 0.005).
Specific symptoms reported during the precoded section of the verbal autopsy are shown in Table 4. Symptoms significantly associated with HIV-positive deaths included fever, rash, weight loss, anaemia, cough, chest pain, abdominal pain and headache. Diarrhoea was not significantly associated with HIV infection, but was mentioned by only three respondents. Symptoms reported for HIV-positive men and women were generally similar, but cough for more than 30 days was reported more often for male deaths (P = 0.02), whereas anaemia and headache were reported more often for female deaths (P = 0.05 and 0.09, respectively).
On the basis of verbal autopsy reports, the WHO clinical case definition for AIDS was satisfied for only 13 (7%) out of 178 cases, of whom only seven were HIV-positive, giving a sensitivity of 11% and positive predictive value (PPV) of 54%. HIV-positive men who died were significantly more likely to satisfy the case definition for AIDS than HIV-positive women (P = 0.01), none of whom satisfied the definition. Using the simplified case definition, 40 (22%) out of 178 deaths were classified as AIDS, of which 27 were HIV-positive, giving a sensitivity of 55% and PPV of 67%.
AIDS/HIV was mentioned during the verbal autopsy interview by only seven repondents (4%), and in five of these cases the deceased was HIV-positive (four men and one woman). The respondent stated that witchcraft was the cause of death in 20 cases (11%), five of whom were HIV-positive.
Discussion
In this rural cohort, the mortality rate among HIV-negative individuals was 6.0 per 1000 person-years, similar to adult mortality rates recorded in other countries in the developing world [24-26]. The overall mortality rate among HIV-positive adults was 93.5 per 1000 person-years or nearly 10% per annum, similar to previous studies in Africa [7-9,27] and higher than in developed countries, and since seroconversion dates were not known we have no direct information on the incubation period. However, the age-adjusted rate ratio of 15.7 confirms the strong association between HIV infection and mortality in this rural African population.
Although mortality increased with age as expected in the HIV-negative persons, there was no clear trend in HIV-positive persons with a rate close to 10% per annum in all age-groups except the youngest. Thus, there was no evidence that the incubation period decreased with age as in some developed country populations. The lower rate in those aged 15-19 years was based on small numbers, but may reflect more recent dates of infection in these young persons.
As nearly 30% of the cohort were not seen at follow-up, these mortality estimates are subject to bias. Apart from the deaths, most of the losses were individuals who had moved away from the community or who were travelling. Mobility is generally higher in the healthy than the sick, suggesting that mortality may have been lower among those lost to follow-up. On the other hand, some individuals move back to their home villages to die in this part of Africa, implying the opposite effect. It is unclear whether the net effect in our cohort would be an underestimation or overestimation of true mortality. However, the similarity in attendance rates and reasons for non-attendance in HIV-positive and HIV-negative individuals, apart from the deaths, argues against a substantial bias in the comparison of mortality in these two groups.
HIV-positive individuals made up only 4% of this adult cohort, but 37% of deaths occurred in this group. PAF calculations indicated that 35% of deaths were attributable to HIV, implying that HIV has increased overall adult mortality in these communities by more than 50%. The higher PAF of 53% in those aged 20-29 years implies that mortality has more than doubled in this age-group. These findings are in accordance with those of rural cohort studies in the Masaka and Rakai districts of Uganda [7-9], in which HIV prevalences of 8 and 13% were associated with PAF estimates of around 50%. In a cohort study of factory workers in urban Mwanza, the prevalence of HIV was 11% and the PAF was 62% [27]. In agreement with these three studies, we found the excess mortality associated with HIV infection to be higher in women than in men. However, in our cohort this reflected lower background mortality among HIV-negative women than men, rather than higher mortality in HIV-positive women.
HIV prevalence in rural Mwanza appears to have increased from about 2.5% in 1990/1991 to 4% in 1992 [15,22]. Between 1992 and 1994, the results of the intervention trial [23] show an incidence of HIV infection in this population of about 1% per annum in areas without the STD control intervention. This implies that new HIV infections should exceed deaths of HIV-positive individuals by a factor of about two. Results from the trials cohort support this conclusion. During 2 years of follow-up, 73 deaths were recorded in HIV-positive individuals, considerably fewer than the 130 seroconversions. This suggests that the epidemic has yet to stabilize in this population, that the prevalence of HIV infection is continuing to rise, and that HIV-associated mortality will increase still further in years to come.
Most deaths in this population occur at home, and medical diagnoses of cause-of-death are not available. To obtain information on the signs and symptoms of cohort members who died, we used an adult verbal autopsy instrument that has previously been applied in Tanzania, Ethiopia and Ghana. It was possible to identify a suitable respondent and to complete the verbal autopsy in all cases, except for 18 (9%) in which the family had moved out of the community since the death. Respondents for HIV-positive deaths were less likely to be spouses or children of the deceased than for HIV-negative deaths. Since close relatives may be better able to recall symptoms, it is possible that some symptoms were underreported for HIV-positive deaths. However, all respondents had looked after the deceased, and appeared able to recall the symptoms preceding death.
The reliability of a verbal autopsy interview in ascertaining signs and symptoms is limited by the ability of respondents to recall and report these observations. Recall may be of special concern in this study, since interviews may have taken place nearly 3 years after death in some cases. Despite these limitations, the instrument was able to detect clear differences in reported symptoms between HIV-positive and HIV-negative deaths, despite the interviewer being blinded to the HIV status of the deceased. Facilities for HIV testing and counselling in this rural population were very limited at the time of the study, and it can be assumed that few individuals in our cohort were aware of their HIV status. It follows that verbal autopsy respondents were also blind to the HIV status of the deceased in most cases.
Although a wide range of symptoms were associated with HIV infection, no single symptom was able to discriminate effectively between HIV-positive and HIV-negative deaths. When symptoms were combined using the WHO clinical case definition, only 13 deaths satisfied these criteria and only seven of these were HIV-positive at baseline. The poor sensitivity and specificity of the clinical case definition is in accordance with other studies [11,28], although few of the deceased were seen at a hospital or clinic before death so that the more definitive signs and symptoms of AIDS could not be recorded. A simplified definition, based on the presence of any two of four stated symptoms, performed rather better with a sensitivity of 42%, but the PPV was still only 67%.
The symptoms used in these case definitions are associated with a range of tropical and infectious diseases that are found commonly in this rural population, including malaria, tuberculosis and pneumonia, and this may help to explain their low specificity. Moreover, their poor sensitivity suggests that many HIV-positive individuals die without manifesting the clinical pattern traditionally associated with AIDS. For example, a surprising finding was that diarrhoea was reported in very few cases, and although this may partly reflect recall or reporting errors, we would not expect respondents to readily forget prolonged or severe episodes of diarrhoea.
These observations are of some relevance to the development of guidelines for appropriate clinical care of HIV-positive patients, since they suggest that rural health workers are unlikely to diagnose HIV disease reliably on the basis of simple signs and symptoms. It is therefore important that HIV serology is made available at health units wherever this is feasible.
Very few respondents reported that the death they had observed was associated with HIV or AIDS, but some reported that witchcraft played a role in the death. Many traditional beliefs are held by rural African communities, and these beliefs can sometimes be used to deny the consequences of HIV infection [29].
In a previous survey in this area, fewer than 25% of respondents thought that HIV/AIDS was a problem in their village [30]. Whereas HIV has clearly had a substantial effect on mortality in this population, the overall adult mortality rate of 9 per 1000 person-years is still relatively low, and the increase may have occurred too gradually to be clearly recognized, particularly in the absence of specific signs and symptoms. In many rural parts of Africa, it has been difficult to achieve substantial changes in risk behaviour to combat the HIV epidemic. This may be due partly to the difficulty of persuading rural dwellers of the threat to health posed by HIV in their villages.
Acknowledgements
We thank the Principal Secretary, Ministry of Health, the Manager of the National AIDS Control Programme, and the Director General of the National Institute for Medical Research, Tanzania, for permission to carry out and publish the results of this study. We acknowledge the support given by the Regional Medical Officer Mwanza, the Director of the National Institute for Medical Research and the Director of Bugando Medical Centre, Mwanza. We thank the field team for their hard work and national, regional, district and community leaders for their help. We are grateful to D. Chandramohn and G. Maude for making the verbal autopsy instrument available for our use. Above all, we would like to thank the families of all those interviewed for their patience and assistance.
References
1. World Health Organization: The current global situation of the HIV/AIDS pandemic. Wkly Epidemiol Rec 1995, 70:7-8.
2. Piot P, Laga M, Ryder R, et al.: The global epidemiology of HIV infection: continuity, heterogeneity and change. J Acquir Immune Defic Syndr 1990, 3:403-412.
3. Rutherford GW, Lifson AR, Hessol NA, et al.: Course of HIV-1 infection in a cohort of homosexual and bisexual men: an 11 year follow up. Br Med J 1990, 301:1193-1198.
4. Goedert JJ, Kessler CM, Aledort LM, et al.: A prospective study of human immunodeficiency virus type 1 infection and the development of AIDS in subjects with hemophilia. N Engl J Med 1989, 321:1141-1148.
5. Colebunders R, Ryder R, Francis H, et al.: Seroconversion rate, mortality and clinical manifestations associated with the receipt of an human immunodeficiency virus infected blood transfusion. J Infect Dis 1991, 164:450-456.
6. Anzala OA, Nagelkerke NJD, Bwayo JJ, et al.: Rapid progression to disease in African sex workers with human immunodeficiency virus type 1 infection. J Infect Dis 1995, 171:686-689.
7. Mulder DW, Nunn AJ, Wagner H-U, Kamali A, Kengeya-Kayondo JF: HIV-1 incidence and HIV-1-associated mortality in a rural Ugandan population cohort. AIDS 1994, 8:87-92.
8. Mulder DW, Nunn AJ, Kamali A, Nakiyingi J, Wagner U, Kengeya-Kayondo JF: Two-year HIV-1-associated mortality in a Ugandan rural population. Lancet 1994, 343:1021-1023.
9. Sewankambo NK, Wawer MJ, Gray RH, et al.: Demographic impact of HIV infection in rural Rakai District, Uganda: results of a population-based cohort study. AIDS 1994, 8:1707-1713.
10. Nelson AM, Hassig SE, Kayembe M, et al.: HIV-1 seropositivity and mortality at University Hospital, Kinshasa, Zaire, 1987. AIDS 1991, 5:583-586.
11. De Cock KM, Barrere B, Diaby L, et al.: AIDS the leading cause of adult death in the West African city of Abidjan, Ivory Coast. Science 1990, 249:793-796.
12. Gregson S, Garnett GP, Anderson R: Is HIV-1 likely to become a leading cause of adult mortality in sub-Saharan Africa? J Acquir Immune Defic Syndr 1994, 7:839-852.
13. Killewo J, Nyamuryekung'e K, Sandstrom A, et al.: Prevalence of HIV-1 infection in the Kagera region of Tanzania: a population-based study. AIDS 1990, 4:1081-1085.
14. Serwadda D, Wawer MJ, Musgrave SD, Savankambo NK, Kaplan JE, Gray RH: HIV risk factors in three geographic strata of rural Rakai District, Uganda. AIDS 1992, 6:983-989.
15. Barongo LR, Borgdorff MW, Mosha FF, et al.: The epidemiology of HIV-1 infection in urban areas, roadside settlements and rural villages in Mwanza Region, Tanzania. AIDS 1992, 6:1521-1528.
16. Colebunders RL, Latif AS: Natural history and clinical presentation of HIV-1 infection in adults. AIDS 1991, 5 (suppl 1):S103-S112.
17. Gilks CF, Otieno LS, Brindle RJ, et al.: The presentation and outcome of HIV-related disease in Nairobi. Q J Med 1992, 82:25-32.
18. Morrow RH, Colebunders RL, Chin J: Interactions of HIV infection with endemic tropical diseases. AIDS 1989, 3 (suppl 1):S79-S87.
19. World Health Organization: WHO case definitions for AIDS surveillance in adults and adolescents. Wkly Epidemiol Rec 1994, 69:273-280.
20. De Cock K, Selik R, Soro B, Gayle H, Colebunders R: AIDS surveillance in Africa: a reappraisal of case definitions. Br Med J 1991, 303:1185-1188.
21. Hayes R, Mosha F, Nicoll A, et al.: A community trial of the impact of improved sexually transmitted disease treatment on the HIV epidemic in rural Tanzania: 1. Design. AIDS 1995, 9:919-926.
22. Grosskurth H, Mosha F, Todd J, et al.: A community trial of the impact of improved sexually transmitted disease treatment on the HIV epidemic in rural Tanzania: 2. Baseline results. AIDS 1995, 9:927-934.
23. Grosskurth H, Mosha F, Todd J, et al.: The impact of improved STD treatment on the incidence of HIV infection in rural Tanzania: a randomised controlled trial. Lancet 1995, 346:530-536.
24. Timaeus JM: Adult mortality: levels, trends and data sources. In Disease and Mortality in Sub-Saharan Africa. Edited by Feachem RG, Jamison DT. New York: Oxford University Press; 1991:87-100.
25. The World Bank: World Development Report 1993, Investing in Health. New York: Oxford University Press; 1993.
26. United Nations: United Nations Demographic Yearbook 1992. New York: UN Publishing Division; 1994.
27. Borgdorff MW, Barongo LR, Klokke AH, et al.: HIV-1 incidence and HIV-1 associated mortality and morbidity in a cohort of urban factory workers in Tanzania. Genitourin Med 1995, 71:212-215.
28. Belec L, Brogan T, Keou F, Georges AJ: Surveillance of acquired immunodeficiency syndrome in Africa: an analysis of evaluations of the World Health Organization and other clinical definitions. Epidemiol Rev 1994, 16:403-417.
29. Nicoll A, Laukamm-Josten U, Mwizarubi B, et al.: Lay health beliefs concerning HIV and AIDS: a barrier for control programmes. AIDS Care 1993, 5:223-233.
30. Munguti K, Grosskurth H, Newell J, et al.: Patterns of sexual behaviour in a rural population in north-western Tanzania. Soc Sci Med 1997 (in press).
© Lippincott-Raven Publishers.