In June 2003, the United States Military HIV Research Program began a 3-year, closed, prospective, combined community-work place cohort study of 2801 adults.1,2 The study was designed to capture a cohort reflecting one that would be recruited in a future vaccine study; assess the suitability of a tea plantation in rural Kericho, Kenya, for the evaluation of candidate vaccines; and, to develop the vaccine, reagents, infrastructure, and databases required to plan and execute these trials in the future. One study primary objective was to estimate the incidence of HIV-1 in adult workers and dependants on the tea plantation. Similar HIV-1 incidence and related epidemiological data have been reported from Uganda, Tanzania, Malawi, Zimbabwe, and elsewhere.3-13
Although related, HIV observational cohort data in Kenya have been published, most, if not all, of the data have been from urban and higher risk populations14-16 or, in the case of incidence rates methods, may utilize estimates based upon the BED HIV-1 Capture EIA (BED) assay with or without supplemental adjustments.17-19 The Kericho HIV cohort study utilized rigorous, gold standard laboratory assays (ie, enzyme-linked immunosorbent assay and Western blot) and provides primary HIV-1 incidence data from a traditionally nonhigh risk population in a rural setting. We present HIV-1 incidence rates and risk factors associated with seroconversion from this study and discuss their implications for future HIV programs and clinical trials.
Description of the Kericho HIV Cohort Study
The Kericho HIV Cohort Study is formally known as the HIV and Malaria Cohort Study among Plantation Workers and Adult Dependents in Kericho, Kenya. This is a 36-month, closed, prospective study aimed at estimating HIV prevalence, incidence, comorbidities, molecular epidemiology, and vaccine feasibility and acceptability among adult tea plantation workers and dependents in Kericho, Kenya. The study was conducted on a large tea plantation on the outskirts of Kericho, a town of approximately 150,000 in the southern Rift Valley Province of Kenya. In this region, the HIV prevalence rates range from approximately 6% to 14%, likely varying in part due to differing surveillance methodologies used and populations studied.17,20,21
Details of recruitment and laboratory testing of this study have been described elsewhere.21,22 Briefly, after full review and approval by both the Kenya Medical Research Institute and the Walter Reed Army Institute of Research Institutional Review Boards, 2801 adult plantation workers and dependent volunteers aged 18-55 years were recruited primarily through the use of open “barazas” (community gatherings) over a 6-month period beginning in June 2003. After providing written informed consent in either Kiswahili or English, study volunteers were enrolled in the baseline cohort and followed every 6 months for 3 years, with the final follow-up ending in December 2006. For this study, all prevalent HIV cases (n = 401) were excluded, resulting in 2400 participants comprising our analytic cohort.
Data, HIV Incidence Case Definition, and Study Related Care
Study volunteers completed extensive baseline questionnaires collecting sociodemographic, medical, behavioral, and HIV risk data. Abbreviated follow-up questionnaires focusing upon interim medical history and HIV risks and laboratory evaluations were conducted at each 6-month follow-up. Incident cases were defined as HIV-1 seroconversions using an algorithm consisting of 2 sequential enzyme-linked immunosorbent assays and confirmatory Western blot testing.22 Volunteers initially identified as HIV incident cases returned for repeat phlebotomy and verification HIV testing. Study volunteers received individual pretest counseling prior to each HIV test including general HIV education and prevention and information regarding confidentiality of HIV test results. Individual posttest counseling occurred approximately 2 weeks after HIV testing or verification where applicable and included education regarding healthy lifestyles for HIV-negative volunteers and immediate and ongoing psychosocial support, and referral for further evaluation, clinical staging, and treatment where indicated for those who tested HIV positive.
Care provided to all volunteers throughout the study included treatment for common medical ailments (eg, respiratory tract infections, dermatologic conditions, parasitic and malarial infections) with referral to the local plantation or Ministry of Health facility depending upon the volunteer's preference when indicated. For volunteers who tested HIV positive, prophylaxis for opportunistic infections (eg, trimethoprim/sulfamethoxazole) and multivitamins were offered. All pregnant HIV-infected women were offered nevirapine to prevent transmission of HIV to their infant based upon guidelines at the time and as part of a Prevention of Mother to Child Transmission of HIV program. Antiretroviral therapy according to the Kenya Ministry of Health guidelines became available during the third 6-month follow-up and was provided by either the plantation or local Ministry of Health facility.
Descriptive statistics were used to compare baseline characteristics between men and women comprising the baseline cohort. Continuous data were compared by Student t test, and categorical data were compared by Chi-square or Fisher exact test where appropriate. Incidence density rates were calculated per 100 person-years (py). Time to HIV seroconversion was calculated using the midpoint imputation (interval defined by the latest screening test that was negative for infection and the earliest screening test that was positive). This procedure can produce biased parameter estimates due to interval-censored data, but the magnitude of this bias is small if the interval width is 2 years or less.23,24
HIV-1 incidence rates per 100 py were calculated overall and by study year based upon the number of volunteers participating in the respective study follow-ups and person-time contributed. HIV survival curves were compared by using the Log-rank test. Four parametric survival distributions including the exponential, Weibull, log-normal, and log-logistic were evaluated to model survival. The fit of the distributions was compared using the Akaike information criterion. On the basis of the Akaike information criterion, all distributions reported quite similar results for goodness of fit. The Weibull distribution was used to model survival and estimate hazard rate (HR) ratios along with their 95% confidence intervals (CIs). The smoothed hazard function was estimated using the Kernel smoothing method. All significant variables identified in univariate analyses were tested for effect modification with HIV status using 1-way interaction terms. All statistical analyses were carried out using STATA version 9.0 (Stata Corporation, College Station, TX).
Most (63.5%) of the 2400 participants were men who were older than women (mean = 30.8 and 28.8 years, respectively) (P < 0.001, Table 1). Four tribes from the region were most commonly represented with more men being from the Kalenjin (32.7% vs. 24.1% ) and Kisii (25.8% vs. 24.3%) tribes (more common in the Kericho region) and more women from the Luhya (20.9% vs. 16.3%) and Luo (20.0% vs. 16.7%) tribes (more representative of western Kenya) (P < 0.001). Overall, men were more likely to have attended high school or greater (48.5% vs. 34.3%), more likely to be employed by the plantation (84.4% vs. 48.9%), and more likely to have been married 2 or more times (17.3% vs. 6.8%) (all P < 0.001).
The majority of men (80.4%) in our cohort were circumcised. More than half (57.2%) of the cohort had their onset of sexual intercourse to be between 15 and 18 years, and approximately half of the cohort (50.6%) had been sexually active for 7-18 years. With 22.9% of the cohort having 9 or more sexual interchanges in the last month, 18.9% reported having sex with individuals who were not their regular partner/spouse only. Although only 5.1% of participants reported a sexually transmitted infection (STI) in the previous 5 years, 10.8% reported STI symptoms in the previous 6 months (12.7% women, 9.9% men, P = 0.04). Nearly 8% of men (7.6%) reported having sex with a commercial sex worker (CSW) in the past 6 months, and overall 7.4%-11.2% reported engaging in some type of transactional sex (sex for goods or food for sex). Few participants (2.7%) reported a past blood transfusion.
Over the 3-year follow-up period, 2196 (91.5%) of participants in the baseline cohort had at least 1 follow-up visit, and 2002 (83.4%) participated in the final follow-up evaluation, resulting in 6285.4 py of follow-up (women = 2279.7 py; men = 4005.7 py). During follow-up, 29 (15 women) seroincident HIV cases were identified during year 1, 20 (5 women) cases during year 2 (49 cumulative), and 14 (3 women) cases during year 3 (63 cumulative, 23 women). The corresponding 12-, 24-, and 36-month HIV incidence rates were 1.41 (95% CI = 0.95-2.02), 1.16 (95% CI = 0.86-1.54), and 1.00 (95% CI = 0.77-1.28) per 100 py of follow-up. By gender, the corresponding cumulative incidence rates per 100 py of follow-up: 12 months: women = 1.98 (95% CI = 1.11-3.25) and men = 1.08 (95% CI = 0.59-1.81); 24 months: women = 1.31 (95% CI = 0.80-2.02) and men = 1.07 (95% CI = 0.72-1.55); and, 36 months: women = 1.01 (95% CI = 0.64-1.51) and men = 1.00 (95% CI = 0.71-1.36).
Most incident cases (60 of 63) had a lag time less than 2 years (median = 6.8 months) between baseline examination and seroconversion. The shape of hazard function of HIV seroconversion revealed a peak around the end of the first year followed by a decreasing trend in the second year (Figure 1). The cumulative hazard rate of HIV infection decreased over time (14% per month).
Several baseline sociodemographic, behavioral, and medical risk factors were significantly associated with HIV seroconversion (Table 2). Members of the Luo tribe, predominately from Western Kenya traditionally not practicing circumcision, were at highest risk of HIV infection (HR = 3.31; 95% CI = 1.65-6.63). Participants who were married more than once were almost 3 times more likely to become HIV infected (HR = 2.83; 95% CI = 1.20-6.69) than those who were never married. Circumcised men had a nearly 70% reduction in HIV acquisition (HR = 0.32; 95% CI = 0.17-0.60) compared with uncircumcised men. Participants with a self-reported history of a STI were more than twice as likely to acquire HIV (HR = 2.40; 95% CI = 1.09-5.26) as those who reported never having had an STI. Persons having transactional sex (ie, providing sex for goods) and a history of alcohol abuse during sex were also at increased risk of incident HIV (HR = 3.30; 95% CI = 1.79-6.09 and HR = 2.44; 95% CI = 1.16-5.13, respectively). No significant HIV status effect modification by baseline sociodemographic, medical, behavioral, or other HIV risk factors was observed.
We found 36-month HIV-1 incidence to be low but notably higher among women in adult plantation workers and dependents in rural Kenya. One major factor potentially contributing to this low incidence rate (and lower incidence compared with women) is that male circumcision is widely practiced in this region of Kenya. About 80% of men in the population from which the cohort was drawn were circumcised, mostly by traditional practices (75%). The significant finding of a lower HIV incidence among circumcised men in this cohort has been previously described.25
The decrease in HIV incidence rates during the 3 years of follow-up, and the sharp decline in the hazard rates between the first and second years of the study are expected findings and have been observed similarly elsewhere.15 Given the Kericho HIV Cohort Study is a closed cohort, individuals at highest risk would be expected to acquire HIV infection early and were not replaced by other high risk individuals from the greater population. Over time, the pool of at-risk individuals in the cohort would naturally decrease until, given sufficient follow-up time, few if any remaining cohort members would be at risk of acquiring HIV. The educational and counseling messages provided to participants at each follow-up visit may also have contributed to a decline over time in HIV incidence. The plateau in the HIV HR is also reasonable to expect given the cohort HIV prevalence rate of approximately 14%.21 New HIV infections would be expected to occur, albeit at a low level, as long as some cohort members continued interactions with individuals in the larger community who are at risk of HIV infection themselves.
The potential for HIV infections to increase as behaviors continue to occur among cohort members and other at-risk individuals, and notable modifiable risk factors (eg, transactional sex, polygamy, extramarital sexual relations, history of STI) provide opportunities for targeted HIV prevention campaigns. Both the Government of Kenya and the US-based President's Emergency Plan for AIDS Relief currently sponsor such campaigns in this region of Kenya.26,27
A strength of this study is the ability to provide rural, primary, long-term HIV-1 incidence rates within the context of a prospective, noninterventional, cohort study. In addition to providing directly observed estimates, the Kericho HIV Cohort Study used rigorous research methods to establish HIV diagnoses (sequential enzyme-linked immunosorbent assay with confirmatory Western blot). Extensive baseline questionnaires allowed assessment of sociodemographic and other (eg, medical) risks for incident HIV infection. Although 83.4% of the cohort was retained for 3 years, a notable achievement for the first study of its type in Kenya, the impact of the approximately 17% loss to follow-up likely represents the primary limitation of our study. This may impart bidirectional biases on our incidence estimates reported that are based upon volunteers participating in the respective study follow-ups: either an overestimated incidence rate if all those lost to follow-up were HIV negative; or, an underestimated incidence rate if all those lost to follow-up were HIV positive. We estimate the potential impact of this 17% loss to follow-up on our overall, unadjusted 36-month incidence rate [1.00 (95% CI = 0.77-1.28)] to result in an overall incidence between 0.09 (if all lost to follow-up were HIV negative) and 3.87 (if all lost to follow-up were HIV positive) per 100 py depending upon the HIV status of those lost to follow-up. In review of the baseline variables comparing those lost to follow-up to those who completed the study (data not shown), those lost to follow-up were younger, had less years of sexual activity, were never married, and were unemployed. These factors and more details related to our cohort retention are described in detail in a dedicated manuscript anticipated to follow this publication. Finally, the relatively few HIV infections may have limited our study power to detect some significant risk factors (eg, nonsignificant HR for persons having sex with known HIV-infected individuals and some traditional risk factors such as young age at sexual debut).
Caution is necessary in drawing conclusions based upon direct, side-by-side comparisons of other published HIV-1 incidence rates due to different populations studied and methods applied, temporal relationships with regard to the HIV epidemic, potential prevention campaigns ongoing, and loss to follow-up rates. However, general observations and comparisons may be made with regard to HIV-1 incidence rates per 100 py observed in the Kericho cohort to other cohorts. Data from 2 similar prospective cohort studies in Uganda and a population based survey with 12-month follow-up in Tanzania reported incidence rates of approximately 2.1 at 1 year in the Rakai cohort,3 1.4 in the larger Kagera region at 1 year,8 and 0.7 at 4 years in the Masaka cohort.6 These are similar and with overlapping confidence intervals to 1- and 3-year incidence rates in the Kericho cohort: 1.4 and 1.0, respectively. In comparison, HIV incidence rates among women of reproductive age in Malawi and Zimbabwe seem to be high (4.5-4.7)11,13 compared with women in rural Kericho (approximately 2.0 at 1 year). Similarly, 12-month incidence rates among male, sugar estate workers in rural Malawi reported at 17.1 in 1994-1995 and subsequently 3.8 in 1998-199912 are higher than those in men in the Kericho cohort, 1.1. Finally and consistent with what would be expected in higher risk populations, incidence rates among prostitutes and truck drivers in Nairobi and Mombasa, Kenya, were markedly higher than corresponding incidence rates in the Kericho cohort.14,16
In conclusion, 36-month incidence rates were relatively low in adult plantation workers and dependents in rural Kenya. Such findings support the use of higher risk populations (eg, STI clinic attendees, CSWs, truck drivers, discordant couples) in long-term HIV prevention studies in this region, particularly those which are endpoint driven (eg, HIV preventive vaccine trials, preexposure prophylaxis studies). Although one would expect incidence to decrease in such a closed HIV cohort study, analyses are planned to evaluate the impact of HIV education and prevention campaigns provided to cohort members on a regular basis on declining HIV incidence rates.
The Walter Reed Army Institute of Research Institutional Research Board human use protocol #855 (RV142), “HIV and Malaria Cohort Study Among Plantation Workers and Adult Dependents in Kericho, Kenya,” is funded through the United States Military HIV Research Program (the Walter Reed Army Institute of Research and the Henry M. Jackson Foundation for the Advancement of Military Medicine Inc). The authors are grateful to the dedicated and committed 2801 participants of the Kericho HIV Cohort Study. This work is published with the permission of the Director, Kenya Medical Research Institute, Nairobi, Kenya.
2. “HIV and Malaria Cohort Study Among Plantation Workers and Adult Dependents in Kericho, Kenya,” The Walter Reed Army Institute of Research Institutional Research Board human use protocol no. 855 (RV142) and Kenya Medical Research Institute Scientific Steering Committee protocol no. 590.
3. Wawer MJ, Sewankambo NK, Berkley S, et al. Incidence of HIV-1 infection in a rural region of Uganda. BMJ
4. Wawer MJ, Serwadda D, Gray RH, et al. Trends in HIV-1 prevalence may not reflect trends in incidence in mature epidemics: data from the Rakai population-based cohort, Uganda. AIDS
5. Wawer MJ, Gray RH, Sewankambo NK, et al. Rates of HIV-1 transmission per coital act, by stage of HIV-1 infection, in Rakai, Uganda. J Infect Dis
6. Kengya-Kayondo J, Kamali A, Nunn A, et al. Incidence of HIV-1 infection in adults and socio-demographic characteristics of seroconverters in a rural population in Uganda: 1990-1994. Int J Epidemiol
7. Mulder D, Nunn A, Wagner H, et al. HIV-1 incidence and HIV-1-associated mortality in a rural Ugandan population cohort. AIDS
8. Killewo J, Sandstrom A, Raden U, et al. Incidence of HIV-1 infection among adults in the Kagera region of Tanzania. Int J Epidemiol
9. Kwesigabo G, Killewo J, Urassa W, et al. HIV-1 infection prevalence and incidence trends in areas of contrasting levels of infection in the Kagera region, Tanzania, 1987-2000. J Acquir Immune Defic Syndr
10. Wambura M, Urassa M, Isingo R, et al. HIV prevalence and incidence in rural Tanzania: results from 10 years of follow-up in an open-cohort study. J Acquir Immune Defic Syndr
11. Kumwenda N, Kumwenda J, Kafulafula G, et al. HIV-1 incidence among women of reproductive age in Malawi. Int J STD AIDS
12. Kumwenda N, Taha T, Hoover D, et al. HIV-1 incidence among male workers at a sugar estate in rural Malawi. J Acquir Immune Defic Syndr
13. Kumwenda N, Hoffman I, Chirenje M, et al. HIV incidence among women of reproductive age in Malawi and Zimbabwe. Sex Transm Dis
14. Fowke K, Nagelkerke N, Kimani J, et al. Resistance to HIV-1 infection among persistently seronegative prostitutes in Nairobi, Kenya. Lancet
15. Baeten J, Richardson B, Martin H Jr, et al. Trends in HIV-1 incidence in a cohort of prostitutes in Kenya: implications for HIV-1 vaccine efficacy trials. J Acquir Immune Defic Syndr
16. Jackson D, Rakwar J, Richardson B, et al. Decreased incidence of sexually transmitted diseases among trucking company workers in Kenya: results of a behavioural risk-reduction programme. AIDS
17. National AIDS and STI Control Programme, Ministry of Health, Kenya. July 2008. Kenya AIDS Indicator Survey 2007: Preliminary Report
. Nairobi, Kenya: NASCOP; 2008.
20. National AIDS and STI Control Programme, Ministry of Health, Kenya. AIDS in Kenya
. 7th ed. Nairobi, Kenya: NASCOP; 2005.
21. Foglia G, Sateren WB, Renzullo PO, et al. High prevalence of HIV infection among rural tea plantation residents in Kericho, Kenya. Epidemiol Infect
22. Foglia G, Royster G, Wasunna K, et al. Use of rapid and conventional testing technologies for human immunodeficiency virus type 1 serologic screening in a rural Kenyan reference laboratory. J Clin Microbiol
23. Law CG, Brookmeyer R. Effect of mid-point imputation on the analysis of doubly censored data. Stat Med
24. Williamson JM, Satten GA, Hanson JA, et al. Analysis of dynamic cohort data. Am J Epidemiol
25. Shaffer DN, Bautista CT, Sateren WB, et al. The protective effect of circumcision on incident HIV in rural, low risk men circumcised predominately by traditional healers in Kenya: two-year follow-up of the Kericho HIV Cohort Study. J Acquir Immune Defic Syndr
26. National AIDS/STD Control Program (NASCOP homepage), Ministry of Health, Government of Kenya. Available at: http://www.aidskenya.org/
. Accessed December 30, 2008.
27. The US President's Emergency Plan for AIDS Relief (PEPFAR homepage). Available at: http://www.pepfar.gov/
. Accessed December 30, 2008.
Keywords:© 2010 Lippincott Williams & Wilkins, Inc.
HIV; incidence; cohort; Kericho; Kenya