National expenditure on health research in South Africa: How has the landscape changed in the past decade?

BACKGROUND: Over the past 18 years, the South African (SA) Ministry of Health has committed to allocate 2% of the national health budget to research, while the National Health Research Policy (2001) proposed that the health research budget should be 2% of total public sector health expenditure. A review was conducted by the National Health Research Committee (NHRC) in 2014 to determine whether these goals had been met, using available data up to 2009/10. It revealed that public sector health research funding remained below 2% of the national health budget, supporting the perception of reduced public sector health research funding. OBJECTIVES: To provide an update on the previous review to investigate changes in the health research landscape since 2009/10 and whether goals have been met. METHODS: Various publicly available sources of information on public and private expenditure on health research in SA were used to investigate health research funding and expenditure. In addition, questionnaires were sent to 35 major national and international funders of health research in SA to obtain data on the level of funding provided and the fields of research funded. RESULTS: Total health research expenditure in SA was ZAR6.9 billion in 2016/17, or 19.2% of gross expenditure on research and development, with 1.7% of the ZAR38.6 billion National Department of Health budget from National Treasury being spent on health research through the South African Medical Research Council (ZAR658 million), corresponding to 0.4% of the consolidated government expenditure on health. However, although the total government plus science council spend on health research in 2016/17 was ZAR1.45 billion, this represents just 0.033% of the gross domestic product (GDP), thus remaining well below the aspirational target of 0.15% of the GDP set by the NHRC in 2014. Based on feedback from the funders, the estimated baseline health research funding in 2016/17 was in excess of ZAR4.1 billion, which is considerably higher than many researchers may realise. Three-quarters of this funding originated from foreign sources, suggesting both strengths and opportunities for health research in SA, but also highlighting increasing dependence on foreign funding. Notably, the majority of funders approached were not able to readily break down expenditure according to disease area. CONCLUSIONS: Health research funding has changed significantly since our previous review, although the government's own commitments to it remain unmet. Improved mechanisms to track health research expenditure are urgently required for better alignment of funding priorities and increased co-ordination between science councils in health research funding.

Good correlation between the Afinion AS100 analyser and the ABX Pentra 400 analyser for the measurement of glycosylated haemoglobin and lipid levels in older adults in Durban, South Africa

Background. The Afinion AS100 analyser is a small bench-top, multi-assay, point-of-care (POC) analyser that is able to measure glycated haemoglobin (HbA1c) and lipid levels.Objective. To assess performance of the Afinion analyser compared with a reference laboratory test for the measurement of HbA1c and lipid levels.Method. The study involved men and women enrolled in a cross-sectional study, Sexual health, HIV infection and comorbidity with non-communicable diseases among Older Persons (SHIOP), which was conducted from February to May 2016. Whole blood was drawn aseptically by a trained study nurse into a serum separator gel tube and an ethylenediaminetetra-acetic acid (EDTA) tube. The EDTA whole blood was used to measure HbA1c levels, and serum to measure total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C) and triglyceride levels. Lin's correlation coefficient was used to assess the agreement between the Afinion and ABX Pentra 400 analysers for each marker.Results. A total of 435 older individuals were included in the study. The proportion of HbA1c results that were correctly classified by the Afinion analyser was 92.2%. Bland-Altman analysis and linear regression analysis showed a very good agreement (correlation concordance 0.89) between the two analysers for the measurement of HbA1c. The two-way scatter plot for TC showed a substantial correlation (0.80). However, a total of 69 cholesterol results that were within the normal range on the Pentra were misclassified as abnormal on the Afinion. The readings obtained for HDL-C levels with the Afinion were shown to be slightly overestimated when compared with the Pentra. However, correlation for HDL-C on the two analysers was 0.93, indicating an almost perfect agreement. Seventy-four LDL-C results were erroneously classified as abnormal on the Afinion but were within the normal range on the Pentra, resulting in a substantial correlation of 0.75. An excellent agreement was observed between triglyceride measurements (0.99).Conclusion.This study supports the use of the Afinion AS100 analyser in POC testing for the measurement of HbA1c, triglycerides and HDL-C in a South African setting

Good correlation between the Afinion AS100 analyser and the ABX Pentra 400 analyser for the measurement of glycosylated haemoglobin and lipid levels in older adults in Durban, South Africa.

BACKGROUND: The Afinion AS100 analyser is a small bench-top, multi-assay, point-of-care (POC) analyser that is able to measure glycated haemoglobin (HbA1c) and lipid levels. OBJECTIVE: To assess performance of the Afinion analyser compared with a reference laboratory test for the measurement of HbA1c and lipid levels. METHOD: The study involved men and women enrolled in a cross-sectional study, Sexual health, HIV infection and comorbidity with non-communicable diseases among Older Persons (SHIOP), which was conducted from February to May 2016. Whole blood was drawn aseptically by a trained study nurse into a serum separator gel tube and an ethylenediaminetetra-acetic acid (EDTA) tube. The EDTA whole blood was used to measure HbA1c levels, and serum to measure total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), low-density lipoprotein cholesterol (LDL-C) and triglyceride levels. Lin's correlation coefficient was used to assess the agreement between the Afinion and ABX Pentra 400 analysers for each marker. RESULTS: A total of 435 older individuals were included in the study. The proportion of HbA1c results that were correctly classified by the Afinion analyser was 92.2%. Bland-Altman analysis and linear regression analysis showed a very good agreement (correlation concordance 0.89) between the two analysers for the measurement of HbA1c. The two-way scatter plot for TC showed a substantial correlation (0.80). However, a total of 69 cholesterol results that were within the normal range on the Pentra were misclassified as abnormal on the Afinion. The readings obtained for HDL-C levels with the Afinion were shown to be slightly overestimated when compared with the Pentra. However, correlation for HDL-C on the two analysers was 0.93, indicating an almost perfect agreement. Seventy-four LDL-C results were erroneously classified as abnormal on the Afinion but were within the normal range on the Pentra, resulting in a substantial correlation of 0.75. An excellent agreement was observed between triglyceride measurements (0.99). CONCLUSION: This study supports the use of the Afinion AS100 analyser in POC testing for the measurement of HbA1c, triglycerides and HDL-C in a South African setting.

Detecting Mycobacterium tuberculosis using the loop-mediated isothermal amplification test in South Africa.

SETTING: In South Africa, KwaZulu-Natal is the epicentre of the human immunodeficiency virus (HIV) epidemic, where approximately 70% of people with tuberculosis (TB) are co-infected with HIV. Undiagnosed TB contributes to high mortality in HIV-infected patients. Delays in diagnosing TB and treatment initiation result in prolonged transmission and increased infectiousness. OBJECTIVE: To evaluate the LoopampTM MTBC Detection kit (TB-LAMP; based on the loop-mediated isothermal amplification assay), smear microscopy and Xpert test with the gold standard of mycobacterial culture. METHODS: Sputum samples were collected from 705 patients with symptoms of pulmonary TB attending a primary health care clinic. RESULTS: The TB-LAMP assay had significantly higher sensitivity than smear microscopy (72.6% vs. 45.4%, P < 0.001), whereas specificity was slightly lower (99% vs. 96.8%, P = 0.05), but significantly higher than Xpert (92.9%, P = 0.004). There was no significant difference in sensitivity of smear-positive, culture-positive and smear-negative, culture-positive sputum samples using TB-LAMP vs. Xpert (respectively 95.9%/55.9% vs. 97.6%/66.1%; P =0.65, P = 0.27). The positive predictive value of TB-LAMP was significantly higher than that of Xpert (87.5% vs. 77.0%; P = 0.02), but similar to that of smear microscopy (94.2%; P = 0.18). The negative predictive value was respectively 91.9%, 92.5% (P = 0.73) and 83.1% (P = 0.0001). CONCLUSION: Given its ease of operability, the TB-LAMP assay could be implemented as a point-of-care test in primary health care settings, and contribute to reducing treatment waiting times and TB prevalence.

Poor Performance of the Chlamydia Rapid Test Device for the Detection of Asymptomatic Infections in South African Men: A Pilot Study.

Background. To the best of our knowledge, there have been no published reports on the diagnostic performance of the Chlamydia Rapid Test (CRT) Device for male urine samples. We evaluated the performance of the CRT Device when compared with that of the BD ProbeTec ET PCR Assay in a population of asymptomatic men. Methods. The study enrolled 100 men between June and July 2015. From each consenting male, 20-30 mL of urine was collected. Sensitivity and specificity of the rapid test compared to PCR were calculated. All analysis was performed in STATA version 13. Results. All men had valid rapid and PCR test results. The test showed a low sensitivity against PCR (20%) (95% CI 3.7-6.2%); however, an excellent specificity was observed (100%) (one sided 97.5% CI: 96.0-100). Conclusions. This test was not found to be suitable as a screening tool for genital Chlamydia infections in men. Our findings emphasize the need for more sensitive POC tests to be developed since the current approach for the management of STIs in Africa is confounded by poor sensitivity and specificity resulting in many infected individuals not being treated.

Spatial clustering of "measured" and "unmeasured" risk factors for HIV infections in hyper-endemic communities in KwaZulu-Natal, South Africa: results from geoadditive models.

Sub-Saharan Africa contains more than 60% of all HIV infections worldwide. HIV prevalence was currently estimated to be at least 15% in KwaZulu-Natal and the epidemic is described as hyper-endemic. Knowledge of spatial clustering of risk factors which are linked to new HIV infections is important for prioritizing areas to change the trajectory of the epidemic. Geoadditive models were used to investigate spatial characteristics of the risk factors from two clinical trial units (Umkomaas and Botha's Hill) in the province of KwaZulu-Natal, South Africa. Study population was a cohort of women who screened and enrolled in an HIV prevention biomedical intervention trial. The results suggest high HIV incidence rates (5.8 and 8 per 100 person-year). Considerable spatial variations in behavioural factors within a relatively small geographical region, low level of education, early age at sexual debut, higher number of sexual partners, not being married/cohabitating with a sexual partner and sexual activity in exchange for money, gift and drugs were all determined to be clustered in certain regions; they were accounted for 25% (Umkomaas) and 65% (Botha's Hill) of the excess new HIV infections in two clinical trial units. Results from our study highlighted existence of significant spatial heterogeneity in "measured" and "unmeasured" risk factors in a relatively small region. As the HIV funding has been declining, identifying, targeting and reaching the most-at-risk individuals will likely play a significant role in developing the most efficient and cost-effective prevention programmes and subsequently will change the trajectory of the epidemic.

Clinical evaluation of the OneStep Gonorrhea RapiCard InstaTest for detection of Neisseria gonorrhoeae in symptomatic patients from KwaZulu-Natal, South Africa.

We evaluated a point-of-care test for the detection of Neisseria gonorrhoeae in patients attending a public health clinic in KwaZulu-Natal, South Africa. The test showed a low sensitivity against PCR and culture (<40%); however, a higher specificity was observed (>95%). This test is unsuitable as a screening tool for gonorrhea.

Provision of long-acting reversible contraception in HIV-prevalent countries: results from nationally representative surveys in southern Africa.

OBJECTIVE: To analyse the current provision of long-acting reversible contraception (LARC) and clinician training needs in HIV-prevalent settings. DESIGN: Nationally representative survey of clinicians. SETTING: HIV-prevalent settings in South Africa and Zimbabwe. POPULATION: Clinicians in South Africa and Zimbabwe. METHODS: Nationally representative surveys of clinicians were conducted in South Africa and Zimbabwe (n = 1444) to assess current clinical practice in the provision of LARC in HIV-prevalent settings. Multivariable logistic regression was used to analyse contraceptive provision and clinician training needs. MAIN OUTCOME MEASURE: Multivariable logistic regression of contraceptive provision and clinician training needs. RESULTS: Provision of the most effective reversible contraceptives is limited: only 14% of clinicians provide copper intrauterine devices (IUDs), 4% levonorgestrel-releasing IUDs and 16% contraceptive implants. Clinicians' perceptions of patient eligibility for IUD use were overly restrictive, especially related to HIV risks. Less than 5% reported that IUDs were appropriate for women at high risk of HIV or for HIV-positive women, contrary to evidence-based guidelines. Only 15% viewed implants as appropriate for women at risk of HIV. Most clinicians (82%), however, felt that IUDs were underused by patients, and over half desired additional training on LARC methods. Logistic regression analysis showed that LARC provision was largely restricted to physicians, hospital settings and urban areas. Results also showed that clinicians in rural areas and clinics, including nurses, were especially interested in training. CONCLUSIONS: Clinician competency in LARC provision is important in southern Africa, given the low use of methods and high rates of unintended pregnancy among HIV-positive and at-risk women. Despite low provision, clinician interest is high, suggesting the need for increased evidence-based training in LARC to reduce unintended pregnancy and associated morbidities.

Discrepancies in diagnosis of incident HIV infection between antibody- and DNA-based tests in a phase III prevention trial in southern Africa.

Dried blood spots (DBS) are widely used to test for HIV in a variety of research and service delivery settings; however, uniform guidelines regarding collection, storage and DNA extraction processes have neither been developed nor evaluated. Previously published reports suggested DBS may be stored at room temperature for up to 60 days, and intensive stability tests have shown that DBS can withstand high temperatures, humidity and freeze-thawing. During the implementation of a large randomized controlled trial (RCT) in southern Africa, with HIV acquisition as the primary endpoint, we observed 65 instances when DBS samples collected from the same day as a positive HIV antibody test yielded negative DNA polymerase chain reaction (PCR) results. The source of this discrepancy may have been due to inadequate specimen volume, filter paper or DNA extraction procedures, but were most likely due to storage conditions that have been reported as acceptable in other settings.

The relative contribution of viral and bacterial sexually transmitted infections on HIV acquisition in southern African women in the Methods for Improving Reproductive Health in Africa study.

We assess the relative contribution of viral and bacterial sexually transmitted infections (STIs) on HIV acquisition among southern African women in a nested case-control study within the Methods for Improving Reproductive Health in Africa (MIRA) trial. Cases were women with incident HIV infection; controls were HIV-uninfected at the time of case seroconversion selected in a 1 to 3 case to control ratio (risk-set sampling), matched on study site and time of follow-up. Conditional logistic regression models were used to calculate adjusted odds ratios (AORs) and population-attributable fractions (PAF). Among 4948 enrolled women, we analysed 309 cases and 927 controls. The overall HIV incidence rate was 4.0 per 100 women-years. The incidence of HIV infection was markedly higher in women who had prevalent Herpes simplex virus type 2 (HSV-2) (AOR: 2.14; 95% confidence interval [CI]: 1.55-2.96), incident HSV-2 (AOR: 4.43; 95% CI: 1.77-11.05) and incident Neisseria gonorrhoeae (AOR: 6.92; 95% CI: 3.01-15.90). The adjusted PAF of HIV incidence for prevalent HSV-2 was 29.0% (95% CI: 16.8-39.3), for incident HSV-2 2.1% (95% CI: 0.6-3.6) and for incident N. gonorrhoeae 4.1% (95% CI: 2.5-5.8). Women's greatest risk factors for HIV acquisition were incident bacterial and viral STIs. Women-centred interventions aimed at decreasing HIV incidence in young African women need to address these common co-morbid conditions.

Acceptability of COL-1492, a vaginal gel, among sex workers in one Asian and three African cities.

OBJECTIVES: To evaluate the acceptability of COL-1492, a vaginal gel containing 52.5 mg nonoxynol-9, in an HIV prevention trial. METHODS: Sex workers participating in a phase II/III triple blind, randomised trial in Benin, Côte d'Ivoire, South Africa, and Thailand were interviewed on the gel's acceptability at monthly scheduled clinic visits. Safer sex counselling, male condoms, and study gels were given at each monthly visit; a gynaecological examination and HIV test were performed. Phase III interviews considered the participants' appreciation of the gel. On the first, second, and fifth follow up visits, the study volunteers completed more extensive questionnaires. RESULTS: Responses were similar between treatment arms. Women indicated not liking their gel in 1.8% of the visits; 98.1% of the women found the gel easy to apply; 30.1% said that it affected sexual intercourse. These effects were mostly improvements (92.6%) by facilitating intercourse (73.6%). Intercourse was more often affected in women reporting painful sexual intercourse (OR: 2.59 (95% CI 1.63 to 4.12)) and in older women. The latter effect differed among centres. CONCLUSION: Most participants found their assigned gel acceptable and the vast majority of reported effects on intercourse were favourable. The type of gel had no significant impact on the findings.

Nonoxynol-9 for preventing vaginal acquisition of HIV infection by women from men.

BACKGROUND: There is a need for female-controlled methods of HIV prevention. Vaginal microbicides, substances inserted into the vagina to prevent women acquiring HIV and sexually transmitted infections (STIs) from men, could be useful in this regard. One potential vaginal microbicide is the widely used spermicide, nonoxynol-9 (N-9). OBJECTIVES: To determine the safety and effectiveness of N-9 in preventing vaginal acquisition of HIV infection by women from men. SEARCH STRATEGY: Extensive searches of electronic databases, conference abstracts, reference lists of relevant studies and contact with experts and funders. SELECTION CRITERIA: Randomised controlled trials meeting pre-determined quality criteria with HIV infection as the outcome. DATA COLLECTION AND ANALYSIS: Data were extracted by one reviewer and checked by the another. Any discrepancies were adjudicated by a third reviewer. MAIN RESULTS: Five trials were included in the review and four contributed to a meta-analysis. Overall, the risk of HIV infection was not statistically significantly different among women receiving N-9 (relative risk [RR] 1.12, 95% CI 0.88-1.42; p=0.4). The risk of genital lesions was statistically significantly greater among women receiving N-9 (RR 1.18, 95%CI 1.02-1.36; p=0.02). REVIEWER'S CONCLUSIONS: There is no evidence that nonoxynol-9 protects against vaginal acquisition of HIV infection by women from men. There is evidence that it may do harm by increasing the frequency of genital lesions.

Nonoxynol-9 for preventing vaginal acquisition of sexually transmitted infections by women from men.

BACKGROUND: The incidence and prevalence of sexually transmitted infections (STI) and other reproductive tract infections (RTI) is high in much of the developing and parts of the developed worlds. STIs and RTIs are associated with the vaginal transmission of HIV. Additional strategies to improve STI control are needed, and vaginal microbicides are a possible strategy. One potential vaginal microbicide is the widely used spermicide, nonoxynol-9 (N-9). OBJECTIVES: To determine the safety and effectiveness of N-9 in preventing vaginal acquisition of sexual transmitted infections (exclusive of HIV) by women from men. SEARCH STRATEGY: Systematic search of electronic databases, conference abstracts, reference lists of relevant studies and contact with experts and funders. SELECTION CRITERIA: Randomised controlled trials meeting pre-determined quality criteria with STI as the outcome. DATA COLLECTION AND ANALYSIS: Data were extracted by one reviewer and checked by another. MAIN RESULTS: Ten of 12 identified randomised controlled trials were included and findings among them were broadly consistent. In meta-analysis, the risks of gonorrhoea (relative risk [RR] 0.91, 95%CI 0.67-1.24), cervical infection (RR 1.01, 0.84-1.22), trichomoniasis (RR 0.84, 0.69-1.02), bacterial vaginosis (0.88, 0.74-1.04), chlamydia (RR 0.88, 0.77-1.01) and candidiasis (RR 0.97, 0.84-1.12) were not statistically significantly different in women receiving N-9 compared with placebo. Genital lesions were more common in the N-9 users (RR 1.17, 95%CI 1.02-1.35). REVIEWER'S CONCLUSIONS: There is good evidence that nonoxynol-9 does not protect against sexually transmitted infections, and there is some evidence that it may be harmful by increasing the rate of genital ulceration. As such, this product cannot be recommended for STI prevention.

Characterization of proteinuria in asymptomatic family members and household contacts of children with hepatitis B virus-associated membranous nephropathy.

The biosocial background in which the hepatitis B virus (HBV) carrier state with membranous nephropathy (MN) develops was studied by evaluating HBV carriage and proteinuria among 195 family members and household contacts of 31 index HBV carrier children with MN. Unrelated individuals from the communities of these index cases who were negative for HBV served as controls (n = 123). HBV was determined by using third-generation enzyme-linked immunosorbent assay, slot-blot hybridization, and nested polymerase chain reaction. Patterns of proteinuria were determined by using sodium dodecyl sulfate-polyacrylamide gel electrophoresis; immunoglobulin G and haptoglobulin were suggestive of MN. Seventy-two members (36.9%) of the study group (n = 195) were HBV carriers; 21 of these carriers (29.2%) had proteinuria. Twenty-eight members (41.2%) of the study group who were HBV negative (n = 68) and 26.8% of the controls showed proteinuria. This lack of association between HBV carriage and proteinuria remained when controlled for sex and family relationship. HBV was not protective against the development of proteinuria. Proteinuria suggestive of MN was strongly associated with an abnormal protein-creatinine ratio (P: = 0.001), but was not significantly different between subjects and controls (8.7% versus 6.5%; P: = 0.5). Genetic influences or environmental exposures in these subjects may be responsible for the proteinuria, suggesting underlying glomerular basement membrane damage. Discordance between the HBV carrier state and patterns of proteinuria in the study group suggest that HBV and MN may not be causally related or may reflect exceptional interaction between specifically vulnerable individuals and HBV.

Conserved domains of subtype C nef from South African HIV type 1-infected individuals include cytotoxic T lymphocyte epitope-rich regions.

We have characterized 43 nef sequences from subtype C HIV-1-infected South Africans and compared deduced amino acid sequences with other subtypes to identify areas of conservation. Our Nef amino acid sequences were aligned with a consensus subtype B, HXB2 reference strain and a consensus subtype C sequence. All were found to be highly homologous to subtype B in the central region of Nef, but more variable at the N and C termini of the molecule. Alignment of a consensus amino acid sequence generated from South African subtype C Nef with subtypes A, B, and D underscores cross-clade conservation in the central domain of the molecule. This domain is also rich in previously described cytotoxic T lymphocyte (CTL) epitopes that are restricted by commonly found HLA molecules in the South African population.

The impact of human immunodeficiency virus type 1 status on human papillomavirus (HPV) prevalence and HPV antibodies in serum and cervical secretions.

Human immunodeficiency virus (HIV) type 1-infected (HIV-positive) and -uninfected (HIV-negative) sex workers were examined for the presence of cervical human papillomavirus (HPV) DNA. Cervicovaginal rinse and serum samples from these women were examined for IgG and IgA antibodies to HPV-16 virus-like particles (VLP-16) by ELISA. The HIV-positive women displayed a significantly higher prevalence of HPV DNA (40/47 [85%]) than did the HIV-negative women (22/52 [42%]; P=.00001). Both HIV-positive and HIV-negative sex workers displayed a high seroprevalence rate for anti-VLP-16 IgG antibodies (27/40 [68%] and 30/43 [70%], respectively), but significantly fewer HIV-positive women than HIV-negative women had anti-VLP-16 serum IgA (6/40 [15%] vs. 17/43 [40%], respectively; P=.012). Significantly more HIV-positive women than HIV-negative women had cervical anti-VLP-16 IgG antibodies (16/49 [33%] vs. 6/63 [10%], respectively; P=.002) but not IgA antibodies (P=.3).

Safety of multiple daily applications of COL-1492, a nonoxynol-9 vaginal gel, among female sex workers. COL-1492 Phase II Study Group.

RATIONALE: COL-1492 is a nonoxynol-9 (N-9)-containing vaginal gel and may be a potential microbicide. As part of an effectiveness trial, an initial toxicity study was conducted. OBJECTIVES: The main objective of the reported study was the assessment of the toxicity of a 52.5 mg N-9 gel, COL-1492, when used a number of times each day by female sex workers. METHODS: This was a randomized, placebo-controlled triple-blinded trial among female sex workers. The participants were asked to use the product for each vaginal sexual act. At each monthly visit a gynaecological examination with sexually transmitted disease sampling and colposcopy was performed. Venous blood was drawn for syphilis and HIV serology. All women received intensive counselling on condom use. Male condoms and sexually transmitted disease treatment were given free of charge. RESULTS: Only blinded results on the colposcopic examinations are reported. The incidence of lesions with or without an epithelial disruption was low: 0.06 and 0.29, respectively, per 100 woman-days in group A; 0.09 and 0.26 respectively per 100 woman-days in group B. There was no significant difference between the two arms. CONCLUSION: The multiple daily use of COL-1492 by female sex workers did not show an increase of local toxicity over that of a placebo. Colposcopy was discontinued in the autumn of 1997 in accordance with a Data Safety Monitoring Board decision. In the currently ongoing effectiveness trial the assessment of the product's toxicity continues to be monitored by simple visual examination.

Challenges in the conduct of vaginal microbicide effectiveness trials in the developing world.

Conducting a phase III trial of a vaginal microbicide in a developing country poses several important and complex ethical challenges. As part of a process to bridge the gap between ethical theory and practice, we share our experiences in performing a phase III trial of Col 1492 (Advantage S) among female sex workers at four sites world-wide; Durban, Abidjan, Cotonou and Hat Yai. The ethical challenges included: (i) difficulties in obtaining informed consent. Participants were unable to grasp the concepts of a clinical trial for several weeks to months. In Cotonou, 30% of the women did not know the gel was tested for HIV prevention. Only 25% understood what a placebo was. In Durban, 70% of the women did not fully understand the study after 3 months; (ii) in sustaining the use of known HIV prevention strategies. Participants at the Durban site had difficulty in sustaining condom use due to financial and client preferences. Sex without condoms was worth more ($20) than sex with condoms ($10); (iii) in maintaining the confidentiality of the subject's HIV status. Novel approaches such as role plays and emphasis on other exclusion criteria were needed to maintain the confidentiality of women not included in the trial due to their HIV status; (iv) in providing care and support to the subjects who became infected with HIV during the trial. Women could only be offered routine sexually transmitted disease treatment and counselling. Anti-retrovirals were not offered. The successes and failures of the solutions attempted are described.