Our search for “RCTs” in the Register identified 834 records. Of these, 154 references described potentially relevant African RCTs. The remaining 680 references referred to RCTs conducted in non-African countries. With the inclusion of two additional cross references, the number of potential African RCTs increased to 156. We were unable to trace the reference for one study 
, and could not obtain full reports for two 
, thus these three studies were excluded. After assessing eligibility, 97 references were included. shows the flow diagram including the reasons for exclusion of 59 references.
Flow diagram showing the assessment of African trials for 2004–2008 in the HIV register.
Of the 97 references included, 68 were primary RCTs reported for the first time during 2004–2008. Eight of the 68 primary RCTs had nine more references published during the same period. Twelve references reported on four RCTs already published before 2004 and included in the previous analysis 
. We identified eight references to published protocols, three of which had primary RCTs published in the period under study. We extracted data on the 68 primary RCTs. Where an RCT was reported in more than one reference, we used data contained in all the related references.
Sources of Trials
All trials included in the analysis were published as journal articles. Nine were published in The Lancet
, seven in AIDS
and six in the New England Journal of Medicine
. Other journals included the American Journal of Clinical Nutrition
and Journal of Infectious Diseases
with four articles each; the American Journal of Obstetrics and Gynecology, BMJ, PLoS ONE
and the Journal of Acquired Immune Deficiency Syndromes
had three articles each. The remaining 26 trials were published in 20 different journals. The 2008 impact factor for the journals with one published article ranged between 1.517 and 7.069 (seven journals). Five journals with two published articles each had an impact factor range between 2.304 and 31.718. The remaining eight journals were not listed in the Institute for Scientific Information Web of Knowledge Journal Citation Report for 2008. All study reports were published in English language journals. Sources of trials are shown in Appendix S3
Location and centre types of African RCTs
Seven RCTs were conducted at multiple sites across 11 countries which included Benin, Cameroon, Ethiopia, Ghana, Malawi, Nigeria, South Africa Tanzania, Uganda Zimbabwe and Zambia. Thirty-six RCTs were conducted in multiple centres in a single country. Of these, South Africa hosted 12 trials, Tanzania and Zambia hosted five trials each, and Malawi hosted four trials, Kenya, Nigeria and Zimbabwe hosted two trials each. Botswana, Burkina-Faso, Rwanda and Uganda each hosted one multi-site trial each.
Twenty-five trials were single-centre RCTs with five conducted in South Africa and five in Zambia. Four trials single-centre trials were conducted in Zimbabwe, three were conducted in Kenya and Uganda each and Malawi hosted two trials. Mozambique, Nigeria and Tanzania hosted one trial each. shows the map of African countries and the number of trials conducted per country.
The countries where the trials were conducted.
The principal investigator (PI) was clearly reported in 21 of the 68 trials. In 17 of the trials the PI was also the corresponding first author and in one trial the corresponding last author. The PI was the first author in one and the last author in two of the remaining trials but not the corresponding author.
When the PI was not clearly stated, we classified the first author as the PI for further analysis. Most principal investigators, both the clearly stated and the assumed, were based in the USA (29) and in the United Kingdom (10), with 18 PIs residing in Africa: in South Africa (8), Zambia (5) Uganda (2), Kenya (1), Rwanda (1), and in Nigeria (1). The other non-African PIs resided in Denmark (3), France (1), Netherlands (1), Canada (1) and India (1). The locations of four PIs could not be identified.
Qualifications of the PIs were generally not reported. Of the 18 reported, seven had PhD degrees, six medical degrees, three had both medical and PhD degrees and two had medical and Masters degrees.
Forty-three RCTs assessed interventions for the prevention of HIV and related infections, while 25 assessed interventions for the treatment of AIDS and related infections.
Of the 43 prevention trials, 15 (35%) investigated the prevention of mother-to-child transmission of HIV. Behavioural interventions and microbicides were investigated in five trials each. Six trials investigated the use of pharmaceutical products for the prevention of opportunistic infections in HIV-infected participants, while one trial evaluated pharmaceutical products for HIV. Other trials investigated the effects of male circumcision (3) and nutritional interventions in HIV-infected participants (4). Four other trials investigated interventions to improve the uptake of HIV counselling and testing, the use of contraception methods, planning for the future by HIV-infected couples and the prevention of malaria in HIV-infected patients. We did not identify any trials for vaccine efficacy.
Twelve (48%) of the 25 treatment trials focused on pharmaceutical products for the treatment of opportunistic infections and three investigated pharmaceutical products for the treatment of AIDS. Others investigated the use of pharmaceutical products for other infections in HIV-infected people (3), the effects of nutritional interventions (4), the effects of exercise (1), delivery of highly-active antiretroviral therapy (HAART) by direct observation (1) and interventions for treating bacterial vaginosis (1).
Twenty-one trials reported both the month and year participant enrolment into the trial began and ended, and the month and year the trial was completed. Nineteen reported the month and year when participant enrolment began and ended, but did not report the trial completion dates. Twelve reported the month and year the trial started and ended, but did not report the dates of participant enrolment. Ten RCTs reported some dates, e.g. trial start and end years with no months or the month and year the trial started without reporting the end dates. In six of the trials no dates were reported.
The first trial commenced in 1994 and investigated the treatment of AIDS-associated Kaposi's sarcoma in Zimbabwe 
. Nine trials commenced before the year 2000. The last five trials began in 2005 and two of these ended in 2007.
Sample size, power calculation and primary outcome
The median sample size was 626 participants. The sample size ranged from 33 participants in a trial investigating the treatment of Kaposi's sarcoma conducted in Durban, South Africa 
to 9645 participants in a behavioural intervention trial conducted in the Mwanza region in Tanzania 
. In 31 RCTs, the number of participants was less than or equal to 500, with four RCTs including more than 5000 participants (see ).
The ranges of the sample size in the trials.
In 54 trials, the sample size and power calculation based on the primary outcome was conducted before the trial began. In fourteen trials there was no report of power calculation. The primary outcome was clearly indicated in 56 trials, while 12 trials did not report which of the outcomes was the primary outcome.
Number of intervention groups and types of randomization
Fifty-seven trials compared two interventions, six compared three and five compared four interventions with each other. In 43 trials participants were randomized at the individual level, while 17 trials randomized mother and child pairs. One trial randomized couples. There were six cluster trials and one cross-over trial. Three cluster trials investigated behavioural interventions, two investigated the prevention of other infections in HIV-infected participants and one investigated the use of pharmaceutical products for the prevention of opportunistic infections in HIV- infected participants.
The quality of methods
Generation of the randomization sequence
We judged the methods used to generate the random sequence to be free of bias in 38 trials. The methods used to generate the randomization sequence included a computer-generated list (31), a randomized list generated by a statistician (2), a random number list/), and a randomized list prepared off-site (1). In one trial, the methods were only described as permuted without any further description 
and coin tossing was used in one trial 
. In 30 trials we could not determine the methods used to generate the random sequence.
Methods used for allocation concealment was assessed to be free of bias in 35 trials. These included the use of sequentially numbered, sealed, opaque or padded envelopes in 14 trials, treatment given in identical, non-transparent and sequentially numbered containers (11), the randomization list kept in sealed envelopes drawn independently and sequentially (5), the code or the list held off-site with no access by the investigators (4) and centralized randomization (1). Methods for allocation concealment were not adequately described in 32 trials. The allocation sequence was inadequately concealed in one trial, where the first 17 participants were assigned to one intervention and randomization started on the 18th participant.
Blinding of providers, participants and outcome assessors
Eighteen trials clearly stated that providers, participants and outcome assessors were blinded to the interventions into which the participants were allocated. In six trials, providers and participants were blinded, but it was unclear whether the outcome assessors were blinded. Twenty-four trials blinded the providers, 33 blinded the participants and 31 blinded the outcome assessors. Providers were not blinded in 30 trials and participants were not blinded in 29 trials, however, outcome assessors were blinded in 12 of these trials. shows the blinding of providers, participants and outcome assessors.
Blinding of providers, participants and outcome assessors.
Consort flow diagram
A CONSORT flow diagram was included in 58 trials, clearly showing loss to follow-up. In 10 trials, the CONSORT flow diagram was not included.
Thirteen trials were terminated earlier than planned. Reasons for early termination included a significant effect of the intervention detected during data monitoring 
and no significant effect during data monitoring 
. In two of these trials, the results of data monitoring were confirmed in a recently published trial employing the same intervention 
. A possible increase in the risk in the intervention arm was cited in two trials 
, in both, the decision to review the data was taken due to the publication of a related trial. In one trial 
, the US Food and Drug Administration recommended against the prolonged use of Nevirapine by women with CD4 counts ≥250 cells/mm3
, thus the intervention was terminated. Non-compliance with the protocol 
and low event rate of the outcome of interest 
were cited as reasons in one trial each.
Ethics approval and informed consent
Sixty-six trials reported receiving ethical approval and two trials did not report on ethical approval. Fifty-seven received approval from both African and non-African ethics review committees, eight reported receiving ethical review from African countries without mentioning international ethics review-boards, while one trial reported receiving ethical approval but did not state the name and location of the approving body. Of the eight trials that reported ethical approval only from African bodies, seven reported non-African funding agencies. Written informed consent was obtained from participants in 52 trials, and one trial obtained consent orally. Eight trials reported receiving consent, but the type of consent was not specified. Informed consent was not reported in seven trials.
Fifty-eight trials were funded by multiple organisations and the primary funder could not be identified. Nine trials had a single funder. One trial 
did not report on funding.
Eight RCTs were funded by African government agencies; five agencies were from South Africa (National Institute for Communicable Diseases; National Research Foundation; Medical Research Council of South Africa; University of Cape Town and Rhodes University), two from Zimbabwe (Ministry of Health and Child Welfare, and University of Zimbabwe) and one from Rwanda (Multi-sectoral AIDS Program). None of the trials was solely funded by African government agencies. The majority of funding was obtained from non-African governments. shows the details of non-African government agencies funding the trials.
Non-African Government agencies providing funding to trials.
Five African non-governmental agencies funded four trials; from South Africa (Africa Centre for Health and Population Studies), Malawi (Blantyre Christian Centre, and Wellcome Trust Laboratories in Malawi) and Zimbabwe (Jewish Humanitarian & Relief Committee and The Salvation Army). Non-African non-governmental agencies included Fogarty International (8), Wellcome Trust (8), and the Bill and Melinda Gates Foundation (4). One trial 
was exclusively funded by a non-governmental agency (Rockefeller Foundation). USA based non-governmental agencies provided funding to 33 trials, United Kingdom based agencies funded eight, while Denmark and Canada based agencies funded five and one, respectively. The locations of 18 non-governmental agencies were unclear.
Pharmaceutical and other commercial companies
None of the trials were exclusively funded by pharmaceutical companies. Fourteen pharmaceutical companies provided funding to 16 trials. Funding from pharmaceutical companies included the provision of drugs, placebo or both. Nine of the trials that received funding from pharmaceutical companies were co-funded by the US government. One trial 
was exclusively funded by a commercial company, Nestle SA.