The initial study idea was created in response to an April 2001 Request for Proposals from the US Centers for Disease Control and Prevention, submitted in June 2001 and funded in October 2001. As the protocol was being developed, we conducted a series of formative research projects over a two-year period in Lilongwe. Using data generated from multiple focus groups, client home visits, and individual in-depth interviews and taste tests of the nutritional supplement among pregnant women, spouses, civic leaders, healthcare workers, HIV-infected breastfeeding women and family members, the team gauged the feasibility of the proposed protocol (49
). This shaped many decisions in protocol design; for example, we learned that mothers considered drawing blood from healthy babies at birth feasible and acceptable but had concerns about the volume to be drawn, and we used this information when formulating the biological sample schedule and volumes. We also learned that mothers preferred a longer weaning period, rather than more rapid weaning, for their own and the babies’ comfort and safety and to assure good infant transition to other foods. This shaped our decision to allow a month-long weaning period at 6 months of infant age. Mothers expressed concern about the impact of weaning on the nutritional status of their infants and so we provide locally produced RUTF from weaning through 48 weeks post-partum. We also tested the acceptability among mothers of the choice of the maternal nutrition supplement. We placed emphasis on mother’s health during breastfeeding, and provided a family food supplement. Without the formative research many of these issues may have gone unaddressed and could have undermined the successful implementation of the BAN study. The result was an evidence-based protocol ensuring participant understanding of the research, safeguards for the participants, and increased feasibility and acceptance of clinical research in this community.
While English is the official language of Malawi, Chichewa is the common local language widely used throughout the country. Though many of our patients do not speak English, all study staff spoke both English and Chichewa. The adult literacy rate in Malawi is approximately 64% (51
). Prior to initial enrollment, the language of our consent forms was dramatically simplified compared to the standard language found in many standard consent forms. Due to language and literacy issues, we felt that a long and complicated consent form was not ethical. For example, we did not mention chemical names of compounds but rather emphasized the potential toxicities they caused. All consents were translated into Chichewa and then back translated by a separate team into English to ensure that they met international standards. All study forms, which contained questions to be asked patients, were in Chichewa.
The study team used the WHO CIOMS Guidelines for Biomedical Research Involving Human Subjects when the study was designed and held an ethics consultative meeting with an outside ethicist prior to submitting the final protocol.
Once the final protocol was developed, it was submitted for review at three separate institutional review boards in 2003: the parent institution, the sponsoring agency, and the host country. Since each IRB had its own comments on the protocol, we reconciled all with final approval by all three IRBs in March 2004, thirty months after the initial award. In parallel with this process, negotiations for donations of study drugs were conducted.
BAN began enrolling participants in April 2004. Initially, enrollment was relatively slow. In order to increase enrollment, we engaged in community outreach to dispel rumors about the study, including discussions with tribal elders and leaders. We noticed that many women who initially expressed interest in the study failed to return for their screening and enrollment visits. Upon further questioning, it became clear that many husbands were not supportive of the study. To address this, we developed community activities specifically targeting men, such as weekly support groups for men. In addition to community activities, two recruitment sites were added in other areas of Lilongwe. Although the main study activities, including all clinical visits, remained at one site, these two additional sites provided a larger pool of potential study participants. By February 2006, enrollment had increased to our target of 15 enrollees per week. It has remained relatively stable since.
Prior to initiating the study we realized that unlike shorter term studies, we would be following these women and their infants for more than 12 months. Usually, such care is provided by the understaffed public sector, which also suffers from frequent stock outs of critical medicines and often lack diagnostic facilities. Since we felt an ethical obligation to provide clinical care for intercurrrent illnesses, we purchased all the medications for intercurrent illnesses, hired additional staff to provide for non-study related care and built a microbiology laboratory. Although we have not completed a formal cost analysis such clinical care consumes an additional 25% of the study costs.
To date, there have been twelve subsequent amendments made to the protocol and submitted to the three regulatory agencies (). Modifications were made to increase toxicity monitoring in light of new data on nevirapine hepatotoxicity (52
), ours and others’ discovery of benign ethnic infant neutropenia (53
) which is a temporary, clinically insignificant laboratory finding, better understanding of nevirapine resistance (55
), FDA advisories (56
), changing national and international guidelines for treatment of adults and need for cotrimoxazole prophylaxis (57
), a change in the NIH Toxicity Tables, and several changes to decrease the protocol burden on the clients and the staff. We also decided that longer follow up may be necessary and thus altered the consent form to allow for contact after the study ended. These amendments can be grouped as those implemented to address toxicities, those to decrease the enormous workload of the study, and those to meet the changing reality of HIV care. After the initial IRB application, which took many months of negotiating between the three IRBs, the process has decreased to 2 months or less for each amendment. Although, each of the IRBs would sometimes ask for clarification or ask us to amend the consent form, there were no excessive delays or unwanted requests. On two occasions, for safety and ethical reasons the IRB’s were notified simultaneously of the problem and our action plan. This applied to our handling of the nevirapine toxicity issue and after the DSMB recommendation to stop one arm.
Amendments for new information regarding toxicities
Within weeks of assigning women to HAART containing NVP, we became aware of a need to increase the monitoring for hepatic toxicity. In February 2004, Boehringer-Ingelheim had sent a “Dear Doctor” letter outlining changes to the Black Box Warning for nevirapine, which stated that women with CD4 counts greater than 250 cells/μL had a 12-fold higher risk of hepatotoxicity, usually within six weeks of initiating therapy. On January 19, 2005 the FDA strengthened the warning to recommend against starting nevirapine treatment in women with CD4 counts greater than 250 cells/μL unless benefits clearly outweigh risk (56
). Among the first 39 women randomized who received nevirapine-containing HAART, we found 3 severe rashes, 1 case of clinical hepatitis and 5 cases of severe elevations in hepatic transaminases. We subsequently have shown that nevirapine drug levels are higher in Africans compared to Americans (59
). Even though we did not have any fatal complications, conducting a study in a resource-limited country requires a higher standard beyond safety and efficacy; that of feasibility of eventual national roll out of the regimen. The requirement for close monitoring of transaminases necessitating a large number of clinical and laboratory staff clearly made this an infeasible regimen for eventual implementation in breastfeeding women with high CD4 counts in Malawi. We initially switched all breastfeeding mothers assigned to HAART from nevirapine to our second-line regimen containing nelfinavir and ultimately, for reasons of availability, safety and potency, to lopinavir/ritonavir (Kaletra).
The NIH toxicity tables were developed by the NIH Division of AIDS (DAIDS) in the beginning of the AIDS epidemic in the United States. Although widely used in clinical trials, we found early in the conduct of BAN that they do not take into account the benign neonatal neutropenia seen in both African-Americans and black Africans (53
). In our first months we noticed that of 206 newborns enrolled, 22 had grade 3 or 4 neutropenia (9 present at birth) without any clinical sequelae and with spontaneous recovery of the absolute neutrophil counts within weeks. We, therefore, modified the infant neutropenia criteria for our protocol to define grade three toxicity for any age infant to include an ANC less than 400 cells/μL, which decreased the incidence of grade 3/4 neutropenia to less than 1%. Subsequently, DAIDS did modify their toxicity table, although not to the extent of our revision ().
NIH Division of AIDS previous and revised toxicity tables for definition of infant neutropenia and BAN modification
In 2006 the NIH Division of AIDS modified infant anemia criteria to state that any hemoglobin less than 9.0 g/dl for HIV negative infants at least 57 days old was at least a grade 3 toxicity. The previous version cited a hemoglobin value of less than 7 g/dl as the cut-point for grade 3 anemia. This new table was adopted by BAN in May 2006 and led to an artificial increase in grade 3/4 anemia from 55 newborn infants to 238 in the first 1233 mother infant pairs randomized.
Amendments to decrease workload
With a weekly accrual target of 15 mother-infant pairs, the overall workload of BAN is considerable. Due to the complex design and multiple objectives, lengthy questionnaires, and the need for close participant monitoring, study visits were taking up to 8 hours to complete, sometimes requiring drivers to take clients home in the evening, in some cases requiring return visits, and placing an undue burden on participants. In addition, the sheer volume of data generated, including over 20,000 pages scanned per month, jeopardized the quality of the data obtained. With the onset of the rainy season each November in Lilongwe, sick (i.e. unscheduled) visits to the study clinic increased precipitously, placing a huge burden on project staff and increasing study costs dramatically. From July 2005 until June 2006, there were 1800 scheduled study visits and 400 unscheduled visits for the first 883 mother infant pairs randomized, usually due to intercurrent illnesses. These additional visits increased annual costs by $362,000 (29%) in additional personnel and laboratory costs.
With this burden on clients, staff and study budget, we quickly realized that several elements of the original protocol were not feasible. Since many of our patients are not literate, our staff would read the consent to them, pausing for questions and clarifications. In addition, the study staff uses a consenting checklist to ensure that important areas have been covered and are understood by the client. The study consent process was taking up to 2 hours to complete. Since 28% of consented women were subsequently determined to be ineligible, (half because they did not meet primary eligibility due to CD4 count or hemoglobin level and half because they delivered at home and did not present in time to the study clinic or had other complications), we decided to split the consent process into a screening registration form and a post-partum randomization form. This both shortened the original consent and the time imposed on the patients as well as sparing those, who would not be eligible, from having to sit through the description of the study after delivery. In addition, the PMTCT program, which refers potential participants to BAN, began routine CD4 testing thereby allowing women with low CD4 counts to be excluded from the study and referred to appropriate care without referral to BAN.
The burden on study participants was also great due to the number and length of visits. In order to complete the study, mothers came to clinic for at least 17 visits and infants for at least 15 visits with blood drawn at every visit. When 900 women-infant pairs had been randomized, we systematically reduced the data collected to focus on key study endpoints. Case report forms across the study had questions reduced. Reductions were made in demographic, delivery history, breastfeeding, infant feeding, infant and maternal anthropometry, ARV adherence, lab and physical exam forms. Twenty of the over 70 forms in the study were modified. After we had randomized 1600 mother infant pairs, we further reduced the case report forms. This reduction was mainly through the elimination of forms or number of times a form, such as the infant and maternal 24-hour dietary recall, is administered. Deleted forms included the breast health history, detailed breast exam, and detailed information about breastfeeding and early weaning practices. At the peak, we estimate that collectively, staff was spending 1941 hours per month filling out forms, which was reduced to an average of 1608 hours after the first reduction in 2006 and to 1135 hours after the second reduction in 2007, resulting in an overall reduction in workload of approximately 40%. Although we had pilot tested many of the forms, in retrospect we realized that in advance of the study it would have been helpful to devote more time to practice runs of the visit procedures including completion of the forms.
We originally planned to use case report forms with diagnostic algorithms for each opportunistic and endemic infection similar to those used in US HIV clinical trials. Even with modifications reflecting locally available tests, it was clear that these forms were infeasible. For instance, almost no one obtained a chest radiograph for initial diagnosis of pneumonia. We stopped using those forms and substituted an adverse events log to make it easier for staff to track onset and completion of various events including adverse events, endemic and opportunistic infections and concomitant medications.
In addition to reducing data collection, we also cut back on specimen collection. Thirty-nine months into the study we stopped freezing cell pellets, peripheral blood mononuclear cells, and pharmacokinetic samples in order to cut costs. Previously, we were freezing on average 5,680 breast milk vials, 3,100 plasma vials, 500 cell pellet vials and 1,400 peripheral blood mononuclear cell vials monthly. After implementing our amendments this decreased to 1,350 plasma vials monthly. None of the primary endpoints required any of the discontinued questions or specimens.
Amendments to meet the changing realities of HIV care
In a study initiated during 2004 in a rapidly changing field, it was unavoidable that new data would appear that might require changes to the protocol. In an effort to minimize development of nevirapine resistance in women randomized to a HAART regimen containing nevirapine, we added a 7-day Combivir tail after the nevirapine was stopped. This was an approximate adjustment for the prolonged serum half-life of nevirapine compared to the relatively short half-life of Combivir.
The Malawi Ministry of Health guidelines for HAART treatment changed the CD4 threshold below which antiretroviral therapy was to be started from a CD4 count of <200 cell/ μL to < 250 cell/μL. In response, we changed the BAN enrollment criteria to exclude women who had CDC counts of <250 cell/ μL; these women were referred for antiretroviral therapy. The Malawi Ministry of Health also instituted routine cotrimoxazole prophylaxis for HIV-infected adults and HIV-exposed infants. As of December 2005, all women in BAN with CD4 counts <500 cell/μL were started on cotrimoxazole after the first trimester. In addition, all children aged 6 weeks or above received cotrimoxazole until 3 months after discontinuation of breastfeeding. For HIV-infected infants the prophylaxis is continued indefinitely.