Boston/Botswana circ. trial update
How did three babies die?
Earlier this year a randomised trial by US researchers comparing different ways to circumcise newborn African boys appeared to alter both its sample size and the number of treatment arms, between registration and publication (click here). Unmentioned in the abstract, three healthy babies also died (click here). My letter to JAIDS and the author’s response (click here, uploaded here Thornton and Plank) have now been published.
Dr Plank says the registration/publication discrepancies resulted from muddle and ambiguity, and I agree. Far from attempting any sort of data driven analysis, she appears to have little interest in the randomised comparison. The trial seems primarily part of a campaign to introduce newborn circumcision to Botswana; it probably reduces sexual transmission of HIV in adults, so let’s get kids done at birth!
She also reveals that one baby died 25 days, and another 10 weeks, post surgery. Neither death appears to have been procedure-related, and again I agree. However, here is Dr Plank on the third death:
“The last death occurred in a baby who died of suspected sepsis on day of life 3. The baby was circumcised using a Mogen clamp on day of life 2 and discharged to home later that day. The following day he was brought to the local health center with respiratory distress and was noted to be febrile and was transferred to the district hospital. The study team was not notified of his admission until the next morning, after he had died. The circumstances of this baby’s death were reviewed in great detail with several groups to obtain independent assessment of the cause of death: the hospital staff, the Botswana Ministry of Health, the Botswana Health Research and Development Committee, the Partners Institutional Review Board (IRB), and our own Data Safety Monitoring Committee (DSMC). All parties agreed that based on all the clinical data available, the most likely cause of death was neonatal sepsis or pneumonia and that it was extremely unlikely that the baby’s death was related to the circumcision procedure.
Autopsies are very rarely performed in Botswana and were not performed in any of the 3 deaths in the study. Detailed diagnostic work-ups are also often not available in resource-limited settings or are not performed (eg, because a baby dies at home). Finally, prenatal screening for group B streptococcus is not routinely performed, and mothers do not receive prophylactic antibiotics.”
A healthy term baby dies 24 hours after a research operation and no tests nor autopsy are done. However the researchers, their own DSMC, and the two IRB’s who had approved the research all conclude “that it was extremely unlikely that the baby’s death was related to the circumcision procedure”!
Am I going mad? “Extremely unlikely”! How can any sane doctor possibly conclude that?
Of course no-one can prove the death was procedure-related. But the time course strongly suggests a relation. Even if the baby had some unrecognised congenital heart disease, or streptococcal infection, as the direct cause of death, it would remain highly likely that the research circumcision was an aggravating factor.
My enquires of the Partners IRB in Boston led to another internal investigation but the same conclusion, (Partners reply) and a refusal to reveal their reasoning; “we don’t publish our deliberations.”
Together with some colleagues I formally requested the United States Office of Human Research Protection to investigate the decisions of the Partners IRB allegation-to-ohrp. They declined to do so ohrp reply.
Perhaps one day, when the United States gets over it’s obsession with male genital mutilation and can look at the issue dispassionately, historians will add a small footnote to the sad history of casual experimentation on brown people. For now it makes you want to weep.
Jim Thornton
Ripe-tomato.org 
(Uploaded with permission) David Gisselquist writes: I note that OHRP’s Borror said only unanticipated problems need to be reported to IRBs. Not sure how that works, because the IRB is responsible for the quality of reporting, so how can they be responsible if they are not getting info on all adverse events? In any case, here’s some relevant language from the page she cites:
http://www.hhs.gov/ohrp/policy/advevntguid.html#Q2 I’ve copied in several paras below.
Also, when the Plank study reports adverse events, shouldn’t they be reporting all adverse events, including deaths? As the US government web page Borror cites makes clear, the definition of adverse events includes things that are not and might not be related to the research.
B. Assessing whether an adverse event is related or possibly related to participation in research […]
Determinations about the relatedness of adverse events to participation in research commonly result in probability statements that fall along a continuum between definitely related to the research and definitely unrelated to participation in the research. OHRP considers possibly related to participation in the research to be an important threshold for determining whether a particular adverse event represents an unanticipated problem. In this guidance document, OHRP defines possibly related as follows:
There is a reasonable possibility that the adverse event may have been caused by the procedures involved in the research (modified from the definition of associated with use of the drug in FDA regulations at 21 CFR 312.32(a)).
OHRP recognizes that it may be difficult to determine whether a particular adverse event is related or possibly related to participation in the research… .
Thank you, Jim Thornton, for revealing how investigator discretion, backed by high sounding rationalisations, can warp and corrupt RCTs. What you found does NOT satisfy my standards of research integrity.