Wednesday, February 22, 2012

Problem with Mathematical Modelling in Circumcision & HIV Prevention

The whole basis of mathematical modelling is based on the accuracy of assumptions they are based on, and their ability to accurately predict the future from these assumptions.  If the assumptions the mathematical modelling are based on, are in any way incorrect, false, distorted or less than assumed, then the mathematical predictions based on those flawed assumptions will be totally inaccurate.

The mathematical modelling which was used by researchers to convince the WHO to endorse circumcision in Africa are fundamentally flawed.  Basically the assumption is this, the results from the clinical trials assumed that circumcision was the only variable that accounted for a reduction in HIV infections and therefore this will translate will translate exactly into the exact proportion of future reductions in HIV infections in the real life circumstances of Africans, .Any innacuracy in assumptions will totally distort any mathematical predictions based on these assumptions.

Lets look at the flaws in these assumptions

That men will not modify their behaviour following circumcision, such as have sex without using condoms believing they are protected by circumcision, less condom use is a completely different assumption not accounted for, and if occurs as some evidence already suggests, will totally change the predicted outcomes of  reduction in hiv infections
The research did not prove life-long protection only partial episodic protection (relative to the trial conditions for only 18months, and did not follow and support them for the course of their lives).
The trials also did not reflect not real world settings, and therefore how can they be applicable to real world settings. Eg. Clinical trilas with education, counselling, educations, 1st world medical treatment.

Rarely reported or emphasised, A significant number of circumcised men in the study became infected with HIV.

The participants were not randomly selected, but selected themselves, creating a potential bias or distortion in the generalisability of the results to any general population.
The participants were paid adult male volunteers who wanted to be circumcised and therefore had a bias in favour of it, and could possibly have been in favour of circumcision because they were high risk candidates who had unprotected sex (no condoms) with multiple sex partners, and were looking for a way to maintain high risk behaviours
Therefore as the participants were not randomly selected, and were a potentially biased self-selected sample of the population, the results cannot be extrapolated to general populations outside of this population sub-group. (Van Howe & Storms, 2011)

Inadequate Controls: Participants in the trials were not treated equally with the circumcised group given more education about healing from surgery, advised to not resume sex for 6 to 8 weeks and therefore, abstained from sex longer, and participants were given greater time and emphasis about wearing condoms during the period of healing from surgery. No control was undertaken to examine possible non-sexual blood exposures by participants. No control was undertaken for dry sex as practised by some african cultures. No control for the sex (gender) of partners, and no control for anal intercourse.
(Van Howe & Storms, 2011)
Unexplained and disrtorting the statistics was the finding that in the first three months of the Kenyan trial, five men became HIV-positive who reported no sexual activity in the period before the seroconversion (0.73/100 person-years, 95%CI=0.30-1.76). (Van Howe & Storms, 2011)
Data suggests a percentage of infections were from non-sexual exposures, with 23 infected men reporting no sexual contact without a condom. No explanations or investigations undertaken for non-sexual exposures to HIV infections. (Van Howe & Storms, 2011)

The African HIV Trial researchers were all pro-circumcision and with a history of activism in the area.

Nearly 10 times as many participants dropped out of the clinical studies as were infected, with HIV status unknown.

The studies were ended early exagerrating effects.

The vast majority of participants in the study were HIV free, therefore, why was no attempt made by researchers to identify the 100% condom users and compare these to the circumcised group, Was 100% condom use more effective than circumcision = Most probably yes!! but researchers did not want to find this and report it.

No long term follow-up possible with all subjects circumcised at end of trial.

Researchers used speculative hypotheses to explain trial findings, such as Langeran present in the foreskin cells are targeted by HIV, whereas later research found Langeran cells actually kill HIV.
(Van Howe & Storms, 2011)

The studies had such high numbers of participants leading to an overpowering of the statistical analysis, inflating the results.
(Van Howe & Storms, 2011)

French demographer Garenne criticised the findings by demonstrating that interventions with a near 50% clinical trial efficacy had very little population effect.

At best the research findings are only valid for adult circumcision volunteers, and populations with high prevalence of HIV, not babies or low prevalence nations. At worst the research is so floored the findings only have validity within similar research conditions and virtually zero validity for real world situations.

1 comment:

  1. You've distilled from Van Howe and Storms many valid criticisms of the African clinical trials. Another long article that is a devastating critique of the African clinical trials is: