It is not, of course, up to the media to decide what is good or bad science. The media was reporting what it heard from scientists [about cold fusion]. Only a tiny fraction of all scientific research is ever covered by the popular media, however, and most scientists go through their entire career without once encountering a reporter. New results and ideas are argued in the halls of research institutions, presented at scientific meetings, published in scholarly journals, all out of the public view. Voodoo science, by contrast, is usually pitched directly to the media, circumventing the normal process of scientific review and debate. ... The result is that a dispro­portionate share of the science seen by the public is flawed.

- "Voodoo Science" by Robert Park, pp26-7

The Randomised Clinical Tests

The three studies compared
Method of circumcision
Loss from study
Non-sexual transmission
Blood-borne transmission
Effect of curtailing the studies
Predictions based on the studies
The Hawthorne Effect
Humans are not lab rats
Other published cautions

National Institute of Allergy and
Infectious Diseases (NIAID)

EMBARGOED FOR RELEASE
Wednesday, December 13, 2006
12:00 Noon ET

Adult Male Circumcision Significantly Reduces Risk of Acquiring HIV
["Significantly" is used here in a strict statistical sense, different from the common meaning.]
Trials in Kenya and Uganda Stopped Early

The National Institute of Allergy and Infectious Diseases (NIAID), part of the National Institutes of Health (NIH), announced an early end to two clinical trials of adult male circumcision because an interim review of trial data revealed that medically performed circumcision significantly reduces a man's risk of acquiring HIV through heterosexual intercourse. The trial in Kisumu, Kenya, of 2,784 HIV-negative men showed a 53 percent reduction of HIV acquisition in circumcised men relative to uncircumcised men, while a trial of 4,996 HIV-negative men in Rakai, Uganda, showed that HIV acquisition was reduced by 48 percent in circumcised men. ["Impressive sounding reductions in relative risk can mask much smaller reductions in absolute risk." - editorial in the British Medical Journal, January 19, 2008]

"These findings are of great interest to public health policy makers who are developing and implementing comprehensive HIV prevention programs,"says NIH Director Elias A. Zerhouni, M.D. "Male circumcision performed safely in a medical environment complements other HIV prevention strategies and could lessen the burden of HIV/AIDS, especially in countries in sub-Saharan Africa where, according to the 2006 estimates from UNAIDS, 2.8 million new infections occurred in a single year."

"Many studies have suggested that male circumcision plays a role in protecting against HIV acquisition," notes NIAID Director Anthony S. Fauci, M.D. "We now have confirmation — from large, carefully controlled, randomized [but not double blinded] clinical trials — showing definitively that medically performed circumcision can significantly lower the risk of adult males contracting HIV through heterosexual intercourse. While the initial benefit will be fewer HIV infections in men, ultimately adult male circumcision could lead to fewer infections in women in those areas of the world where HIV is spread primarily through heterosexual intercourse."

The findings from the African studies may have less impact on the epidemic in the United States for several reasons. In the United States, most men have been circumcised. Also, there is a lower prevalence of HIV. Moreover, most infections among men in the United States are in men who have sex with men, for whom the amount of benefit [if any] provided by circumcision is unknown [but is likely to be much less, because HIV is known to be more readily transmitted to the receptive male partner]. Nonetheless, the overall findings of the African studies are likely to be broadly relevant regardless of geographic location: a man at sexual risk who is uncircumcised is more likely than a man who is circumcised to become infected with HIV. Still, circumcision is only part of a broader HIV prevention strategy that includes limiting the number of sexual partners and using condoms during intercourse. [In that case, any benefit provided by circumcision would only apply in the rare cases where a condom breaks or comes off.]

The co-principal investigators of the Kenyan trial are Robert Bailey, Ph.D., M.P.H., of the University of Illinois at Chicago, and Stephen Moses, M.D., M.P.H., University of Manitoba, Canada. In addition to NIAID support, the Kenyan trial was funded by the Canadian Institutes of Health Research and included Kenyan researchers Jeckoniah Ndinya-Achola, M.B.Ch.B., and Kawango Agot, Ph.D., M.P.H. The Ugandan trial is led by Ronald Gray, M.B.B.S., M.Sc., of Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland. Additional collaborators in the Ugandan trial were David Serwadda, M.Med., M.Sc., M.P.H., Nelson Sewankambo, M.B.Ch.B., M.Med.M.Sc., Stephen Watya, M.B.Ch.B., M.Med., and Godfrey Kigozi, M.B.Ch.B., M.P.H.

Both trials involved adult, HIV-negative heterosexual male volunteers assigned at random to either intervention (circumcision performed by trained medical professionals in a clinic setting) or no intervention (no circumcision). All participants were extensively counseled in HIV prevention and risk reduction techniques.

[With AIDS running at 4.10% in the population (according the the CIA's World Factbook), selecting men who are HIV-negative means that already

• they are likely to have some natural immunity
• they are likely to be more careful than the average person and

the fact that they volunteer implies they have more concern about HIV/AIDS than others. These introduces biases that make circumcision likely to be less effective when applied to the general population.]

Both trials reached their enrollment targets by September 2005 and were originally designed to continue follow-up until mid-2007. However, at the regularly scheduled meeting of the NIAID Data and Safety Monitoring Board (DSMB) on December 12, 2006, reviewers assessed the interim data and deemed medically performed circumcision safe and effective in reducing HIV acquisition in both trials. They therefore recommended the two studies be halted early. All men who were randomized into the non-intervention arms will now be offered circumcision.

[For statistical reasons, effectiveness of a treatment declines with the passage of time. Cutting the experiment short gives a falsely optimistic outcome.]

"It is critical to emphasize that these clinical trials demonstrated that medical circumcision is safe and effective when the procedure is performed by medically trained professionals and when patients receive appropriate care during the healing period following surgery," notes Dr. Fauci.

[But once the meme "Circumcision prevents HIV" is loose in the community, this will be forgotten and circumcisions will be done under unhygienic conditions with shared instruments, quite possibly under duress.]

Researchers have noted significant variations in HIV prevalence that seemed, at least in certain African and Asian countries, to be associated with levels of male circumcision in the community. In areas where circumcision is common, HIV prevalence tends to be lower; conversely, areas of higher HIV prevalence overlapped with regions where male circumcision is not commonly practiced.

[These correlations require highly selective use of statistics. There are many exceptions: HIV is rare in Cuba, where circumcision is also rare, and common in Lesotho, where circumcision is common, and common among both the Zulu of South Africa who do not circumcise, and the Xhosa, who do.]

Results of the first randomized clinical trial assessing the protective value of male circumcision against HIV infection, conducted by a team of French and South African researchers in South Africa, were reported in 2005. That trial of more than 3,000 HIV-negative men showed that circumcision reduced the risk of acquiring HIV by 60 percent. The trial was funded by the French Agence Nationale de Recherches sur le Sida (ANRS) (see http://www.anrs.fr/).

[Earlier studies claimed an eight-fold reduction. As each new study corrects the errors of its predecessors, the claimed benefit goes down. In this, it resembles parapsychological research. The suspicion arises that when all confounding factors have been allowed for, circumcision will confer no benefit at all.

The Relative Risk Reduction of 53% seems impressive, but when the rates of HIV infection in the experimental and control populations are considered, the results are less impressive.

Cut infection rate in 12 months	1.58%
Intact infect. rate in 12 months	3.38%
Absolute risk reduction	1.8 (95% CI: 0.64-2.95)
Relative risk reduction	53% (95% CI: 23-72)
Odds ratio	0.45 (95% CI: 0.27-0.77)
Number needed to treat in 1 year	56 (95% CI: 34-155)
In other words, you would have to circumcise 56 men to prevent one of them contracting HIV in one year.   Click on the image for an enlarged version   And the number needed to prevent HIV longer term is higher. Doctors could spend their time better spent treating people with ulcerative disease and malaria, which make HIV transmission easier and using the money saved to promote safer sexual practices. Few accepted medicines have such a high NNT. On this basis, the NNT in developed countries such as the USA, where the HIV rate is relatively low (0.6% compared to 4.1% in Uganda), would be much higher - it would take 380 circumcisions in the US to prevent one case of HIV.

For more information on the Kenyan and Ugandan trials of adult male circumcision, see the NIAID Questions and Answers document at http://www3.niaid.nih.gov/news/QA/AMC12_QA.htm.

Abstract
Impact of male circumcision on the female-to-male transmission of HIV Auvert B.¹, Puren A.², Taljaard D.³, Lagarde E.⁴, Sitta R.⁴, Tambekou J.⁴
1 UVSQ - INSERM U687 - APHP, ST Maurice CEDEX, France, 2 NICD, Johannesburg, South Africa, 3 Progressus CC, Johannesburg, South Africa, 4 INSERM U687, St Maurice, France

Introduction: Observational studies suggest that male circumcision could protect against HIV-1 acquisition. A randomized control intervention trial to test this hypothesis was performed in sub-Saharan Africa with a high prevalence of HIV and where the mode of transmission is through sexual contact.

Methods: 3273 uncircumcised men, aged 18-24 and wishing to be circumcised, were randomized in a control and intervention group. Men were followed for 21 months with an inclusion visit and follow-up visits at month 3, 12 and 21. Male circumcision was offered to the intervention group just after randomization and to the control group at the end of 21 month follow-up visit. Male circumcisions were performed by medical doctors. At each visit, sexual behavior was assessed by a questionnaire and a blood sample was taken for HIV serology. These grouped censored data were analyzed in an “intention to prevent” univariate and multivariate analysis using the piecewise survival model, and relative risk (RR) of HIV infection with 95% confidence interval (95% CI) was determined.

Results: Loss to follow-up was <11%; <1% of the intervention group were not circumcised and < 2% of the control group were circumcised during the follow-up. We observed 45 HIV infections in the control group and 15 in the intervention group, RR=2.77 (95% CI: 1.56 – 4.91; p=0.0005). When controlling for sexual behavior, including condom use and health seeking behavior, the RR was unchanged: RR=2.93 (p=0.0003).

Conclusions: Male circumcision provides a high degree of protection against HIV infection acquisition. Male circumcision is equivalent to a vaccine with a 63% efficacy. The promotion of male circumcision in uncircumcised males will reduce HIV incidence among men and indirectly will protect females and children from HIV infection. Male circumcision must be recognized as an important means to fight the spread of HIV infection and the international community must mobilize to promote it.

Some factors casting doubt on the findings:

1. A summary of the study includes this:

   Inclusion criteria: ... Consenting to avoid sexual contact (except with condom protection) during the 6 weeks following the medicalized circumcision

   The experimental (circumcised) men, but not the control group (left intact), were told:

   When you are circumcised you will be asked to have no sexual contact in the 6 weeks after surgery. To have sexual contact before your skin of your penis is completely healed, could lead to infection if your partner is infected with a sexually transmitted disease. It could also be painful and lead to bleeding. If you desire to have sexual contact in the 6 weeks after surgery, despite our recommendation, it is absolutely essential that your (sic) use a condom.

   So: 1. The circumcised experimental group, but not the intact control group, got into the HABIT of using condoms
   2. They learnt HOW to use condoms
   3. They had to make sure they HAD condoms (which are in scandalously short supply in South Africa), and
   4. last but not least, they were PROTECTED by condoms.
   

   The researchers could hardly say to the experimental group, "but after that you don't have to use condoms" could they?

   Meanwhile the intact control group was not required to use condoms for the first six weeks of the study, just sent out to take their chances.

   This throws the results into, er, a cocked hat.

2. The circumcised men would have had to take some time away from any sexual activity, reducing their exposure to HIV.

3. The circumcised men would inevitably get more exposure to safe-sex information during their time in medical hands, waiting for and recovering from, their operations.

4. Humans are not lab rats. They have sex in non-random ways. Many of the men in the study would put themselves at little or no risk of contracting HIV, a few at great risk, so the effective sample size is much smaller than it appears, making the margin of error much larger.

5. Because HIV-positive men were excluded from the study, there would have been a higher proportion of men with natural immunity in both groups than the general population, reducing the effective sample size still further.

6. Because all the subjects did not just agree to be circumcised but wished to be, they were not a representative sample of the general population.

7. 11-14 percent of the original participants (360 - 458 men) were lost to study or disqualified from continuing. Their HIV status and/or circumcision status might not be typical of the total (for example, if they dropped out because they lost interest in the experiment when they found circumcision had not protected them), introducing sufficient bias to refute the claimed finding.

8. Michel Garenne of the Institut Pasteur in Paris points out that the claimed 63% efficacy is not comparable to that of a vaccine and would result in considerable infection over time.

9. Jennifer Vines, MD, of the Oregon Health & Science University in Portland, comments "...the authors did not control for other sources of HIV transimission such as blood transfusions or exposure through infected needles. ... Controlling for this route of infection could result in a smaller difference between HIV infection rates in the circumcised versus uncircumcised groups, indicating that circumcision may not be as effective at decreasing HIV transmission as the article suggests."

10. Columnist Stephen Strauss (below) points out that the study was cut short before even half as many men were infected as were infected before it began.

11. The Lancet (which earlier published a strident call for circumcision by Robert Bailey) refused to publish the study (apparently with ethical concerns about not telling men they had HIV). The study has been published by the Public Library of Science, an "open source" online medium.

12. The Abstract of the AIDS Conference in Rio reported 15 seroconversions from the circumcised group and 45 seroconversions in the uncircumcised group. (The New Scientist, 6 August, reported 15 seroconversions in the circumcised group but 51 in the uncircumcised group. On 29 July the Science and Development Network reported 18 seroconversions in the circumcised group and 51 in the uncircumcised group.) On 23 October, PLoS reported that there were 20 seroconversions in the circumcised group and 49 in the uncircumcised group. From the official figures: 15-45 at the AIDS Conference in Brazil and 20-49 in the PLoS Journal, between 1 August and 23 October there appear to have been 4 seroconversions among the uncircumcised and 5 seroconversions among the circumcised: in less than 3 months, a 3:1 difference has shrunk to 2.45:1 difference.

13. We've seen it many times before. Circumcision is touted as the great panacea for this or that dreaded disease of the age - but as the studies are refined, the advantage withers away.

14. The rampant evangelism of the Conclusion suggests that the experimenters are not altogether detached.

The foreceps-guided method, in which the foreskin is pulled forward and cut, removes significantly less mucosa than the sleeve procedure in which a strip of tissue is taken from behind the glans (and a method like the forceps-guided has been blamed for the high rate of HIV infection in Lesotho, where most men are circumcised). Yet the degree of HIV reduction is substantially the same for the two methods - suggesting circumcision is not what is causing the difference.

All three trials had significant numbers "lost from study", their HIV status unknown (orange bars in the graphs below) - 100 circumcised subjects (6.5%) in South Africa, 87 (10%) in Kenya and 140 (3.5%) in Uganda. (The figures are presented confusingly in the studies because the men did not all enter the trials together, but each trial was stopped at a stroke.) Those figures are high enough in themselves to cast doubt on the validity of the results, but circumcised men who found they had HIV would be disillusioned with the trials and less likely to return. It would take only 25, 25 and 23 such men respectively to completely nullify the trials, and fewer to render the results non-significant.

The orange part of each of the three right-hand bars (below the dotted lines) represents the much-hyped "60% protection" conferred by circumcision. If just those men, whose HIV status is unknown, proved in fact to be HIV+ (red), circumcision would certainly have no protective effect whatever, but it would not take all of them to reduce the effect below statistical significance.

In the South African trial, one third (23 of 69) of the HIV infections occurred in men who reported no unprotected sex during the period from their last negative test to their first positive test. In Uganda, 16 of 67 new infections occurred in men who reported no sex partners (6 infections) or 100% condom use (10 infections). The trial in Kenya did not report how sexual exposures related to HIV incidence, except for seven men infected in the first three months (sensitive tests did not find HIV in the men's blood at the beginning of the trial). Five of those seven, including three of four who had been circumcised, reported no sexual exposures from the beginning of the trial until their first HIV-positive test.

The studies ignored exposure to HIV by blood. In the two studies that reported information on genital symptoms, 30-43% of infections with HIV occurred during intervals when men reported genital ulcers or other genital symptoms or problems. Because genital symptoms were more common in uncircumcised men, they may have been more likely to contract HIV from skin-piercing procedures such as injections to treat genital symptoms, but the studies did not consider that possibility. None of the studies reported on injections or on any other blood exposures during follow-up. In the Kenyan trial, four men became HIV-positive a month after circumcision, so the circumcision itself might have infected them, but the study did not mention that possibility.

For example...

If the African studies had not been stopped early and long-term results had been obtained, the HIV infection rate might very well have become statistically insignificant between the circumcised and non-circumcised groups. Look at the progression in the number of cases of HIV in the Kisumu study:

The number of cases in each period for each group is small, so their relative sizes are affected greatly by random variation. It appears from the data that the rate of infection is lower among the circumcised men in the first 18 months following circumcision, but that there's little difference beyond 18 months. If the study had not been terminated early at 24 months, it is quite likely that the number of HIV cases between the groups would have become insignificant. The decision to terminate the studies early prevented any future comparison of the progression of HIV in the circumcised and control groups and the very real possible invalidation of the alleged "proof".

One of the researchers (Gray) has the nerve to extrapolate the figures into the future from his truncated study, claiming to show that the rate of "protection" increases over time:

One probability is that the incidence in the first six months is higher because they got HIV from their circumcisions! - if there is any non-random causal relationship at all.

It is utterly innumerate to extrapolate anything from such tiny numbers of cases, p-values or not. If he'd done the same to the intact men, he'd find the "protection" from being intact increased over time too!

Medscape

Dr. Wainberg: Are we ready as a world to make recommendations in regard to more widespread surgical procedures such as male circumcision?

Dr. Auvert: The answer is no. For sure we have a clear scientific answer about the association between circumcision and HIV infection. For sure we have demonstrated that in South Africa and this part of the world we did see a population level reduction of HIV infection in this trial, but we are not ready to use this as a prevention method right now. The situation in Africa is quite complex -- you've got a lot of different cultural situations and it's not possible.

The Lancet 2006; 368:1236

DOI:10.1016/S0140-6736(06)69513-5

Correspondence

Cautious optimism for new HIV/AIDS prevention strategies

Edward Mills ^a and Nandi Siegfried ^b

b. Clinical Trial Service Unit, Department of Medicine, University of Oxford, Oxford, UK

The Hawthorne effect refers to the phenomenon that when people are observed in a study, their behavior or performance temporarily changes. Others have broadened the definition to mean that people’s behavior and performance change, following any new or increased attention. The term gets its name from a factory called the Hawthorne Works in Illinois, where a series of experiments on factory workers were carried out between 1924 and 1932. Most notably, production went up when the lighting was increased, and it went up when the lighting was decreased: it was the attention the workers were getting when the measurements were taken, not the lighting, that caused the effect.

The Randomised Controlled Trials are subject to the Hawthorne Effect because they were not double blind: all the subjects knew which group they were in, and what effect this was supposed to have. The Hawhtorne Effect could not have directly affected the extent to which they were infected with HIV, but it could have affected their sexual behaviour, making the circumcised men more aware of safer sexual practises (having part cut off one's penis concentrates the mind wonderfully), and perhaps more likely to implement them. They reported no change in their sexual behaviour, but self-reporting may not be accurate: their reporting of homosexual behaviour, for example, is so low it attracts the strong suspicion that they were under-reporting it.

"At Toronto, sociologists and anthropologists in particular were sceptical of the narrow form of ''science'' being touted as the only form of evidence needed. Activists and practitioners, e.g. people living with HIV and AIDS, those working in the non-governmental sector and prevention workers - those who comprise the bulk of the ''AIDS community'' - were concerned with a potential undercutting of their hard-won shifts in sexual cultures, in many places, toward safe sex practices." ...

"After all, these trials were not test tube experiments but experiments conducted in clinical settings. Such settings are profoundly social moments with real human interactions and complex components, even if RCT design in principle tries to circumvent such inputs. For example, how do we assess the fact of these trials not being double-blinded: the men in each arm clearly knew their circumcision status? That known difference could have affected how the men responded behaviourally, psychologically and sexually."

A literature search found a much greater proportion of the studies of circumcision were of adverse effects, ethics, ethnology, history, legislation and jurisprudence, than (the proportion) of the studies of appendectomy ("the surgical removal of a part of the body seen as somewhat unimportant") or hysterectomy ("a more serious and controversial sexual and reproductive health operation") .

From the conclusion:
"We believe we need to know much more about male circumcision for HIV prevention before adopting it as a population health measure. The WHO/UNAIDS Statement is cautious in noting the existence of caveats and gaps, but it argues that it is time to go ahead. We would argue that there is still much work to do before national authorities and the global HIV/AIDS community can feel confident about proceeding."

Reproductive Health Matters 2007;15(29):33-44