Flawed study: NARI,Pune study on the effect of circumcision on HIV transmission

[Guest guest]

Flawed study: Indian Council of Medical Research (ICMR)-National AIDS Research

Institude (NARI) study on the effect of circumcision on HIV transmission. The

Lancet.

_____________________

Male circumcision and risk of HIV-1 infection

Sir-- Reynolds and colleagues (Mar 27, p 1039)1 have recast old

data from the Mehendale group in Pune, India, on female-to-male

infection with HIV-1. They purport to show a significant reduction in

female-to-male infection if the man is circumcised.

The exhaustive Cochrane review2 of the evidence for possible

protection from female-to-male sexual transmission of HIV-1 by

circumcision concluded that " insufficient evidence exists to support

an interventional effect of male circumcision in heterosexual males " .

In reviewing previous studies from the Mehendale group in Pune, the

Cochrane review noted a low or unstated participation rate, failure

to control for religion, and an imbalance between circumcised and

uncircumcised groups. As in these previous studies, Reynolds and

colleagues' report is also affected by selection bias because it

involved high-risk groups from sexually transmitted disease clinics

rather than the general population.

The present study also contains major statistical flaws. Although

Reynolds and colleagues attempt to control for religion, a confounder

remains because only three participants in the uncircumcised group

(0•6%) were Muslim. As Reynolds and colleagues state, " When non-

Muslim men were assessed separately, the protective effect was not

significant " .

Furthermore, the discrepancy in sample sizes (191 circumcised men vs

2107 genitally intact men) clearly suggests heterogeneity of

variance, such that reported " statistically significant effects "

might be little more than statistical artifacts, especially with the

large overall sample size (n=2298) and resultant high level of

experimental power.

Although multiple univariate effects are reported, the corresponding

multivariate effects are not reported (in the absence of a

significant multivariate effect, interpretation of univariate effects

is likely to be difficult).3 At the very least, basic Bonferroni

corrections should be done to keep to a minimum the risk of

claiming " significant effects " that are due to chance alone.3 Also, I

am surprised that no effect sizes are reported.

Researcher bias should not be ignored. There is a strong tendency to

defend the culture of origin.4 The Cochrane review

notes: " Circumcision practices are largely culturally determined and

as a result there are strong beliefs and opinions surrounding its

practice. It is important to acknowledge that researchers' personal

biases and the dominant circumcision practices of their respective

countries may influence the interpretation of their findings " .2

The study also notes a substantial failure to use condoms during

visits with female sex workers. Clearly, irrespective of circumcision

status, the HIV-1 infection rate would approach zero in both groups

if condom use were universal.

The investigators tacitly acknowledge that circumcision would not be

culturally acceptable to Hindu men. In addition, there are other

factors to consider before taking any decision to introduce

circumcision. These include potential adverse medical and

psychosexual effects, as well as legal, ethical, and human rights

issues.5

The statistical inadequacies highlighted weaken the validity of

Reynolds and colleagues' study. Therefore it would be prudent to

await the results of the three randomised controlled trials now

underway2 before any conclusions are drawn about HIV-1 transmission

and male circumcision status.

J Boyle

Bond University, Gold Coast, Queensland 4229, Australia (e-

mail:gboyle@...)

1 Reynolds SJ, Shepherd ME, Risbud AR, et al. Male circumcision and

risk of HIV-1 and other infections in Pune, India. Lancet 2004; 363:

1039-40. [Text]

2 Siegfried N, Muller M, Volmink J, et al. Male circumcision for

prevention of heterosexual acquisition of HIV in men (Cochrane

Review). In: The Cochrane Library, issue 3. Oxford: Update Software,

2003.

3 Keppel G. Design and analysis: a researcher's handbook. Englewood

Cliffs, NJ: Prentice Hall, 1991.

4 Goldman R. The psychological impact of circumcision. BJU Int 1999;

83 :(suppl 1) 93-103. [PubMed]

5 Boyle GJ. Issues associated with the introduction of circumcision

into a non-circumcising society. Sex Trans Inf 2003; 79: 427-28.

[PubMed]

____________________

Sir-- Reynolds and colleagues1 claim that their study supports

a biological, rather than behavioural, explanation for the

observation, in some studies, of a lower incidence of HIV-1 in

circumcised than in uncircumcised men. Their analysis is

unconvincing.

Religion (an important determinant of both behaviour and social

interaction) is clearly a confounding factor in their study. Showing

a lack of association between circumcision status and selected

sexually transmitted infections does not exclude the possibility that

another religion-related difference (other than circumcision) might

be responsible for a difference in the rate of acquisition of HIV-1

by men of different religious cultures.

A previous study in Pune2 found other factors associated with HIV-1

infection among men attending sexually transmitted disease clinics.

These included ulcerative and non-ulcerative sexually transmitted

diseases and education levels. Some studies in which religion was not

a confounding factor have not shown an association between

circumcision and HIV-1, or have shown a positive association. The

population-based study at Carletonville, South Africa,3 found that

the main risk factors for HIV were related to sexually transmitted

infections, sexual behaviour, and age. Men who were seropositive for

herpes simplex virus type 2 (HSV-2) were seven times more likely to

be HIV positive than men who were HSV-2 seronegative. No protective

effect of circumcision on HIV prevalence was shown. The UK Gay Men's

Sex Survey4 of more than 14 000 men showed a significant association

between being circumcised and being HIV positive--an association that

was consistent across all ethnic groups and age groups. Perhaps there

is a biological reason why circumcised men are more vulnerable to

HIV?

The probability of acquiring HIV-1 is a function of exposure to risk

and the likelihood of transmission when exposed to risk. The study is

silent on this point, but few, if any, of the Muslim men in the study

acquired HIV-1. It follows that the prevalence of HIV-1 among Muslims

in Pune must have been very low. If their sexual relations did not

expose the Muslim men to risk then it is likely that this, rather

than any possible circumcision-dependent difference in the

transmission rate, explains the difference in the rate of acquisition

of HIV-1 by men of different religions. One could argue that HIV-1

prevalence among Muslims is low precisely because most Muslim men are

circumcised, but this is the very point requiring proof. Pooling

Muslim men with non-Muslims to find a lower incidence of HIV-1 in

circumcised men obscures that requirement, besides introducing a host

of confounding factors.

Throughout their study the authors refer to the " protective effect "

of circumcision as though it were an established fact, rather than

the hypothesis being tested. In a systematic review5 of published and

unpublished studies, the Cochrane Library found insufficient evidence

to support the claim, noting that observational studies are

inherently limited by confounding.

Darby

15 Morehead Street, Curtin, Australian Capital Territory 2605,

Australia (e-mail:robjld@...)

1 Reynolds SJ, Shepherd ME, Risbud AR, et al. Male circumcision and

risk of HIV-1 and other sexually transmitted diseases in India.

Lancet 2004; 363: 1039-40. [Text]

2 Rodrigues JJ, Mehendale SM, Shepherd ME, et al. Risk factors for

HIV infection in people attending clinics for sexually transmitted

diseases in India. BMJ 1995; 311: 283-86. [PubMed]

3 B, Auvert B, Carael M, et al. HIV infection among youth in

a South African mining town is associated with herpes simplex virus-2

seropositivity and sexual behaviour. AIDS 2001; 15: 885-98. [PubMed]

4 Reid D, Weatherburn P, Hickson F, et al. Know the score: findings

from the National Gay Men's Sex Survey 2001. Portsmouth: Sigma

Research (University of Portsmouth), 2002.

5 Siegfried N, Muller M, Volmink J, et al. Male circumcision for

prevention of heterosexual acquisition of HIV in men (Cochrane

Review). In: The Cochrane Library, issue 3. Oxford: Update Software,

2003.

_____________________

Sir-- Reynolds and colleagues1 allege that they have proved the

value of circumcision in preventing female-to-male infection with HIV-

1. They have failed in their mission.

Franco2 observes: " It is a daunting job to verify whether a

particular behavior or medical intervention truly operates on the

etiologic pathway of HIV infection when there are so many confounders

along the way. There are many alternative scenarios that are

consistent with the observed data. These alternatives need to be

shown as implausible before one can establish that male circumcision

protects against HIV infection later in life. The burden of proof is

on the epidemiologist to show that the study design is as free of

selection biases as possible, that risk factor information has been

measured with the best available instruments, and that careful (even

obsessive) statistical analysis has zealously controlled for every

possible confounder " .

Gray and colleagues3 have shown that genital ulceration and viral

load are the main determinants of infection in coital acts. Although

Reynolds and colleagues make some attempt to control for genital

ulceration, they completely ignore viral load. Furthermore, the

participants in their study were attendees at three sexually

transmitted disease clinics, so the selection is heavily biased.

Efforts to control HIV-1 infection must be consistent with human

rights.4 The right to bodily integrity, therefore, should be

protected. Even if circumcision is eventually shown to be protective,

consent of the individual, which can only be given by adults, must be

sought before doing circumcision. This might prove a heavy burden to

overcome, since there is increasing awareness of the sexual handicaps

conferred by circumcision,5 and doctors have an ethical duty

to " first do no harm " .2

Marilyn Fayre Milos

National Organization of Circumcision Information Resource Centers,

PO Box 2512, San Anselmo, CA 94979, USA (e-mail:nocirc@...)

1 Reynolds SJ, Shepherd ME, Risbud AR, et al. Male circumcision and

risk of HIV-1 and other sexually transmitted infections in India.

Lancet 2004; 363: 1039-40. [Text]

2 Franco EL. Male circumcision and AIDS in Africa: primum non nocere

versus the collective good. Epidemiology 2004; 15: 133-34. [PubMed]

3 Gray RH, Wawer MJ, Brookmeyer R, et al. Probability of HIV-1

transmission per coital act in monogamous, heterosexual, HIV-1-

discordant couples in Rakai, Uganda. Lancet 2001; 357: 1149-53.

[Text]

4 United Nations High Commissioner for Human Rights. HIV/AIDS and

human rights. Geneva: Office of the United Nations High Commissioner

for Human Rights, 2002. http://www.unhchr.ch/hiv/index1.htm (accessed

Apr 7, 2004).

5 Shen Z, Chen S, Zhu C, et al. Erectile function evaluation after

adult circumcision. Zhonghua Nan Ke Xue 2004; 10: 18-19 (in

Chinese). [PubMed]

______________________

Authors' reply

Sir-- Boyle and Darby suggest that our study is flawed

because religion is a confounder. Since circumcision is practised

according to religious affiliation in most parts of Africa and India,

one would have expected that confounding would be present for all the

pathogens studied, not only for HIV-1. It was the specificity of the

observed protective effect of circumcision for HIV-1 in our findings

that was so provocative and which supported a biological rather than

behavioural explanation. The male participants in this study,

irrespective of religion, attended the same sexually transmitted

infection (STI) clinics in Pune and reported high-risk behaviour with

prostitutes from the same brothels. It is difficult to think of a

confounder, restricted to Muslims, that would differentially block

HIV-1 transmission, but not transmission of the other three pathogens

considered (herpes simplex virus type 2, Treponema pallidum, and

Neisseria gonorrhoeae).

Boyle also claims that the discrepancy in sample sizes between

circumcised and uncircumcised men in our study creates statistical

artifacts. However, the statistical methods used are valid whether

sample sizes are equal or not. Boyle also claims that " effect sizes "

were not reported. However, such values were reported in table 2 in

the column labelled " adjusted relative risks " . Circumcised men were

shown to be at 0•15 times the risk of HIV-1 infection as

uncircumcised men. The p value was less than 0•01, showing that even

if a Bonferroni adjustment for the four pathogens examined was used,

the effect would still be significant.

In answer to Marilyn Fayre Milos's point about viral load, this

variable has proved a major determinant of transmission among HIV-1-

serodiscordant couples in Rakai, Uganda.1 Ours was not a discordant-

couples study and we could not control for viral load given our

methods. However, for viral load to be a potential confounder, the

viral load of the sexual partner would need to be associated with our

participant's circumcision status, which we feel is not biologically

plausible.

This study was done in patients attending STI clinics in Pune as part

of an investigation to understand the risk factors for HIV-1

infection in the Indian setting. As with most observational studies

of this nature, this limits the generalisability of the findings to

other populations; however, it does not affect the internal validity

of our findings.

Our objective was to add to worldwide information on the biology of

sexual transmission of HIV-1, especially in view of the Cochrane

review2 concluding that insufficient evidence exists from

epidemiological findings about the protective effects of

circumcision. The definitive word on the effects of male circumcision

of HIV-1 transmission will probably come from three clinical trials

currently underway in Uganda, Kenya, and South Africa.

J Reynolds, E Shepherd, S Brookmeyer, Sanjay M

Mehendale, * C Bollinger

*Division of Infectious Diseases, s Hopkins University,

Baltimore, MD 21205, USA (SJR, RCB); s Hopkins Bloomberg School

of Public Health, Baltimore, MD, USA (MES, RSB); National AIDS

Research Institute, Pune, India (SMM) (e-mail:rcb@...)

1 Gray RH, Wawer MJ, Brookmeyer R, et al. Probability of HIV-1

transmission per coital act in monogamous, heterosexual, HIV-1-

discordant couples in Rakai, Uganda. Lancet 2001; 357: 1149-53.

[Text]

2 Siegfried N, Muller M, Volmink J, et al. Male circumcision for

prevention of heterosexual acquisition of HIV in men (Cochrane

Review). In: The Cochrane Library, issue 3. Oxford: Update Software,

2003.

http://www.thelancet.com/journal/vol363/iss9425/full/llan.363.9425.cor

respondence.29822.1