Subscribe to:
POZ magazine
E-newsletters
Join POZ: Facebook MySpace Twitter Pinterest
Tumblr Google+ Flickr MySpace
POZ Personals
Sign In / Join
Username:
Password:

WHO and UNAIDS Bless Crusade to Circumcise Africans

| 25 Comments


One of the most striking aspects of these trials is that the figures purporting to show that such programs can reduce HIV transmission are those for relative risk reduction, not absolute risk reduction. While a 60% relative risk reduction may sound impressive, a 1.3% absolute risk reduction is not even statistically significant. Why are we being given selective and highly misleading data about circumcision if it is as important an intervention as we are told it is by its proponents?


The simple answer is that there is no credible evidence in favor of mass male circumcision; it needs to be manufactured. In fact, there is evidence that circumcision substantially increases the risk of transmission of HIV from males to females and more than a hint that men risk being infected with HIV through the circumcision operation itself. As a result of the well funded propaganda surrounding these circumcision campaigns, a whole set of myths is emerging. Some people, male and female, think circumcision protects against HIV; it is claimed that it is not possible to be hygienic if uncircumcised; there are allusions to circumcision being more aesthetically pleasing; it is implied that the operation is 'fashionable' or modern, etc.


The levels of misinformation being spread about male circumcision are astounding. Arguments for adult male circumcision have even been used for infant circumcision, although the claimed effects of adult circumcision have not been demonstrated for infant circumcision. 'Experts' extol the multiple virtues of circumcision, ignoring the lack of evidence for their claims, indeed, apparently blind to the entirely unscientific nature of many of the claims. After stating that  "a circumcised [male organ] is definitely cleaner than an uncircumcised one" 'Dr' Khumbulani Moyo, Clinical Director of Population Services International goes on to say "Circumcised men are also more likely to be assertive sexually as awareness of a good body image is a very important factor in building self confidence." I wonder what his doctoral thesis was on; yoga perhaps?


Boyle and Hill note that the three trials purporting to show the effectiveness of circumcision were carried out in countries where it was already clear that HIV prevalence was higher among uncircumcised men. However, there are just as many countries where HIV prevalence is higher among circumcised men. They ask why the evidence to support a program that may aim to circumcise as many as 38 million men is so selective and point out that with less selective analysis, the program would not be supported. There are so many biases and inadequacies in the data that it can not be used to justify carrying out what is likely to be a dangerous, unnecessary and perhaps even counterproductive program.


It's hard to do this lengthy and well researched paper justice in a short blog post, but it's worth mentioning that one of the many flaws in the research is that non-sexual transmission of HIV was not reported. Quite a number of the men infected with HIV during the trials were probably not infected sexually and could have been infected through unsafe healthcare, perhaps even the treatment they received through taking part in the trial. Mass male circumcision enthusiasts claim that the operation reduces sexual transmission, but many men (and women) might face high non-sexual risks in addition to any sexual risks. But trials into circumcision and other HIV prevention interventions rarely seem to consider non-sexual risk.


There is a substantial body of evidence showing that male circumcision either doesn't reduce HIV transmission or even that it increases transmission. This evidence is not often mentioned by those whose aim appears to be to promote the strategy at all costs. In contrast, there is evidence that 'circumcising' women may be associated with some reduction in HIV transmission without this giving rise to the same enthusiasm for female genital mutilation. There is something of the crusader about the circumcision enthusiasts, something cabalistic in their methods. But what appears to be entirely lacking is science and logic.


25 Comments

Show Comment(s)

Comments on Simon Collery's blog entry "WHO and UNAIDS Bless Crusade to Circumcise Africans"

"a 1.3% absolute risk reduction is not even statistically significant."

Can you explain what you mean by that? How did you calculate it?

Refer to the Boyle and Hill paper cited, pages 7 and 11. If you're confused about the meaning of 'statistical significance', refer to Wikipedia.

The paper just makes the same false claim that you do in this post:

p7: "That the very small absolute reduction of about 1.3% was not statistically significant"

p11: "the absolute decrease in HIV infection was only 1.31%, which is not statistically significant."

The wikipedia article explains how a p value is calculated, which is indeed correct and exactly how the p values were calculated in all three randomized controlled trials of circumcision in HIV prevention; all three results are statistically significant. To state that the absolute decrease in risk was not statistically significant is simply a lie.

Unless you are able to explain how you, or Boyle and Hill, calculated that the reduction was not statistically significant. Do you care to have another try?

I suggest you contact the authors.

You're making a claim here about statistical significance that you can't defend, because it isn't true.

“What does the frequently cited ‘60% relative reduction’ in HIV infections actually mean? Across all three female-to-male trials, of the 5.411 men subjected to male circumcision, 64 (1.18%) became HIV positive. Among the 5.497 controls, 137 (2.49%) became HIV positive, so the absolute decrease in HIV infection was only 1.31%, which is not statistically significant”p.326

I read most of the paper, I think Simon is referring to this? Thats how I see it, I'm still trying to figure out where they got the 60%?

The 60% figure is relative risk reduction; doesn't give you any basis for making a decision about whether to get circumcised or not but it sounds a lot better than 1.31%.

http://www.cdc.gov/botusa/Success-Stories/0307-MaleCircumcision-Botswana.htm

WHY is this a "success story?" Read on. "Kebaabetswe and her colleagues put Botswana ahead of the game by conducting a study in 2001 to determine the acceptability of circumcision in the country, as well as the preferred age and setting for male circumcision." 2001!... as a guessing game to promote circumcision, BEFORE the highly questionable trials.

Simon has it exactly right, "There is something of the crusader about the circumcision enthusiasts, something cabalistic in their methods. But what appears to be entirely lacking is science and logic."

As his comment about statistical significance shows, Simon is no position to judge anything about these data. To assert that the reduction in absolute risk was not statistically significant is a lie - statistical significance is calculated based on the number of endpoints in each arm of a trial, and the difference was highly statistically significant in all three trials.

In a population of 1,000,000 people, a difference between 2.49% and 1.18% would mean a difference between 24,900 infections and 11,800 infections.

In my anger, I've been guilty of talking out of my arse too, statistical significance can be calculated for absolute risk:

http://books.google.com/books?id=AoltVRLgW1MC&pg=PA6&lpg=PA6&dq=%22#v=onepage&q&f=false

Although my stridency suggested a facility with statistics that I cannot in any way claim, my accusation was not unjustified:

http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1114210/

"When the treatment effect is significant at the 5% level, the 95% confidence interval for the absolute risk reduction will not include zero"

There seems to be some confusion between "statistically significant" (not likely to be due to chance) and significant in the ordinary sense "worth taking notice of". After the Boyle/Hill paper says "not statistically significant" it adds "(relevent from a policy implementation perspective)".

Since the 60% (73/137) and the 1.3% (73/~5,400) refer to the same number, if one is statistically significant, the other must be too. But just because the difference is not likely to be due to chance, that doesn't mean it was due to circumcision. As Gissellquist points out (http://tinyurl.com/7deqtap), non-sexual transmission has been greatly understated in these studies. If non-circumcised men were more likely to get medical treatment - because they are less afraid of doctors? - that could explain the HIV difference. I suspect that (statistically) signficantly many circumcised men dropped out of the trials after they found out they were HIV+ - wouldn't you?

Nor is a 1.3% difference necessarily big ("significant") enough to add it to HIV prevention measures - let alone promote it with the kind of hysterical fervour that is being "rolled out" across Africa, to the exclusion of measures known to work, such as condoms. In Malawi, the Catholic Bishops are endorsing circumcision while condemning condoms, and this is simply - I don't exaggerate - murderous.

While the results of the RCTs were "statistically" significant, they fail the "So What! Who Cares?" test. Absolute risk reductions is so small as to be almost meaningless. We use these RCTs in teaching our medical students and medical residents as an example of how an RCT is not necessarily the gold standard of medical evidence and why one of the lower grades of evidence places a poorly done RCT as having less value than most observational trials. When we hand out the articles to the students and residents we ask them to find to tell us if these trials were ethical and if not why not. We have a prize for the learner who identifies the most forms of bias. Usually at the end of the discussion the students wonder how these studies ever got funded or published, especially in the Lancet. After reading the studies, one student commented, "I don't think I can ever trust what I read in the Lancet again."

The studies all have nearly identical methods and share the same built-in forms of bias, all of which increased the estimated treatment effect. The first big problem was that half of the men did not get HIV from sexual contact, so that would reduce the absolute risk reduction from 1.3% to 0.65%. The lead-time bias would reduce the ARR by another 17% to 0.54%. Early stoppage of studies is associated with an over estimate of treatment effect by 30%. so the ARR becomes 0.38%. There is also the expectation bias in both the researchers and participants and the Hawthorn effect this produced. This could have reduced the treatment effect by another 10% at least, so the ARR is now 0.34%. The studies were powered to find a 1% ARR if it exists. The studies would not find a 0.34% difference, so the people designing the study had to find a study design so they could "find" the the 1% difference they needed.

This does not even account for the missing data problem in which there were several men lost to follow-up for every man who became infected. The accepted approach to missing data is to calculate the extremes and note that the truth is within that range. So if one assumes that all of the control group men lost to follow-up became HIV infected and none of the intervention group became infected the relative risks in the three studies would be:

Auvert: 0.10, 95%CI=0.07-0.16
Gray: 0.07, 0.04-0.11
Bailey: 0.14, 0.09-0.21

If all of the intervention group men lost to follow-up were to become infected and none of the control group men lost to follow were to become infected the relative risks would be:

Auvert: 2.52, 1.81-3.51
Gray: 5.60, 4.33-7.71
Bailey: 3.13, 2.25-4.35

The bottom line is these studies have a very serious missing data problem, to the point where the results are basically worthless.

As pointed out in a previous comment, for the absolute risk reduction to be statistically significant the 95% confidence interval should not cross 0. In this particular case, the 95% confidence interval for the absolute risk reduction of 0.0131 is 0.0081 to 0.0182, so it is statistically significant.
Knowing the absolute risk reduction also allows one to calculate how many men must be circumcised to prevent one infection. That number is 76.

As it is statistically significant the absolute risk reduction of 1.31% achieved by circumcision is a reliable measure of just how poorly effective circumcision is in preventing HIV.
In line with Hugh7’s comment, one can only wonder what’s driving the zealous promotion of a poorly effective intervention that carries risks of its own, with the almost complete neglect of prevention education that encourages and supports continued condom use, which is the most effective means we have of preventing the sexual transmission of HIV, and also the safest and least expensive.
One might say much the same about pre-exposure prophylaxis – PrEP. In terms of absolute risk reduction PrEP doesn’t work very well, it’s expensive and carries risks, yet it’s being triumphantly trumpeted as a “game changer”, while budgets for prevention education are being cut.
Resources are barely allocated to the most effective intervention we have to prevent the sexual transmission of HIV, while being very generously devoted to promoting PrEP, the use of which may very well result in an increase in new HIV infections.
Newer prevention interventions should of course be studied, but care should be taken to continually strengthen support for condom use.
A future generation will wonder why we neglected what we know can work in the fervid portrayal of a much less effective intervention as a “game changer” with its relentless promotion

The confusion is because the authors of the paper lied. Statistical significance is not subject to reinterpretation based on agenda-driven subjectivity. Nor are the data fungible based on the ego-driven whims of some guy in an armchair in Pennsylvania.

Nor should the master plan to circumcise the planet come from a small, influentially placed circumcision enthusiasts. It should be based on good science, not what was generated by a bunch of circumcision fanatics in their biased, poorly designed studies. The problem is that "significant" has two meanings. One is the statistical sense (p<.05), the other is the non-statistical sense of being of great importance. The findings in the RCTs, while significant in the statistical sense, but is not significant in the non-statistical sense. What happened in an armchair in Pennsylvania did not make the RCTs poorly designed and of little or no scientific value. This happened when the circumcision enthusiasts turned "researchers" failed to determine the source of the infections that occurred during the trials and built-in lead time bias, and paid the participants in the trials what would be worth over $10,000 if the trials were conducted in the United States. Your harping on this small error in word usage does not alter the reality that the WHO, UNAIDS, and a variety of medical journal editors continue to promote the circumcision solution, even when they know the science is bad.

Isn't the reinterpretation of 'statistical significance' continually being done by the pro-circumcision lobby pushing their agenda for mass circumcision? I think the armchair is at Johns Hopkins. These researchers seem to have been given a license to lie and are rarely held accountable. Almost every pro-circumcision article I have reviewed makes claims that are not justified by the data and evidence. Why are the standards for research and publication so different when it comes to those promoting circumcision, while those opposing it can rarely even get published in the U.S.?
What are the psychological motivations behind cutting the genitals of innocent children and unsuspecting adults in third-world countries? These are the questions I want answered. The three African RCT's, if you can call them RCT's, found nothing of any clinical significance and probably not of any statistical significance either when adjusted for all the biases as noted above.

@Joseph Sonnabend

"As it is statistically significant the absolute risk reduction of 1.31% achieved by circumcision is a reliable measure of just how poorly effective circumcision is in preventing HIV."

So this argument acknowledges that the results are statistically significant. Surely the whole point of calculating statistical significance is to provide a reasonably objective measure of an effect, but you are saying that the significance just allows you to be sure that the intervention is "poorly effective"--in your entirely subjective opinion. And since this opinion is based on the absolute risk reduction, it largely depends on the absolute risk in the population, which was low (provision of counseling and condoms might have had something to do with that).

Is there a level of absolute risk in a population that would cause you to think that an intervention that reduced relative risk by 60% was better than "poorly effective"?

"Knowing the absolute risk reduction also allows one to calculate how many men must be circumcised to prevent one infection. That number is 76."

How many would it be over a twenty year period?

Richard Jefferys is in need of an epiphany because he repeatedly doesn't seem to understand difference between the statistical and clinical realms. The statistical realm tells you there might be a real difference and the chance that the difference happened randomly are small. Some differences are inconsequential. That is a clinical determination. The number needed to treat can help put the clinical realities in numbers. In this example 75 men would have the most sensitive part of the penis cut off, suffer the consequences of possible complications, and receive absolutely no benefit from the amputation. Such a yield from a surgical intervention is considered unacceptable. Let's compare that to appendectomy for someone with suspected appendicitis. 20% will not have appendicitis and will have undergone a surgical procedure with no benefit. The number needed to treat would be 1.67 (1/(80%-20%)).

How about circumcising HIV positive men. In a randomized controlled trial 12% of the female partners of HIV-positive men who were not circumcised became HIV infected, while 18% of the female partners of HIV-positive men who were circumcised became HIV-infected. This study was stopped, rightly so, before the difference became statistically significant. The absolute risk risk reduction was 6%. The number of circumcisions needed to infect a female partner (also know as number needed to harm) is 16.7. So basically for every 17 circumcisions performed someone died. This clearly a clinically important finding that was not statistically significant that is repeatedly dismissed by the researchers, turned lobbyists, who performed the research.

Considering that there are more effective methods of prevention available that do not remove body parts and increase the risk to female partners, why is circumcision even part of the discussion. The money could be better spent on clean medical equipment and needles, condoms, and secondary prevention with ARTs.

The RCTs were overpowered on purpose so the researchers, turned lobbyists, could boast about the significant findings, when what they found was inconsequential.

If you can predict what will happen over the next 20 years with less than 24 months worth of data, you should not waste your time reading blogs, because you have a special gift that no one else has.

No one knows the trajectory the data would take because the studies were stopped prematurely (in part because they were overpowered and had found a difference that was statistical but inconsequential).

"This study was stopped, rightly so, before the difference became statistically significant."

Why rightly? They could not prevent any more infections by circumcising the control group - on the contrary, perhaps. At that stage they just had to go on observing. All the men were HIV+ so they should been giving all the women equal warnings.

The suspicion arises that they stopped the trial SO THAT it could not become statistically significant that circumcising men INcreases the transmission to women. This is a possiblity they have not addressed, but there is good reason to suspect it might. The keratinised circumcised glans would be more likely to create microtears in the vaginal wall than the smooth mucosa of the intact glans.

They rightly stopped the study early because the intervention was killing more people than in the control group. I agree that it is morally reprehensible and academically dishonest for them to claim that the reason for stopping the study was futility and for them to recommend circumcision of HIV+ men so that they look like everyone else, when they knew doing so would kill more women. I guess for Marie Wawer, African women have even less value than African men.

They should certainly have stopped circumcising, but there was nothing to stop them continuing to observe, to see whether the partners of the circumcised men continued to get HIV at a greater rate, or that was just a consequence of premature resumption of sex before the wounds had healed, as they prematurely claimed. Gisselquist at http://tinyurl.com/7deqtap has something to say about the ethics of this.

A lot of ideology by a number of people who do not seem to understand the science or know how clinical research is conducted. Who bothered reading any of the paper on MC? Here is the Auvert one conducted in South Africa, freely available.

http://www.plosmedicine.org/article/info:doi/10.1371/journal.pmed.0020298

It explain where the 60% comes from: There were 20 HIV infections in the circumcion group and 49 in the control group. This result is equivalent to saying that during the period of trial, the intervention prevented six out of ten potential infections.

It also details how much data were missing:
The fraction of participants lost to follow-up (who did not complete the study) was 8.0 % (251/3128), with 6.5% (100/1546) in the intervention group and 9.5% (151/1582) in the control groups.

The MC effect was sustained for at least 42 month in the Kisumu, Kenya Tria which initially reported on 24 months http://bit.ly/LJO9wR

The study was stopped because it would have been unethical to carry on knowing the intervention works. Studies are closely monitored by indepedent monitoring committees and ethics committees to ensure participants health is not put at risk (read more about trials here http://www.ct-toolkit.ac.uk/)

Data on the protective effect of MC does not come from only 3 major trials, but fron 20 years of scientific evidence. Too much asking checking them?

Of course it is much easier to ignore the hard facts, and battle on ideological grounds. In the meantime, people get infected, infect their partners and die.

The facts are out there.

Some of the facts are, as you say, out there. And much of what you and the papers you cite say may be true. But the purported aim is to circumcise many millions of men on the grounds that HIV is almost always transmitted sexually and that circumcision reduces sexual transmission (from women to men). The research does not show that those infected were infected sexually. This undermines their claim about sexual transmission if they don't know what modes of transmission were involved. My argument is that much of the evidence is unclear and that it is premature to roll out such a potentially risky intervention until some of the commonly expressed worries about circumcision and sexual HIV transmission have been addressed.

As for ethical matters, I suggest you read the following article. Ethics committees don't always fulfill their expressed aims:

http://dontgetstuck.wordpress.com/2012/02/11/denied-withheld-and-uncollected-evidence-and-unethical-research-cloud-what-really-happened-during-three-key-trials-of-circumcision-to-protect-men/

However, I accept that the link you supply dates from 2004 and much of the circumcision trial data was being collected many years before that. But your certainty about ensuring that participants' health is not put at risk may be challenged. Many trial participants were not told they were HIV positive, there was no requirement for them to find out their status, nor any requirement for their sexual partners to be informed.

Leave a comment



Archives

 

My Favorite Links

Subscribe to Blog

About this Entry

This page contains a single entry by Simon Collery published on March 6, 2012 1:56 PM.

Timberg and Halperin's Tinderbox: a Veritable Candle on a Sunny Day was the previous entry in this blog.

To What Extent Does Male Circumcision Reduce HIV Transmission? is the next entry in this blog.

Find recent content on the main index or look in the archives to find all content.

Pages

Disclaimer

The opinions expressed by the bloggers and by people providing comments are theirs alone. They do not necessarily reflect the opinions of Smart + Strong and/or its employees.

Smart + Strong is not responsible for the accuracy of any of the information contained in the blogs or within any comments posted to the blogs.



© 2014 Smart + Strong. All Rights Reserved. Terms of use and Your privacy