Aviva Health Care cover vasectomy.

Great news. Aviva HC will pay 350 Euros towards the cost of a vasectomy for Aviva members with level 2 or above cover.

Dr John O’Keeffe wins prestigious prize

Against stiff opposition, Dr John O’Keeffe of the Morehampton Clinic Dublin was awarded the prestigious 2014 Ross Ardill Memorial Plate for his presentation to the Irish Primary Care Surgeons Associaton. The title of his paper was ‘Keyhole No Scalpel Vasectomy – a new approach’.


Ross Ardill Prize


Vasectomy and Prostate Cancer – a possible link?

There has been much interest in a recent American study suggesting a link between vasectomy and Prostate cancer. Initially this seemed very worrying, but on review by statisticians the link seems unfounded.

Here is a rather long and technically involved rebuttal of these claims by Professor Michel Lebreque, a prominent Canadian academic

The debate is very similar to the one “we” had in 1993 when the first two Giovannucci articles were published. Actually the current article is simply the extension of one of these articles (attached).

Just to remind you the figures from the 1993 article, involving much less cases than in 2014 of course:

Giovannucci 1993 No vasectomyN=37800 VasectomyN=10055 RR non adjusted RR age adjusted
Prostate cancer (AI excluded) 225 (0.60%) 54 (0.53%) 1.13* 1.56
Prostate cancer C+D 104 (0.28%) 22 (0.22%) 1.27 N/A


In the initial article, they had also reported a larger association with longer time since vasectomy (22 yrs ), but the method used to reach this conclusion was flawed. An interaction term should had calculated to evaluate heterogeneity between sub-groups instead of relying on the RR on each sub-group.

For a discussion on subgroup analysis, see

The mains results of the current (2014) article are the following:

Siddiqui 2014 No vasectomyN=37084 VasectomyN=12321 RR RR age-ajusted RR multi-factor adjusted
Prostate cancer (AI excluded) 4499 (12.1%) 1524 (12.4%) 1.02 Non reported 1.10
Prostate cancer (advanced cases) 900 (2.4%) 244 (2.0%) 0.83 Non reported 1.20

Those are the potential biases in this report

Confounding bias

As David pointed out in the vasectomy cohort, there is LESS risk of having an advanced (including lethal) cancer that in the non-vasectomy cohort. It is only after multi-factor adjustment that the difference appears positive. This is a very large “swing” in the RR due to adjustment. This means that confounding is very important and when I look at the data of the factors for which adjustment was done, I do not understand. To be a confounding factor, a variable needs to be a risk factor for the disease (prostate cancer ) AND be associated with the exposure (not being distributed equally in vasectomy and non-vasectomy group). Look at Table 1. You will see that only small (or no) difference between groups for race, height, BMI, physical activity, smoking, diabetes, family history of prostate cancer, multivitamin, Vit E, alcohol, so this cannot really explain the difference between adjusted and non-adjusted RR. Only the age and psa testing history (related to age at Dx) is really different. I wish they had reported the age-only adjustment…

I was then surprised to see factors such as height, smoking, diabetes, family history of prostate cancer taken into account as potential confounders in the models as their distribution is exactly the same in both groups, as reported in table 1… I do not understand why they did those unnecessary adjustments. Furthermore, height and diabetes are NOT known risk factors of prostate cancer. See

How come they were then considered as potential confounders? If their data showed that these were indeed risk factors, this is rather strange and this question the validity of some data. In addition, they excluded region, religion, history of std, ejaculation, tomato sauce, BMI at age 21 as potential confounders because they were not associated to vasectomy. Region and religion are associated to vasectomy according to many studies, so this is again strange that it was observed in their data…

So although residual confounding may be possible, they did a good job at considering all possible potential confounding factors. However , I am surprised by some of the choice they made in conducting their analyses and question the reliability of some data (information bias).

Information bias

There could be some errors in assessment of vasectomy (self-report by health professionals, so probably not bad), of Dx of prostate cancer (also self-reported but confirmed with medical records in 90% of cases with an end committee reviewing the cases- no mention however who was on that the committee and, very importan,t if the members were blinded from vasectomy status [usually you write it when it is so, and it is not!]), and of all the potential confounding variables. From table 1 it seems that they have all the data from everybody. This is impossible! As a clue, in the 1993 paper, the sample size of the cohort was 47,885 men (instead of 49,405 now) because they had excluded mean with not adequately completed dietary questionnaire. This represent 1,550 men (3,1% of the 49,405). So I really doubt that they have ALL the information for ALL men. Nothing is reported on missing data and how they were handled.

Selection bias

Difference in characteristics of participants cause either a selection or a confounding bias. In a cohort study, selection bias occurs when there is a differential loss-to-follow-up between the exposed and non-exposed cohort. Here they started with 49,405 men. In the regression model (Cox proportional hazard model) a patient was censured until ca prostate Dx, death or end of follow-up. They do not appear to have lost any participants during the 24-year follow-up! This is either unusual and exceptional or an error in not reporting. So we cannot assess the risk of selection bias.

Subgroup analysis

As for sub group analysis, they repeated the error from the 1993 paper. They present sub group analysis for time since vasectomy and age at vasectomy. They did not find any difference this time, but again they did not perform an adequate analysis to assess heterogeneity between subgroups. This is the same for intensity of prostate cancer screening. They isolated a subgroup with high intensity screening, but the adequate method is to categorize the whole cohort in subgroups according the intensity of screening and see if there is a statistically significant difference between the subgroup (the statistical significance of the interaction term in the model). This is not reported.

Effect size

Although some differences are statistically significant, the clinical significance is questionable as many already mentioned, even if the results were valid (true). However, this is question of judgment. Many people in the urological community believe that the results the ERSPC trial are highly clinically significant when in fact the RR is 0.79, the same 20% relative difference that was calculated adjusted relative difference calculated by Siddiqui et al. . Of course in the screening study it is a 20% decrease and in the vasectomy study it is a 20% increase. In the screening trial, the absolute decrease is 1 out of 1000 men screened, while the increase in advanced disease with vasectomy would be an increase of 5 out of 1000 men (20% over the 2.4% of advanced cancer in the non- exposed group = 0.5%). So, if Siddiqui analysis is valid, vasectomy would cause advanced prostate cancer in 5 men out of 1000 who had a vasectomy!

But is causation can be inferred? Let’s look at the Hill-Bradford criteria.

1: Strength of Association. The stronger the relationship between the independent variable and

the dependent variable, the less likely it is that the relationship is due to an extraneous variable.

A strong association is higher than 2-3. So the answer is NO

2: Temporality. It is logically necessary for a cause to precede an effect in time.

YES, they excluded all men with prevalent cancer at baseline

3: Consistency. Multiple observations, of an association, with different people under different

circumstances and with different measurement instruments increase the credibility of a finding.


4: Theoretical Plausibility. It is easier to accept an association as causal when there is a rational

and theoretical basis for such a conclusion.


5: Coherence. A cause-and-effect interpretation for an association is clearest when it does not

conflict with what is known about the variables under study and when there are no plausible

competing theories or rival hypotheses. In other words, the association must be coherent with

other knowledge.


6: Specificity in the causes. In the ideal situation, the effect has only one cause. In other words,

showing that an outcome is best predicted by one primary factor adds credibility to a causal



7: Dose Response Relationship. There should be a direct relationship between the risk factor

(i.e., the independent variable) and people’s status on the disease variable (i.e., the dependent



8: Experimental Evidence. Any related research that is based on experiments will make a causal

inference more plausible.


9: Analogy. Sometimes a commonly accepted phenomenon in one area can be applied to another



The global answer is NO.

The AUA perform a review and meta –analysis of the available literature in the vasectomy guideline. If the AUA were to repeat the meta-analysis (see below) by changing Giovannucci 1993 prospective study with the Siddiqui 2014 study (it is the same study), I bet there will be no change!!! The conclusion would be the same… or maybe all negative studies were wrong?