Though the data previously cited are insufficient to establish a definitive death rate on their own, there is enough available information to calculate an estimate. Not all of the reported 35.9 deaths out of 1,243,392 circumcisions can be attributed to related causes.
(Here Bollinger references a figure he has provided previously: "Hospital discharge records reveal that, during the 1991–2000 decade, on average 35.9 boys died from all causes each year during their stay (average 2.4 days) in the hospital in which both their birth and circumcision occurred (Thompson Reuters, 2004).")
What portion, then, is circumcision-related and how may we extrapolate to the number of deaths after hospital release? What we can safely assume is that it is unlikely that any of these infants would have been subjected to the unnecessary trauma of circumcision if they had been in critical condition, or that they would have been circumcised after their death.
This is indeed a relatively safe assumption, though it is not one that actually gets us any closer to an answer.
Gender-ratio data can help extrapolate a figure. Males have a 40.4% higher death rate than females from causes that are associated with male circumcision complications, such as infection and hemorrhage,4 during the period of one hour after birth to hospital release (day 2.4), the time frame in which circumcisions are typically performed (CDC, 2004). Assuming that the 59.6% portion is unrelated to gender, we can estimate that 40.4% of the 35.9 deaths were circumcision-related. This calculates to 14.5 deaths prior to hospital release.
This is extraordinary! Bollinger is, in effect, assuming that the difference between male and female death rates is due entirely to circumcision. But it is a well-established fact that male babies are more susceptible to deaths than females, and there is no evidence that this is due to circumcision. Indeed, if circumcision alone were responsible for the difference, then we might expect countries with low circumcision rates to have the same infant mortality rates among males and females. But in fact, that's not the case, as the following table shows:
|Country||Est. neonatal circ. rate||IM (male)||IM (female)||IM m:f ratio|
|United Kingdom||< 5%||5.40||4.28||1.26|
Table: Infant mortality (IM) rates for selected countries. Derived from female rate table and male rate table.
Clearly, infant mortality rates are consistently higher among males regardless of circumcision rates. So Bollinger's approach is clearly flawed. When he is trying to estimate the risk due to circumcision he is actually estimating the risk due to being male!
But it gets even worse (this would be laughable if the subject weren't so serious). Even if we assume that Bollinger's method is sane and appropriate (in spite of evidence to the contrary), he manages to miscalculate those attributable to being male. If the rate is 40.4% higher among males then the observed rate (35.9) will be the rate in females plus 0.404 times that rate again (or 1.404 times the rate in females). So, to find the rate in females:
1.404f = 35.9
f = 35.9 / 1.404 = 25.57
And so the rate attributable to being male will be 40.4% of that, which is 10.33.
But, as noted, this is the rate attributable to being male, not to circumcision.
But as is often the case with hemorrhage and infection, some circumcision-related deaths occur days, even weeks, after hospital release. The CDC’s online searchable database, Mortality: Underlying cause of death, 2004 (CDC), lists causes by various age ranges and reveals that the percentage of deaths after release, compared with deaths before, is 772% greater. This ratio is comparable to Patel’s (1966) 700% postrelease infection rate.
Bollinger expresses this with less than optimal clarity, but what he seems to be saying is that the ratio between deaths in the hospital stay (which Bollinger identifies as typically 2.4 days) and those after the hospital stay (but presumably within the first 28 days of life) is 7.72.
Frankly, that shouldn't be surprising. There are 10.7 times as many days in the latter period than there are in the former, so one would ordinarily expect more deaths simply due to there being more time in which people can die.
Multiplying the 772% adjustment factor for age-at-time-of-death by the 14.5 hospital-stay deaths calculated above, the result is approximately 112 circumcision-related deaths annually for the 1991–2000 decade, a 9.01/100,000 death-incidence ratio.
This multiplication is irrational. It stands to reason that there would be more deaths in the first 28 days than the first 2.4 days, simply because there is more time in which infants can die. If we look at the first 100 years of life, then the ratio will be even greater (in fact, the mortality rate over that period will be almost 100%), but would it make any sense to apply that ratio? Of course not — people die of other things than circumcision, and it wouldn't make any sense.
It doesn't make sense to apply this multiplication, either. Yes, a certain number of circumcision-related deaths will likely occur some time after the event, but it doesn't make any sense to assume, in effect, that any deaths in the period must be due to circumcision.
Applying this ratio to the 1,299,000 circumcisions performed in 2007, the most recent year for which data are available (HCUP, 2007), the number of deaths is about 117. This is equivalent to one death for every 11,105 cases, which is not in substantial conflict with Patel’s observation of zero deaths in 6,753 procedures. It is more than some
other estimates (Speert, 1953; Wiswell, 1989),
It is perhaps a little disingenuous to refer to these as "estimates". These are observations showing 1 death in 566,000 circumcisions (Speert), no deaths in 100,000 boys (Wiswell). Similarly, King reported no deaths in 500,000 circumcisions. So if we use 1 in 500,000 as a reasonable estimate, we would expect 2.6 deaths in 1.3 million circumcisions. Bollinger's errors have led him to a figure some 45 times greater than that which can be extrapolated from actual statistics!
but less than the overstated 230 figure derived from Gairdner (1949). Breaking this statistic down further, about 40% of these deaths (47) would have been from hemorrhage, and the remainder (70) from sepsis, using a hemorrhage-to-sepsis ratio for infant mortality (NCHS, 2004).
Yes, I suppose the nice thing about imaginary numbers is that there is an inexhaustible supply of them.