IFIN BULLETIN #156: A response to the Phipps study in the BMJ.

Oct 10, 2000.

Dear All,


The Phipps study published in the British Medical Journal on Oct 6, 2000,
is one of the studies on hip fracture which I examined as part of my
invited peer review of the York report ( A summary of the York report was
also published in the same issue of the BMJ). The York team incorporated
her data in their meta-analysis of bone fracture. Phipps' paper is one of
the 18 (3 unpublished and one abstract) studies done since 1990. 10 of
these studies show an assoication between increased hip fracture and
fluoridation and 8 do not. Phipps is one that did not.

I am glad that Phipps does not hide the her enthusiasm for water
fluoridation, when she declares without a supporting reference that, "While
the benefit of fluoridation in the prevention of dental caries has been
overwhelmingly substantiated". However, the notion that in addition to the
marginal benefits to teeth that fluoride may have is the protection against
hip fracture is  a stretch in the context of the other 17 studies. While it
is true that her study is superior to some because she controls for 13
variables, and she uses bone mineral density as an indirect confirmation
that those who have more exposure to fluoridated water have accumulated
more fluoride, she still lacks the actual levels of fluoride that have
accumulated in the bones. Instead she relies for her comparisons on the
number of years of exposure to fluoridated water. For the study period
1971-1990, the authors give the mean age of the women examined as 74.5,
74.2 and  73.9 years in the three groups observed: no exposure, mixed
exposure and continuous exposure, respectively. If this was their age in
1990, then the mean age of those exposed was 53.9 years in 1971. Thus most
of their known exposure was after menopause, which according to  some
authors could make a difference with respect to fluoride's impacts
(Danielsen, 1992), although not all agree on this point (Kurtio, 1999).

However, the most disturbing aspect of the report is how much attention is
given to the DECREASE of hip fracture incidence and how little attention is
given to the INCREASE in the incidence of wrist fracture in the group
exposed for 20 years of water fluoridation. The ostensible reason for this
is that the decrease in hip fracture incidence is deemed statistically
significant while the increase in  wrist fracture is deemed statistically
insignificant. However, when one considers the basis of the claims of
significance and insignificance the difference between the two results is
very slender indeed. This is particularly important when it is recognized
that the "significance" for the hip fracture decrease and the
"non-significance" of the wrist fracture increase only emerges after
adjustment for 12 variables. It raises the question of how accurate these
adjustments were, if such fine distinctions are going to be made.

Here are the details. In Table 5, after age adjustment, the authors report
a relative risk of 0.85 for hip fractures for the continuously exposed
group, with a 95% confidence interval (CI) of 0.63 to 1.14 and a p-value of
0.287. Translated this indicates a statistically non-significant decrease
of 15% in the incidence of hip fracture. It is statistically
non-significant because the CI overlaps with 1.00 and the p-value is
greater than 0.05. In the same table, the authors report the relative risk
for hip fracture, after adjustment for 12 more variables, as 0.69, with a
CI of 0.50 to 0.96 and a p-value of 0.028. This translates to a decrease in
hip fracture of 31% and it is now statistically significant because the 95%
confidence interval (0.55 to 0.96) no longer overlaps with 1.00 and the
p-value is less than 0.05.

If we now compare this with the wrist fracture figures we find the
following. The age adjusted relative risk for wrist fracture for the women
continuously exposed is 1.36 (CI: 1.07 - 1.73), p-value 0.012. This
translates to a 36% increase in wrist fracture which is statistically
significant, because the 95% CI does not include the value of 1.00, and the
p-value is less than 0.05. After adjustment for 12 more variables, these
figures become 1.32 (CI: 1.00 to 1.71) with a p-value of 0.051. This is now
declared as a non-significant finding because, even though the relative
risk has hardly changed, despite the consideration of 12 variables, the CI
just overlaps with 1.00, in fact the lower value is actually at 1.00 and
the p-value is just over 0.05 at 0.051! This is about as close as one get
to significant result without actually calling it a statiscally
significant, as you can get. This is so close in fact, that this result
must bring into question how accurately these adjustments, and the
assumptions on which they were based, were performed.

In this respect it is intriquing that when the York team considered these
same results from Phipps et al they recorded the adjusted figure as 1.3
(1.02, 1.7)
(see their appendix C8 "Bone Studies: Individual Study results") and this
is the number reported in the final version of the York Review (Oct 6,
20000. This table can be examined on the web at
http://www.york.ac.uk/inst/crd/fluores.htm. The table can be found in the
appendix identified as appc8.doc.

The fact that this number makes the difference between a so-called
significant result and a non-significant result, raises a serious question
as to why two reports published on the same day have different values - one
significant the other non-significant. How did this number get changed and
by whom?

Another question I have is this. In my efforts to secure the Phipps report
before publication in the BMJ, as part of my peer review of the York
report, I discovered that Kathy Phipps had provided testimony in person
before the York team. There have been 18 studies since 1990 pursuing a
possible relationship between increased hip fracture in the elderly and
fluoride in drinking water, why was this one author (Kathy Phipps) pulled
out for special treatment?

These issues need very careful evaluation because many news outlets and
health wire services are broadcating the "good news" that water
fluoridation is good for bones. In my view this promotion is reckless based
upon such slender evidence, and in the context of several important
factors: a) 50% of all fluoride ingested accumulates in our bones b) water
fluoridation is not the only source of fluoride we are exposed to c) high
doses of fluoride used to treat patients with osteoporoisis in an effort to
harden their bones has led to an increase not a decrease in hip fracture d)
this study also indicates an increasse in wrist fracture, as close to
significance as you can get and e) there are 10 studies (3 unpublished)
which indicate an increase in hip fracture associated with fluoride in
water and f) one of these studies shows an almost dose-reponse increase
above 1 ppm exposure ( Li et al, 1999, unpublished).

All the references cited above can be found in articles I have authored or
co-authored on our webpage http://www.fluoridealert.org


Paul Connett.