On Lunar phases and dog bites: analysing differential exposure.
In response to two articles with contradictory results that were
published in the BMJ (1,2) on the effect of lunar phases on the frequency
of animal bites, and the discussion that followed them, we undertook an
analysis in order to test the same hypothesis with data derived from a
Emergency Department Injury Surveillance System (EDISS) in Greece. The
first of the above articles stratified the number of bites in 10 periods
of moon and found a significantly increased risk of bite at or near full
moon. No adjustment for weekday or other variables was made. The second
article stratified number of bite cases by weekday, and full versus no
full moon, and found no significant
increase in risk. In comparing the two methods, therefore, the one
addressed possible weekday effect but not the continuum among lunar
phases, whereas the other addressed the continuum of lunar phases, but not
the weekday effect.
We believe that both, weekday effects and the continuum of lunar
phases, should be addressed simultaneously in such a problem, and that
ignoring either issue, as done in both articles, can possibly create
extraneous relations between
number of bites and full moon simply because of differential exposure. For
example, adjusting for weekday is necessary because more people tend to be
outdoors (which could imply higher exposure to bites) systematically on
certain days (e.g., weekends) and not on others. Similarly, in certain
cultures, more people tend to be outdoors at full moon, though not
necessarily at periods near full moon, and the adverse impact of such
differential exposure would be less
when looking at the seasonality of lunar phases, rather than simply at
full versus no full moon.
To address these two issues simultaneously, we have conducted a more
appropriate analysis, and with a new from injury cases inflicted by dog
bites, and registered in EDISS, Greece. A total of 2642 such cases had
occurred within the window of earliest to latest full moon between
1/1/1996 and 12/31/1999. For each case, we obtained the gap time between
the day of bite injury and the day of the
immediately preceding full moon. We classified these gap times, which
ranged from 0 to 29 days, to 10 periods, in analogy to the first article,
with the minor exception that each of our periods covered three days: a
bite occurring at
gap of 0 days (full moon), 1, or 29 days was labeled period 5; gaps of 26-
28 days were assigned period 4; and so on for the other periods. Finally,
we cross-classified the cases into the two-way table of moon period (1-
10), by the
weekday at which the bite occurred. We analyzed the data with a Poisson
regression allowing for possible simultaneous (i) effect of weekday, and
(ii) effect of seasonality of the 10 moon periods on bite number, whereby
we estimated the time (which period) of peak and the intensity of the peak
seasonality (procedures and data are available upon request).
The results showed evidence for seasonality, with a relative risk of
1.13 for highest versus lowest risk of bite (P=0.028; 95% CI: [1.01,
1.26]). However, the timing of this highest incidence adjusting for
weekday effect, was estimated at period 1 (17-19 days), with 95% CI:
periods [10,1,2], i.e., [14-22] days after full moon, thus excluding
periods at or near full moon.
Although additional unknown factors may underly the discerned
seasonality pattern during days 14-22, this is always possible in
observational studies, and informed judgment is based on the existing
evidence after known concerns are suitably addressed. The evidence based
on our results is that there is no increases risk of bites at or near full
1. Bhattacharjee C, Bradley P, Smith M, Scally AJ, Wilson BJ. Do
animals bite more during a full moon?
Retrospective observational analysis. Br Med J 2000; 321:1559-61.
2. Chapman S, Morrell S. Barking mad? Another lunar hypothesis bites
the dust. Br Med J 2000; 321: 1561-3.
Competing interests: No competing interests