Pet birds and lung cancer
BMJ 1996; 313 doi: https://doi.org/10.1136/bmj.313.7067.1218 (Published 16 November 1996) Cite this as: BMJ 1996;313:1218- John Britton,
- Sarah Lewis
- Reader in respiratory medicine Statistician University of Nottingham, Division of Respiratory Medicine, City Hospital, Nottingham NG5 1PB
Now no evidence of a link
The question of whether keeping pet birds increases the risk of lung cancer was first raised by Holst.1 2 Having noted a higher occurrence of lung cancer among bird owners in a Dutch general practice population, he and colleagues carried out a case control study comparing hospital patients with community matched controls and showed a 6.7-fold increase in risk.3 Two subsequent studies produced further evidence in support of this observation: one reported a twofold increase in risk of lung cancer in relation to exposure to pet birds in a German population,4 while the other, from Scotland, showed no significant association with exposure to pet birds in general but a 3.5-fold increase in risk of lung cancer in those who kept pigeons.5
At the time, we argued that, although of great importance if valid, these observations might have arisen from residual confounding by cigarette smoking.6 All of the case-control studies had controlled for smoking in their analysis but categorised cases and controls as smokers or non-smokers; the degree of smoking was not analysed quantitatively. In the ensuing correspondence, the authors of the two more recent reports argued that residual confounding was unlikely to have been a major influence because smoking histories seemed to have been similar in cases and controls.7 8 Others pointed out further methodological and confounding factors that might also have contributed to false positive results.9 10 In the end it was probably fair to conclude that the relation between bird keeping and lung cancer needed to be assessed further, in studies with sufficient power and appropriate measures of confounding variables to separate out the independent effects involved.
Two such studies, one from Missouri and the other from Sweden, are published in this issue of the BMJ (pp 1233, 1236).11 12 Like their predecessors, both are case-control studies, but their strengths are that, with 652 and 380 cases respectively, they are considerably more powerful; they have ascertained incident cases only and thus avoided potential survival bias in relation to bird keeping in prevalent cases; they have measured smoking quantitatively and used this quantitative information in their analysis; and they have controlled for confounding by diet and socioeconomic status. They found no evidence whatever of an increase in risk of lung cancer in relation to any measure of bird keeping or exposure. Indeed, in some of the subgroups of exposure to pet or farm birds the risk of lung cancer was significantly reduced. The Missouri study is limited to women only,11 but the Swedish study includes both sexes and shows no evidence of a gender difference in effect.12 While there are inevitable potential criticisms of both studies, which the authors acknowledge, in our view none of these is likely to have resulted in odds ratios so close to unity in both reports. We therefore accept the authors' conclusion that keeping domestic birds does not seem to be associated with any excess risk of lung cancer.
So what is the explanation for the discrepancy between these new findings and the previous evidence? We argue, as before, that residual confounding by smoking is highly likely to have distorted the risk estimates in previous studies. In the new Swedish study the estimated odds ratios for lung cancer in current smokers of 20 cigarettes a day were about 32 and 43 respectively for men and women, ratios which far exceed the estimated effect of keeping pet birds in all earlier investigations. In the presence of effects so strong, the potential for bias in estimating the influence of other exposures even loosely related to smoking is substantial. An alternative explanation advanced by the authors is that exposure to pet birds in the populations they have studied is somehow different in intensity or in terms of associated exposures such as infections, infestations, or other factors related to bird keeping in different countries.
Another potential explanation for the discrepancy is that neither of the two new studies has looked specifically at pigeon keeping, which was the only significant association with lung cancer in the study from Scotland.5 In Britain at least there is likely to be a substantial difference in the number of birds kept, and consequent degree of exposure, between those keeping pigeons for racing and those who keep one or two birds as indoor domestic pets. It is perhaps still possible that the previously reported associations relate to an underlying effect that is specific to pigeon keeping.
To resolve these uncertainties it would be necessary to repeat these studies yet again in the populations that gave rise to the original observations of an association, taking particular care to deal with confounding effects. However, these two new studies seem to be conclusive in two respects. First, they provide further evidence that, irrespective of any perceived misconception,13 cigarette smoking remains by far the single strongest and most commonly encountered avoidable cause of lung cancer. Secondly, and importantly for vast numbers of people, they show that keeping domestic pet birds such as budgerigars, canaries, and parrots does not seem to be associated with an increased risk of lung cancer. The question of whether heavier exposure to pigeons carries an increased risk remains unresolved.