Long haul flights and deep vein thrombosis: who is at risk?

The conclusion of the paper 'Long haul flights and deep vein thrombosis: a significant risk only when additional factors are also present' (Arya et al, 2002) is based on a flawed analysis. The study reports on 568 individuals referred for suspected thrombosis who underwent objective testing. Of these, 185 proved to have thrombosis and 383 did not. Overall, 8·7% of the patients with thrombosis and 7·6% of the patients without thrombosis had travelled by plane in the preceding 4 weeks (odds ratio 1·2). In those who travelled by plane and had an additional risk factor, thrombotic risk was increased 2·7-fold. From this analysis, it remains obscure whether travel contributed to this risk, i.e. the risk increase could simply and solely be the result of the additional risk factor. The appropriate comparison would have been between travellers and non-travellers, restricted to those with an additional risk factor. Therefore, the suggestion to use chemical thromboprophylaxis based on these data is unfounded. It is interesting that, while several studies with various designs [case–control studies (Ferrari et al, 1999; Samama, 2000; Rosendaal et al, 2001), cohort studies (Sarvesvaran, 1986; Lapostolle et al, 2001), and randomized trials of asymptomatic thrombosis (Scurr et al, 2001; Cesarone et al, 2002)] showed positive associations, two case–control studies were negative. Both the Amsterdam study (Kraaijenhagen et al, 2000) and the study by Arya et al (2002) had an unusual design (Kraaijenhagen et al, 2000; Arya et al, 2002), where referred patients who tested positive were compared with referred patients who tested negative. This design is ideally suited to rule out referral bias and recall bias (Bloemenkamp et al, 1999), under the assumption that the exposure of interest does not cause a phenocopy of the disease of interest. However, if flying leads to symptoms that make a traveller or physician think of thrombosis, e.g. leg oedema, preferential referral of travelling controls would lead to an artificially increased frequency of flying among control subjects and the odds ratio would be biased towards the null. Although it is not possible to test for this bias or referral and recall bias in the other studies, an enrichment of control subjects with a history of travel does seem likely in the study by Arya et al (2002), as 8% of the control subjects had undertaken a trip of more than 3 hours in the previous 4 weeks. This seems a high figure for the general population of Camberwell.