A total of 194 English-language publications were initially identified, of which 49 were selected after scanning study titles and excluding comments, letters, editorials, case series and reports, and reviews. After reviewing the abstracts, a further 32 were excluded because they were either interventional or descriptive articles not estimating the risk of travel exposure, or because they estimated risk specifically for a high-risk population. Of the remaining 17 studies, two did not estimate the risk of travel exposure but were, however, retained for review for their relevance and recent publication, instead of for a meta-analysis of risk estimates.[9,11] Another was excluded because both cases and controls were similarly exposed to travel.
Ultimately, nine case-control, two prospective controlled, and five other observational studies were selected for data abstraction and critical appraisal ( Table 1 ). One case control study was included in another pooled risk estimate of case-control observations,[13,14] and another earlier prospective controlled was a pilot study for the larger later study.[15,16]
Excluding the study included in a pooled estimate by ten Wolde and colleagues,[13,14] the case-control collection included a total of 2,486 patients with VTE and 4,543 controls. Not known for two studies,[7,17] the minimum mean or median age of patients who had recently traveled in these studies was a median of 40 years.
Six of nine case-control studies examined exposure to any mode of travel and associated venous thromboembolic events ( Table 2 ).[1,13,14,17,18,19] One did not define 'long distance' travel, and two restricted the studies to air-travel exposure only.[7,20] By definition, one of them would have ignored unofficial/private air travel in cases and controls. Overall mean/median travel time was reported in or could be obtained from four studies,[1,14,18,20] and it ranged from a median of 5 (interquartile range 4.5-9 h) to 7 hours (interquartile range 4-10 h).[14,18] Mean/median time from the end of travel to onset/presentation of VTE (in cases) was reported in three studies: Ferrari and colleagues, mean 13 days (range 4-22 d); Hosoi and colleagues, median 1 day (interquartile range 0-4 d); and Kraaijenhagen and colleagues, mean 5 days (range 2-13 d).
Five studies considered PE along with deep vein thrombosis (DVT) as case definitions.[1,7,14,17,20] However, in one study PE cases were grouped not only with cases of DVT but also with superficial thrombophlebitis as 'case definition 1' (see Table 2 ). Data corresponding to 'case definition 2' from that study, that is DVT of lower legs, is therefore included in this study.
Four studies employed VTE suspects as controls in whom VTE was later ruled out.[13,14,18,19] Except for two studies that did not describe diagnostic techniques,[7,17] and that of Ferrari and colleagues, which did not state how PE was diagnosed, VTE was diagnosed objectively. However, only two studies explicitly stated examining for compressibility of veins in diagnosis of DVT with ultrasonography.[13,14]
In all, five case-control studies concluded that travel is a risk factor for DVT and/or PE based on odds ratio (see Table 2 ).[1,2,17,18,20] In one of these studies, the risk was not significant and was attributed to those with established preexisting VTE risk factors. That study employed a DVT suspect control group in whom DVT was ruled out. These positive studies produced less precise risk estimates with larger confidence intervals compared with negative studies (see Table 2 ). However, the latter showed lower overall exposure to travel, with the exception of the study by Dimberg and colleagues ( Table 3 ).
Female gender distribution in cases versus controls was higher and unadjusted for in one of the five positive studies concluding higher VTE risk with travel (see Table 2 ). This could not be ascertained for two studies,[2,19] whereas another adjusted for gender in its analysis. Also, two of the negative studies did not adjust for a much higher female representation in control groups.[14,19] Few studies identified the location of venous thrombosis and differentiated between proximal and distal DVT (see Table 3 ).[1,2,18,20] Further, most patients who were passengers and developed VTE were probably not free of other VTE risk factors (see Table 3 ). In fact, in five studies that provide this data,[1,2,18,19,20] at least one of the established VTE risk factors was found in 25 to 87% of patients who had traveled.[1,18]
Specifically, air travel-associated risk for VTE could be obtained from five of the six studies that considered all common modes of transportation.[13,14,17,18,19] None of these particularly examined prolonged air travel exposure. Putting aside the study by Kraaijenhagen and colleagues, which was included in a pooled estimate, the absence of increased risk of VTE associated with air travel from three studies contrasted with the relatively imprecise but incriminating risk estimate from the study by Rosendaal (OR = 5.8 [2.0-16.6]).[14,17,18,19] Interestingly, negative studies included a patient-based control, whereas Rosendaal's study employed spouses of cases as controls while considering travel exposure 2 months prior to the event. Further, the prevalence of travel was low (< 15%) in the negative studies (see Table 3 ). On the other hand, the case-control analysis by Martinelli and colleagues was restricted to prolonged air travel alone, which equivocally incriminates air travel as a risk factor for VTE (see Table 2 ).
Of case-control studies, three were clearly not designed to specifically examine VTE association with travel.[2,17,18] Five studies lacked well-defined eligibility criteria.[1,2,7,17,19] Prevalent rather than incident cases were included in four studies and in one of the three case-control analyses in the pooled estimate that investigated association of PE with travel.[1,7,14,17,20] Except in three studies,[7,17,20] controls were either patients with other diseases or those in whom VTE was ruled out.[1,2,13,14,18,19] In Dimberg and colleagues' World Bank study, the study population was healthier than the general population.[7,21] Travel history from participants was obtained after diagnosis of VTE in four studies.[1,7,17,20] Five studies compared some or most of the established VTE risk factors between cases and controls,[1,2,13,19,20] but only one adjusted for these confounders besides age and gender; another adjusted for age, gender, and body mass index, and yet another for age and gender only. Only one case-control analysis was carried out based on prior sample size calculation. Although powered to detect a minimum increase in VTE relative risk of 3 associated with all air travel, it lacked power to estimate VTE odds associated with air travel > 8 hours, in particular. Other important results from these case-control studies are summarized in Table 3 .
Two prospectively controlled studies, one an earlier pilot for the later study,[15,16] examined the incidence of lower limb venous thrombosis including isolated calf muscle vein thrombosis within 48 hours of arrival on returning flights of not less than 8 hours' duration ( Table 4 ). Volunteers did not use protective measures and had no baseline thrombosis on compression ultrasound examination. Mobility in the aircraft was encouraged. A comparison was made with the external controls matched for some of the established VTE risk factors. The controls were comparably evaluated, while long distance travel in the previous 3 months was ruled out in controls in the later study. Eligibility criteria were appropriate. Communications between subjects and the sonographer were not particularly prevented. In the pilot study, frequency of a prothrombin gene mutation in passengers was eightfold that in controls; however, more controls had family history of VTE. In the other study, significantly more controls had protein S deficiency and varicose veins and were of female gender, but both groups included about 17% of participants with a past or family history of VTE. Of the eligible passengers in that study, 113 were excluded; in the control group, the number was 33. Venous thrombosis detected on second ultrasound examination was treated with anticoagulants in the larger study. No DVT was observed in the pilot study. In the second study, distal DVT was associated with other VTE risk factors, and the unadjusted relative risk for all thrombotic events was 2.8 (1.5-5.5).
Another three studies provided risk estimates for symptomatic VTE or PE associated with air travel.[8,22,23] One of these examined the change in VTE event rate over time in arriving passengers using case-crossover analysis that employed conditional probability distribution of the time to event after probabilistically linking hospital data with travel records. Others investigated the risk of symptomatic PE associated with flight duration from a review of cases among air travelers (see Table 4 ).
In the record linkage study, bias was reduced by restricting risk estimates to Australian citizens who were less likely to be lost to follow-up. However, in-flight fatal thrombotic events, death on arrival, and outpatient treatment of DVT could have been missed. A sensitivity analysis using different time periods ruled out any bias related to time trends. Further, it was shown that the Australian travelers were healthier than the general population.
The other two studies excluded lone DVT as an outcome measure.[8,23] They were limited to a short duration of observation periodduring the flight and at the airport. Further, the two studies likely missed less severe disease, in-flight/airport deaths, and late-onset events. Most patients (> 50%) in these studies had established VTE risk factors.
Two before-after prospective studies were not particularly designed to estimate risk of exposure to prolonged travel.[9,11] However, one clearly suggested an association between prolonged air travel and VTE. Since prolonged air travel was examined in passengers with only low to intermediate VTE risk factors, the number of incident cases could be an underestimate for the general population ( Table 5 ). Further, the New Zealand Air Traveller's Thrombosis (NZATT) study observing returning passengers most probably had underestimated the incidence of VTE as a substantial proportion of cautioned passengers employed VTE prophylactic measures. Underestimation was also likely as 39 passengers were lost to follow-up, some of whom might have developed VTE or succumbed to it, and also because those with milder or no symptoms were not investigated further. However, changes in factors affecting VTE after return flights were not explored. The 'business-class versus economy-class syndrome as a cause of thrombosis' (BEST) study was suitably powered to detect incidence of VTE at least as low as 3%. However, evaluation for post-flight VTE could not be completed (see Table 5 ). Although no VTE was identified, many reasons for missing thromboembolism can be ascertained, such as absence of ultrasound scans in all of those with raised post-flight D-dimers; the use of ultrasonography alone for calf DVT, for which the test is not very sensitive; and an incomplete 6-month follow-up. An understanding based on this study is all the more confusing in the light of the fact that preflight baseline D-dimer data were not available for 45% of passengers.
It should be noted that both studies employed compression ultrasonography and therefore could have produced false-negative results and missed small thrombi. In these studies, screening for eligibility based on D-dimer is considered appropriate since the test has a high negative predictive value.
J Travel Med. 2005;12(3):142-154. © 2005 International Society of Travel Medicine
Cite this: Traveler's Thrombosis: A Systematic Review - Medscape - May 01, 2005.