Our comparison of enoxaparin versus UFH in real-world clinical practice yielded results more favorable for enoxaparin than those from clinical trials, which have concluded equivalence between these two prophylaxis methods [3–7]. This difference in outcome may be explained in part by differences in the drug dosing used in the trials versus those we observed in clinical practice. The two trials in which enoxaparin was nominally (but not significantly) less efficacious than UFH administered dosages of 20 and 36 mg/day of enoxaparin and 15,000 IU/day of UFH, in contrast to the median dosages of 60 mg/day and 10,000 IU/day, that we observed in practice [4, 5]. Two other trials that found enoxaparin to be more efficacious (but not significantly) than UFH, used dosages of 40 mg/day for enoxaparin and 15,000 IU/day for UFH, which are consistent with current recommendations [1, 6, 7]. The reasons for the use of higher-than-recommended dosages of enoxaparin and lower-than-recommended dosages of UFH that we observed in real-world practice are unknown, but may reflect differences in ease of administration and monitoring, as well as perceptions about the relative safety of the two prophylaxis methods.
When we stratified patients by dosage of anticoagulant received we found a significant (p < 0.05) trend of fewer VTEs with higher dosing among UFH patients, but the opposite trend among enoxaparin patients. The reason for this seemingly contradictory finding is unclear and may suggest that some of the patients receiving higher dosages of enoxaparin were receiving treatment rather than prophylaxis for VTE. This is consistent with our finding that the 90 patients receiving a dosage of UFH >15,000 IU to 20,00 IU not included in our analysis had a higher rate of VTE (12%) than the patients receiving dosages between 5,000–15,000 IU included in the study. In any case, the results of the stratified analyses suggest that some cases of VTE among UFH patients may be related to inadequate UFH dosing. However in sensitivity analyses that compared patients receiving the recommended dosages of 40 mg/day of enoxaparin and 15,000 IU/day of UFH, the occurrence of VTE among enoxaparin patients remained significantly lower than that among UFH patients.
We similarly stratified patients by duration of prophylaxis and found a significant trend of fewer VTEs with longer duration of thromboprophylaxis. Because patients receiving UFH prophylaxis tended to have fewer days of prophylaxis than those receiving enoxaparin, some cases of VTE among UFH patients may be related to shorter duration or prophylaxis. We note, however, that the occurrence of VTE was lower among enoxaparin versus UFH patients at all levels of treatment duration.
Despite the significantly lower overall risk of VTE among enoxaparin versus UFH patients, we observed no economic benefit of thromboprophylaxis with enoxaparin versus UFH in terms of reduction in either length of hospital stay or total inpatient costs. When pooled across the two treatment groups, patients with VTE did have significantly longer inpatient stays than those without (14.8 days vs. 9.9 days; p = 0.001) and correspondingly higher inpatient costs ($26,605 vs. $17,241; p = 0.003); however, these differences did not translate into overall lower costs for enoxaparin patients. This lack of apparent economic benefit for enoxaparin prophylaxis versus UFH prophylaxis may be due to the relatively small number of patients experiencing VTE, our inability to isolate costs directly related to VTE from the total costs of hospitalization, and the high degree of variability in both length of stay (ranging from 6 to 54 days) and inpatient costs (from $3,000 to $384,000) among the patients in this sample.
Our study is subject to the limitations inherent in the use of retrospective administrative data. First, we identified and classified study subjects based on drug records and ICD-9-CM coding of diagnoses and procedures taken from the inpatient record. If these records were inaccurate or incomplete, subjects may have been misidentified as eligible, or misclassified by diagnosis or receipt of prophylaxis. Similarly, the incidence of VTE was derived from information recorded by the treating institutions; therefore, the validity of our findings depends on the accuracy of their record keeping and on our interpretation of the data. Drug records were used both to classify patients receiving prophylaxis, and to identify patients experiencing VTE, and assumptions regarding treatment patterns and dosing were necessary. Sensitivity analyses demonstrated, however, that our results were not highly dependent on these assumptions, and our conclusions were similar when we used narrower definitions of prophylaxis and treatment.
We further note that because VTE was identified from administrative records, it was not possible to distinguish between symptomatic and asymptomatic VTEs. Attempts to identify patients receiving diagnostic testing for VTE suggested low levels of reporting for these tests, and available data did not allow us to determine the outcome of the test except as inferred from treatment received. In addition, because of the limited information available for the bleeds identified from ICD-9 diagnostic coding, we cannot be certain that these outcomes were caused by administration of either study drug.
Because patients were not randomly assigned to enoxaparin versus UFH prophylaxis, underlying differences in the two groups may have influenced both the selection of thromboprophylaxis and the occurrence of VTE. We note, however, that for our results to have been produced solely by "confounding by indication", treating physicians would have to systematically administer enoxaparin prophylaxis to patients at lower risk of VTE than those receiving UFH prophylaxis. In fact, the enoxaparin patients in our sample had nominally higher frequencies of nearly every risk factor for VTE examined. Moreover, using propensity score techniques to control for confounding, we found that enoxaparin patients had lower risks of VTE than UFH patients across all strata. Nevertheless, the extent of uncontrolled confounding in this study remains unknown.
Finally, in selecting patients for this study, we excluded groups such as patients aged <40 years, pregnant women, and patients with certain comorbid conditions; therefore, extrapolation of our finding to these populations may not be appropriate.