Background: Although Cochrane reviews should not include recommendations, their impact depends on usefulness in guiding clinical practice. A focus on surrogate outcomes is therefore problematic because it does not permit a tradeoff between patient-important desirable and undesirable consequences of alternative management strategies. This problem has become particularly evident in guidelines for anti-thrombotic therapy that have previously used reductions in asymptomatic venous thrombosis, measured by venography, to guide recommendations. Methods: A group of thrombosis experts and methodologists explored options for estimating the impact of prophylactic anti-thrombotic therapies on symptomatic venous thrombosis and pulmonary embolism. Results: We found that most randomized trials of anti-thrombotic prophylaxis were powered only for asymptomatic thrombosis. Trading off reduced thrombosis with increased bleeding requires estimates of patient-important events. We identified three possible strategies: i) Pooling estimates of symptomatic thrombosis from randomized trials. Limitations of this approach include few events for many questions. ii) Using estimates of the ratio of asymptomatic to symptomatic events to estimate the occurrence of the latter. Imprecision of estimates and variability in threshold for asymptomatic events across venography interpreters limits this approach. Both these first two approaches are also limited by underestimation of symptomatic events as a result of treatment of venographically discovered events. iii) Applying relative risk reductions from a composite of asymptomatic and symptomatic events to baseline risks from observational studies. Methodological limitations of observational studies, and uncertainty about whether relative risk reductions in asymptomatic events apply to symptomatic events limit this approach. When observational studies have enrolled representative patients and conducted surveillance allowing accurate estimation of symptomatic events, this last approach is probably superior. Conclusion: Venous thrombosis prevention illustrates challenges in using impact of therapy on surrogates to estimate impact on patient-important outcomes. Although every approach has limitations, this estimation is necessary to make trade-offs required to guide clinical practice.