Validation of the Khorana score for predicting venous thromboembolism in 40 218 patients with cancer initiating chemotherapy

The Khorana score is recommended for guiding primary venous thromboembolism (VTE) prophylaxis in cancer patients, but its clinical utility overall and across cancer types remains debatable. Also, some previous validation studies have ignored the competing risk of death, hereby potentially overestimating VTE risk. We identified ambulatory cancer patients initiating chemotherapy without other indications for anticoagulation using Danish health registries and estimated 6-month cumulative incidence of VTE stratified by Khorana levels. Analyses were conducted with and without considering death as a competing risk using the Kaplan-Meier method vs the cumulative incidence function. Analyses were performed overall and stratified by cancer types. Of 40 218 patients, 35.4% were categorized by Khorana as low risk (score 0), 53.6% as intermediate risk (score 1 to 2), and 10.9% as high risk (score >= 3). Considering competing risk of death, the corresponding 6-month risks of VTE were 1.5% (95% confidence interval [CI], 1.3-1.7), 2.8% (95% CI, 2.6-3.1), and 4.1% (95% CI, 3.5-4.7), respectively. Among patients recommended anticoagulation by guidelines (Khorana score >_2), the 6-month risk was 3.6% (95% CI, 3.3-3.9). Kaplan-Meier analysis overestimated incidence up to 23% compared with competing risk analyses. Using the guideline-recommended threshold of >= 2, the Khorana score did not risk-stratify patients with hepatobiliary or pancreatic cancer, lung cancer, and gynecologic cancer. In conclusion, the Khorana score was able to stratify ambulatory cancer patients according to the risk of VTE, but not for all cancer types. Absolute risks varied by methodology but were lower than in key randomized trials. Thus, although certain limitations with outcome identification using administrative registries apply, the absolute benefit of implementing routine primary thromboprophylaxis in an unselected cancer population may be smaller than seen in randomized trials.