PURPOSE: To determine the optimal criteria for referral to adjuvant chemotherapy after bladder cancer surgery (cystectomy). METHODS: Adjuvant chemotherapy is commonly given after cystectomy. The standard decision rule is that patients with stage pT3, pT4 or positive lymph nodes receive chemotherapy. Recently, a multivariable prediction model ("bladder nomogram") has been published that predicts recurrence on the basis of several clinical and pathological characteristics. We wished to determine whether use of the nomogram to determine eligibility for chemotherapy would lead to better clinical results than the current standard. First, we defined three rules for chemotherapy based on nomogram cut-offs of 10%, 25% and 70%. We then used 4,462 patients from the bladder nomogram data set to estimate: a) the proportion of patients who meet each rule, and who would therefore receive chemotherapy; b) the proportion of patients who would recur, calculated by adjusting the risk of patients who recurred, but would have received chemotherapy. The number of treatments and number of recurrences were combined using a simple net benefit formula: reduction in recurrence rate with treatment minus number of treatments / NNTt, where NNTt is the threshold number of patients who would be treated to prevent one recurrence. A two-way sensitivity analysis was conducted varying both the effectiveness of adjuvant therapy (relative risk between 0.50 - 0.99) and its toxicity (NNTt 1 – 250). RESULTS: The proportion of patients eligible for adjuvant chemotherapy using the standard criteria, 10%, 25% and 70% nomogram cut-offs were 55%, 91%, 41% and 8% respectively. For adjuvant therapy associated with a 80% relative risk, the reduction in event rate would be 4.4%, 5.5%, 4.0% and 1.3%. Comparing the 25% cut-off to the standard rule, for example, the large decrease in the number of treatments compared to the standard decision rule (14%) outweighs the small increase in recurrences (0.4%) for NNTt up to 35; at high NNTt, however, the larger reduction in recurrences for the 10% cut-off makes this the optimal strategy. One or other nomogram cut-offs had superior net benefit to the standard model at all plausible values for adjuvant therapy toxicity and effectiveness. CONCLUSIONS: Use of a multivariable model to determine eligibility for chemotherapy will lead to better clinical results than use of a standard decision rule based on pathological stage.