The authors examined the effect of demographic and clinical variables on time to treatment response in geriatric depression. One hundred and one patients, aged 60-92 years, with nonpsychotic, nonbipolar major depression were treated in an open fashion with 6 weeks of nortriptyline followed, if necessary, by 2 weeks of lithium augmentation. Univariate Cox proportional hazards analyses showed that 3 of 19 variables predicted time to response: high baseline anxiety was associated with delayed response (median of 5 weeks vs. 4 weeks for patients with low anxiety scores), whereas hospitalization for the index episode of depression and attempted suicide predicted shorter time to response. In the final multivariate Cox regression model, baseline anxiety and inpatient status were most predictive of outcome; attempted suicide did not significantly improve the predictive power of the model. Our findings strengthen existing evidence that concomitant anxiety can adversely affect the outcome of geriatric depression.