Background: Tolerance of AI therapy can be poor due to treatment-emergent toxicities and can lead to early discontinuation (non-persistence). Patients often switch from one AI to another when toxicities develop; however, limited prospective data exist on patients who switch AI. Here we describe the effect of switching from E to L or L to E on tolerance of and persistence with therapy. Methods: Postmenopausal women initiating AI therapy were enrolled on the ELPh trial and randomized to E or L. Those that stopped their AI for self-reported intolerance were offered crossover to alternate AI after a 2-6 week washout. Kaplan-Meier estimates of proportions on AI after 1, 3, and 6 months were assessed during 1^st and 2^nd AI. Associations between time on 2^ndAI and clinicopathologic factors were analyzed using univariable Cox proportional hazards model. To evaluate effect of crossover on patient-reported outcomes, multiple questionnaires, including a pain visual analog scale (VAS), were assessed serially. Results: 83 women, mean age 60 years, 45% prior chemotherapy, and 31% with prior tamoxifen use, participated in the crossover protocol. 71% reported improvement in symptoms a mean 4.72 weeks after discontinuing 1^st AI therapy. Median time on 1^st AI was 6.8 months (95% CI 5.8-9 months), and on 2^nd AI was 11.5 months (6.9-24.2). The probability of persistence at 1, 3, and 6 months for the 1^st AI was 94%, 76%, and 55% and for the 2^nd AI was 89%, 73%, and 62%, respectively. There was no significant association between duration on 2^nd AI and 1^st AI (L vs. E), duration on 1^st AI, age, body mass index, or prior therapies. The change in pain VAS from baseline to 1 or 3 months was not significantly different during treatment with the 1^st or 2^ndAI. Conclusions: Although all AI medications have similar mechanisms of activity, nearly two-thirds of patients who are intolerant of one AI are able to maintain therapy for at least 6 months following switch to 2^nd AI. Switching is a reasonable approach for women who cannot tolerate 1^st AI that may improve persistence with therapy. The mechanisms for intrapatient variation in tolerance warrant further study. Clinical trial information: NCT00228956