Purpose: CEREBEL compared the incidence of CNS metastases as first site of relapse in patients with human epidermal growth factor receptor 2-positive metastatic breast cancer receiving lapatinib-capecitabine or trastuzumab-capecitabine. Patients and Methods: Patients without baseline CNS metastases were randomly assigned (1:1) to receive lapatinib-capecitabine (lapatinib 1,250 mg per day; capecitabine 2,000 mg/m² per day on days 1 to 14 every 21 days) or trastuzumab-capecitabine (trastuzumab loading dose of 8 mg/kg followed by an infusion of 6 mg/kg every 3 weeks; capecitabine 2,500 mg/m² per day on days 1 to 14 every 21 days). The primary end point was incidence of CNS metastases as first site of relapse. Secondary end points included progression-free survival (PFS) and overall survival (OS). Results: The study was terminated early with 540 enrolled patients (271 received lapatinib-capecitabine, and 269 received trastuzumab-capecitabine). Incidence of CNS metastases as first site of relapse was 3% (eight of 251 patients) for lapatinib-capecitabine and 5% (12 of 250 patients) for trastuzumab-capecitabine (treatment differences, -1.6%; 95% CI, -2% to 5%; P = .360). PFS and OS were longer with trastuzumab-capecitabine versus lapatinib-capecitabine (hazard ratio [HR] for PFS, 1.30; 95% CI, 1.04 to 1.64; HR for OS, 1.34; 95% CI, 0.95 to 1.64). Serious adverse events were reported in 13% (34 of 269 patients) and 17% (45 of 267 patients) of patients in the lapatinib-capecitabine and trastuzumab-capecitabine arms, respectively. Conclusion: CEREBEL is inconclusive for the primary end point, and no difference was detected between lapatinb-capecitabine and trastuzumab-capecitabine for the incidence of CNS metastases. A better outcome was observed with trastuzumab-capecitabine in the overall population. However, lapatinib-capecitabine efficacy may have been affected by previous exposure to a trastuzumab regimen and/or when treatment was given as first- or second-line therapy in the metastatic setting.