Pembrolizumab (Pembro) in combination with chemotherapy has been approved for the treatment of pretreated advanced NSCLC in the United States and China for its significant efficacy. However, the cost-effectiveness is unknown considering Pembro's high price. The impact of programmed death ligand 1 (PD-L1) test on the cost-effectiveness is also unknown. The current study assessed the cost-effectiveness of combination therapy for nonsquamous NSCLC from the United States and China public payers' perspective. A literature-based Markov model was conducted using KEYNOTE-189 trial data to compare cost and quality-adjusted life years (QALYs) of three treatment strategies for nonsquamous NSCLC: Pembro-chemotherapy combination and chemotherapy strategy without PD-L1 test, and treatment strategy according to their PD-L1 status. In base case analysis, the combination strategy generated an additional 0.78 QALYs and 0.59 QALYs over chemotherapy in the United States and China respectively, resulting in an ICER of $132 392/QALY in the United States and $92 533/QALY in China. In the PD-L1 ≥1% base case, the ICERs were $77 754/QALY and $56 768/QALY respectively in the United States and China for PD-L1 test strategy. In the PD-L1 ≥50% base case, the ICERs were $44 731/QALY and $34 388/QALY respectively in the United States and China for PD-L1 test strategy. Lowering Pembro price can also partly decrease the ICERs. Compared with chemotherapy, the combination strategy is not cost-effective for the treatment of NSCLC in the American and Chinese health care system at WTP threshold of $100 000/QALY for the United States and $27 351/QALY for China. Using PD-L1 test for patient selection and price reduction could improve the cost-effective probabilities of immunotherapy for nonsquamous NSCLC. © 2019 The Authors. Cancer Medicine published by John Wiley & Sons Ltd.