Immune checkpoint inhibitors (ICIs) have been an encouraging treatment method in non-small cell lung cancer (NSCLC). However, bone and liver metastases are considered to restrain immunotherapy efficacy. Since serum alkaline phosphatase (ALP) is associated with bone and liver metastases, it was investigated whether serum ALP could be a novel biomarker to predict the efficacy of ICIs treatment. In the present study, 143 patients with NSCLC receiving ICIs treatment were retrospectively analyzed. The objective response rate (ORR) was compared between the ALP high and low groups, bone metastasis and non-bone metastasis groups, and liver metastasis or non-liver metastasis groups. The associations between clinical characteristics, including ALP level, bone or liver metastasis and median progression-free survival (mPFS) time were analyzed by univariate and multivariate Cox regression analysis. It was found that bone metastasis was associated with a lower ORR (24 vs. 43%; P<0.05) and shorter mPFS (10.2 vs. 17.3 months; P=0.010) in patients with NSCLC receiving ICIs. Liver metastasis was associated with lower ORR (22 vs. 38%; P<0.05), but not with mPFS (P=0.119). The ALP level was higher in patients with bone or liver metastasis than in those without (119.6 or 103.6 vs. 83.3 U/l, respectively; P<0.05). Higher ALP levels were also associated with bone or liver metastasis, lower ORR (20 vs. 39%; P<0.05) and shorter mPFS (8.5 vs. 15.4 months; P=0.009). Cox regression analysis demonstrated that ALP was an independent prognostic indicator of mPFS (hazard ratio, 1.856; 95% confidence interval, 1.030-3.343; P=0.040). In conclusion, pretreatment levels of serum ALP might be a predictive indicator of clinical outcome in patients with NSCLC after ICIs treatment.