Objective  Meta-analysis was used to systematically review the predictive effect of functional movement screen(FMS) on sports injury.

  Methods  The MIDAS module of STATA 15.0 software was used to estimate sensitivity and specificity of 21 Chinese and Enghish literatures as well as 27 groups of data including true positive, false positive, true negative and false negative values, and to aggregate the summary receiver operating characteristic(sROC), area under curve (AUC) and Fagan nomogram to estimate the diagnostic accuracy of FMS. The sources of heterogeneity were analyzed by meta-regression analysis.Deeks' funnel diagram asymmetry test was used to test publication bias.

  Results  The combined value of sensitivity is 0.40 (95% CI 0.32-0.48), and the combined effect value of specificity is 0.80 (95% CI 0.74-0.85). The P values of sensitivity and specificity Q test are less than 0.01, indicating some heterogeneity. The aggregate predictive test probability of positive/negative likelihood ratio (LR) is 50%, and the positive/negative test probability 67% and 43% respectively, the AUC 0.68 (95% CI 0.63-0.72), all indicating a general diagnostic accuracy. For sensitivity, literature type and injury risk thresholds may be sources of heterogeneity (P < 0.05). For specificity, sample size, observation period, literature quality, and study object may be sources of heterogeneity (P < 0.05). Deeks' funnel plot showed that P=0.06 < 0.1 for the 21 included literatures, indicating the possible existence of publication bias.

  Conclusion  The evidence level of correlation strength between FMS comprehensive score and subsequent injury is insufficient to support the idea that FMS comprehensive score can be directly used as a tool for sports injury prediction.