The modified Mallampati score is used to predict difficult tracheal intubation. We have conducted a meta-analysis of published studies to evaluate the Mallampati score as a prognostic test. A total of 55 studies involving 177 088 patients were included after comprehensive electronic and manual searches. The pooled estimates from the meta-analyses were calculated based on a random-effects model and a summary receiver operating curve. Meta-regression analyses were performed to explore sources of possible heterogeneity between the studies. The summary receiver operating curve demonstrated an area under the curve of 0.75. The pooled odds ratio for a difficult intubation with a modified Mallampati score of III or IV was 5.89 [95% confidence interval (CI), 4.74-7.32]. The pooled estimates of the specificity and sensitivity were 0.91 (CI, 0.91-0.91) and 0.35 (CI, 0.34-0.36), respectively. The pooled positive and negative likelihood ratios were 4.13 (CI, 3.60-4.66) and 0.70 (CI, 0.65-0.75), respectively. The meta-analyses had statistical and clinical heterogeneity ranging from 87.2% to 99.4%. Meta-regression analyses did not identify any significant explanation of the heterogeneity. We conclude that the prognostic value of the modified Mallampati score was worse than that estimated by previous meta-analyses. Our assessment shows that the modified Mallampati score is inadequate as a stand-alone test of a difficult laryngoscopy or tracheal intubation, but it may well be a part of a multivariate model for the prediction of a difficult tracheal intubation.