A meta-analysis of studies on intracranial tumors and mobile phone use published by the end of 2012 was performed to evaluate the overall consistency of findings, assess the sensitivity of results to changes in the dataset, and try to detect the sources of between-study heterogeneity. Twenty-nine papers met our inclusion criteria. These papers reported on 47 eligible studies (17 on glioma, 15 on meningioma, 15 on acoustic neuroma), consisting of either primary investigations or pooled analyses. Five combinations of non-overlapping studies per outcome were identified. The combined relative risks (cRRs) in long-term mobile phone users (≥10 years) ranged between 0.98 (0.75-1.28) and 1.11 (0.86-1.44) for meningioma, with little heterogeneity across studies. High heterogeneity was detected across estimates of glioma and acoustic neuroma risk in long term users, with cRRs ranging between 1.19 (95% CI 0.86-1.64) and 1.40 (0.96-2.04), and from 1.14 (0.65-1.99) to 1.33 (0.65-2.73), respectively. A meta-regression of primary studies showed that the methodological differences embedded in the variable "study-group" explained most of the overall heterogeneity in results. Summary risk estimates based on heterogeneous findings should not be over-interpreted. Overall, the results of our study detract from the hypothesis that mobile phone use affects the occurrence of intracranial tumors. However, reproducibility (or lack of) is just one clue in the critical appraisal of epidemiological evidence. Based on other considerations, such as the limited knowledge currently available on risk beyond 15 years from first exposure, or following mobile phone use started in childhood, the pursuance of epidemiological surveillance is warranted.