Summary Background and objective: Training in manual material handling has been the subject of many requests in various workplaces. However, according to five meta-analyses published between 2007 and 2014, the training given to manual material handlers is of questionable value, despite being widely disseminated. A reading of these analyses does not reveal why this is so, because the training programs inventoried and evaluated were described very briefly or not at all. Having more information about handling training programs would make it easier to explain the reported lack of effectiveness and, subsequently, propose avenues for improvement. This was the objective pursued by the authors of the present study. Methodology: Seventy-seven papers covered by the five meta-analyses were analyzed using 86 variables. The training programs were first categorized according to where they took place, i.e., in the workplace, in a laboratory or in a training institution. Workplace programs were described in greater detail since they were the most numerous (51 out of 77). Categories were created based on four quality criteria supported by a theoretical framework: content adapted to context, motor engagement, contextualized practice, and workplace ergonomic transformations to complement the training. The existence of a relationship between a program’s effectiveness and the extent to which it met these criteria was verified. Lastly, a hypothesis was formulated that the meta-analyses might contain a bias related to the program selection criteria. Principal results: Training programs for manual handlers differ greatly in form, using a wide variety of measures in a broad range of contexts. The content, on the other hand, is surprisingly uniform, with a consistent emphasis on learning and adopting a safe handling technique known as “straight back, bent knees.” This standardized content is part of a training approach that focuses on the learners and their behaviours, paying little attention to the individual learner’s interaction with a changing work environment and the efforts of self-regulation this requires. In most programs, training content is predetermined and exportable from one workplace to the next, despite the differences in actual working conditions. Of the four quality criteria accepted, only those related to transformations made concurrently to training and, to a lesser extent, training content adapted to the context, yielded improvements in terms of effectiveness. Ten percent of the studies met all the quality criteria. Despite the impressive number of studies devoted to the evaluation of handler training, the meta-analyses based their conclusions on a small number of them, assigning greater weight to those deemed to be of higher methodological quality, i.e., about one training program in ten. The results show that these higher-quality evaluation designs assessed the effectiveness of programs considered to be easy to evaluate, relatively simple, and generally of lower quality, which may have impacted the conclusions of the meta-analyses. In addition, the results regarding effectiveness, as reported by the studies' authors, paint a more optimistic picture than the conclusions reached by the authors of the meta-analyses. Discussion: The limitations of existing training programs are discussed and possible explanations are provided as to why they are reportedly so ineffective. It is important to emphasize that what should be questioned is not the relevance of offering training programs, but rather the type of training that is focused solely on learning and adopting safe handling techniques. These types of programs have been criticized, but arguments in their favour have also been provided. The techniques themselves should not be rejected, but should no longer constitute the focus of training. Lastly, concrete recommendations are offered to improve material handling training programs. The contradiction between the quality of the evaluation approach and the quality of the programs evaluated under this approach is also discussed. Arguments are presented about the need to develop appropriate evaluation methods to assess the effectiveness of programs considered to be more complex and hence of higher quality. Lastly, the limitations of this study and a conclusion are presented. Highlights Currently, a large majority of handling training programs consist in teaching a basic technique known as “straight back, bent knees.” Few alternatives are presented. The effectiveness of this approach is questionable, say the authors of the meta-analyses, basing their opinion on a limited number of studies whose evaluation designs were deemed to be of high quality, that is, about one in ten training programs. By comparison, the studies report positive effects from more than half the training programs examined, which contradicts the conclusions of the meta-analyses. The “straight back, bent knees” technique therefore does seem to play a certain role in accident prevention. There is a risk that the meta-analyses were biased in their selection of the study samples. Indeed, our results show that the studies using a higher-quality evaluation design expressed opinions on the effectiveness of programs deemed to be of lower quality, since they were “simpler” to evaluate. This situation reflects the limitations of evaluation designs borrowed from biomedicine (i.e., those considered to be of higher quality) when applied to more complex training given under real conditions, which is nonetheless an indication of higher quality. Handling training should be re-examined if it is to remain relevant in a comprehensive prevention program. The content should no longer focus mainly on safe techniques, but should be rounded out with other skills specific to the handler’s job context. This context should be analyzed and understood prior to the training. A skills-based approach rather than the learning of predetermined techniques is advocated. Handlers must be able to adapt to frequent and unpredictable changes in the demands of the job.