Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
The findings of our systematic review (see page 277) showed that ergonomic interventions are not effective for preventing or reducing low back pain (LBP) and neck pain among non-sick listed workers. In this systematic review only randomised controlled trials (RCTs) were included, but Westgaard (see page 217) questions whether study designs other than RCTs (eg, quasi-experimental and qualitative studies) would be also suitable for evaluating the effectiveness of ergonomic interventions in the workplace. For a long time, the conduct of a systematic review on RCTs only was not possible because RCTs on ergonomic interventions were lacking. Therefore, reviews also included study designs that were suspicious for bias (ie, pre–post trials, prospective cohort studies, controlled trials and quasi-experimental trials).1 2 However, in recent years, more and more information from RCTs on ergonomic interventions has become available and this allowed us to conduct our systematic review. Although we agree that other study designs can add to the existing knowledge on ergonomic interventions, we believe that the RCT is the gold standard for investigating the effectiveness of different interventions untainted by bias.3
In his commentary, Westgaard points out that although the purpose of the RCT is to control for most unforeseen factors, interventions conducted in complex environments may be affected by organisational changes, financial problems, lack of management support or other issues and, as a result, study results may be influenced. In our opinion these factors could hamper evaluation of the potential effects of ergonomic interventions in all types of studies except for those carried out in laboratory settings. These unforeseen factors are in fact an inevitable part of applying ergonomic interventions in real (working) life. A possible solution to reduce the influences of these factors is to perform cluster randomisation at the level of the workplace (department or working unit). Similarly to individual randomised trials, the cluster randomised trial also minimises the risk of bias. Moreover, by performing a cluster randomisation, contamination between workers in the intervention group and those in the control group is avoided.4
We strongly support the opinion of Westgaard that researchers should conduct a process evaluation alongside their RCT. Not only can process evaluation help researchers to understand unexpected study results,5 but it can also shed light on whether the intervention was delivered as intended and resulted in the implementation and use of ergonomic measures (ie, implementation, compliance, satisfactions and experiences) and on the successes and failures of the intervention.6 We found that implementation of ergonomic measures was poorly reported in ergonomic intervention studies, while the effectiveness of ergonomic interventions is strongly determined by its end-users. To improve compliance, future ergonomic interventions should use an adequate implementation strategy.7 Furthermore, researchers should improve reporting on compliance.
Westgaard questions whether the performance of our meta-analysis was appropriate. In our opinion, a meta-analysis was possible because the studies used similar questionnaires to measure incidence/prevalence. Furthermore, only physical ergonomic interventions (interventions aimed at redesigning the workplace) with more or less similar interventions were pooled, for example, the provision of a new mouse, new kitchen equipment or arm supports, and adjustments to desk heights. Moreover, the I2 (a measure that quantifies inconsistency across studies) supported the performance of pooling in a meta-analysis.8
We agree with Westgaard's final point that the conclusions of our review have to be interpreted with caution. First of all, we found limited studies per outcome measure and secondly, most studies were conducted in an office setting and study populations consisted of both symptomatic and non-symptomatic workers. Therefore, our results cannot be generalised to the whole population.