Prevalence and incidence of work-related musculoskeletal disorders in secondary industries of 21st century Europe: a systematic review and meta-analysis

The PubMed, Web of Science, ScienceDirect, Cochrane library and Scopus databases, were searched for eligible articles. The final search of the databases was performed on the eight of March 2021. Three authors (RG, BT, JG) developed the search strategy in accordance with the PECO framework [15] that comprised key-words related to prevalence and incidence numbers (epidemiology OR incidence OR prevalence), industrial work (industry OR industrial worker OR industrial work OR industrial workplace OR industrial task) and musculoskeletal disorders (musculoskeletal pain OR occupational injury OR injury OR musculoskeletal disorder OR musculoskeletal disease OR musculoskeletal complaint OR musculoskeletal pain OR cumulative trauma disorders). No filters were added with the exception of publication year set from 2000 to 2021 to only include articles that researched WMSDs in the twenty-first century. In addition, reference lists of studies included in this review were screened for relevant articles not provided by the initial search strategy. Detailed descriptions for each database are displayed in supplementary figure A.

Studies were included if they (i) provided prevalence or incidence data, defined in accordance with the definitions of incidence and prevalence provided by the Centers for Disease Control and Prevention (CDC) [16], (ii) focused on manual work in secondary industry (manufacturing), (iii) included countries of the European Union (EU-28), (iv) reported WMSDs that corresponded to an anatomical region (e.g. neck, back, hip, etc.), (v) adopted an observational study design (cross-sectional, cohort, or health surveys), (vi) used validated or non-validated questionnaires and (vii) were published in peer-reviewed journals. Studies were excluded when they (i) did not provide an overall number of WMSDs in secondary industry, (ii) did not make a clear distinction in manual or administrative workers for reported WMSD data, (iii) failed to clearly differentiate between industry sector, (iv) presented prevalence/incidence numbers based on claims or hospital records and (v) were published before the year 2000.

Databases were searched by one author (RG) and articles were imported in the Rayyan web application for duplicate removal and screening [17]. First, articles were screened on title and abstract by one author according to above mentioned and predefined eligibility criteria. Next, remaining articles were screened on full text by two independent researchers (RG and BT). Disagreements were solved through discussion. If consensus could not be accomplished, a third researcher (JG) would take part in the process to make inclusion by majority possible. When full texts could not be found or data of interest for this meta-analysis was missing, authors would be contacted through e-mail to request full text or data.

One author (RG) extracted the following data of included full texts to answer the research question: study design, type of industry, period of measurement, response rate, demographic characteristics of included participants (age and gender), tools for examination of WMSDs, examiner (e.g. self-report or occupational physician), type of WMSDs (e.g. neck pain or shoulder pain), prevalence and incidence data per WMSD. Occupational physicians were selected as representative for the examiner variable when options between different healthcare professionals were given. Furthermore, self-reporting prevalence data took precedence over physical examination data when studies reported both. Results will be subdivided in a Prevalence section containing a qualitative and quantitative analysis, and an Incidence section limited to a quantitative analysis.

Risk of bias was assessed in accordance with Hoy et al. [18]. They used a risk of bias tool specifically developed for prevalence studies. Risk of bias was verified through ten questions that can be answered with “high risk” or “low risk”. Questions one to four assess the selection and nonresponse bias (external validity), five to nine the measurement bias and ten the bias related to the analysis. When articles provided insufficient information to answer a question a “high risk”- score was assigned to that item. Risk of bias was analyzed by two independent researchers (RG and JG). Consensus was established through discussion.

In order to perform a meta-analysis, studies were included if they reported (i) sufficient demographic information regarding sample size and (ii) data for specific anatomical locations i.e. neck, shoulder, elbow, wrist/hand, back or leg WMSDs. Twelve subgroups of WMSDs were formulated: neck, shoulder, shoulder/neck, elbow, wrist/hand, upper back, lower back, back (for studies that did not make a distinction between upper or lower back), hip, knee, ankle/feet, leg (for studies that did not make a distinction between hip, knee or ankle/feet). Pooling of data occurred in several stages. First, overall prevalence percentages or incidence ratios were calculated for studies reporting data related to subgroups of the investigated sample size i.e. age, gender, skill level (e.g. unskilled versus skilled), workload (e.g. low or high) or type of manual workers (e.g. welder, metal worker, other manual workers). Overall prevalence percentages were calculated using following formula where “p” corresponds to prevalence, “n” to sample size and “n_tot” to the sum of all sample sizes: [(p₁ + … + p_x) (n₁ + … + n_x)]/n_tot. Overall incidence ratios were calculated using following formula where “i” corresponds to incidence cases, “n” to sample size and “y” to the persons-years at risk: (i₁ + … + i_x)/ [(n₁ + … + n_x) (y₁ + … + y_x)]. In order to minimize heterogeneity, no overall prevalence percentages or incidence ratios were calculated for studies that included prevalence periods (e.g. 12-month or 7-day prevalence). Second, standard errors were calculated for each prevalence rate or incidence ratio using following formula with “p” corresponding to either prevalence rate or incidence ratio and “n” to the sample size: sqrt [p (1-p) / n)]. All calculations were performed in Microsoft Excel (version 2002). Third, mean prevalence and incidence values with associated 95% confidence intervals and heterogeneity (I² statistics) per WMSD-subgroup were calculated through the random-effects model of the R software program (version R-4.0.2). I² statistics displays the variation across included studies that is due to heterogeneity rather than chance [19].

A total of 4371 articles were retrieved. After removing duplicates, 3509 articles were subsequently screened by title and abstract. The remaining 88 articles were evaluated on full text as well as six additional articles, obtained from consulting the references of included articles. A total of 35 studies were included in the qualitative analysis of which 24 authors [20‐43] discussed prevalence of WMSDs, seven authors [44‐50] researched incidence of WMSDs and four authors [51‐54] reported both incidence and prevalence rates. Figure 1 displays the results of the screening process in more detail. Due to the limited amount of incidence reports, a meta-analysis could only be performed for studies that reported prevalence data and complied with the predetermined eligibility criteria for quantitative analysis (n = 17).

Risk of bias assessment resulted in 12 low risk scored studies [20, 22, 25, 31, 32, 34, 39‐42, 53, 54], 20 studies with moderate risk scores [21, 23, 24, 26, 28‐30, 33, 35‐38, 43, 45‐49, 51, 52] and 3 studies that received high risk scores [27, 44, 50].

Overall back WMSDs obtained the highest prevalence values of all subgroups (60%). This finding is corroborated by the European Agency for Safety and Health at Work (EU-OSHA) with a 12-month prevalence value of 55% reported in industrial workers (i.e. plant and machine operators, assemblers), indicating the significant susceptibility of this population for developing back WMSDs [1]. Further, lower back WMSDs were more frequently reported (47%) compared to upper back WMSDs (22%). This corresponds to values reported in the systematic review of Briggs et al. [56] where a 12-month prevalence range between 3 to 55% and medians around 30% were obtained for upper back WMSDs in most occupational groups. This difference between prevalence of lower and upper back WMSDs may explain the growing interest in lower back preventive measures since this anatomical location is clearly more prone to develop WMSDs.

It is assumed that the presence of biomechanical risk factors (e.g. lifting heavy loads or performing repetitive task) and psychosocial risk factors (e.g. low job control or level of support from colleagues) increases the development of WMSDs [2, 57]. This correlation was found in the majority of studies [20, 35, 37, 42]. Therefore, to effectively prevent this type of WMSDs, implementation of robotic devices e.g. exoskeletons or collaborative robots could offer great potential by reducing biomechanical loads. Further, studies implementing the Nordic musculoskeletal questionnaire (NMQ) reported higher prevalence values in the lower and upper back subgroup [22, 26, 31, 35, 37, 42, 43]. It is possible that this examining method provided a more thorough self-evaluation than non-validated questionnaires. However, this was not the case for the overall back WMSDs subgroup. Since other factors could have influenced these results (e.g. characteristics of sample size, presence of risk factors, etc.) and high heterogeneity was obtained for all back WMSDs subgroups, it is recommended to interpret these observations with caution.

Regarding incidence of back WMSDs, Häkkänen et al. [49] reported incidence ratios of 0.19 for males (indicating that over the course of 1 year 19 out of 100 persons reported new back WMSDs) and 0.24 for females working in the transport industry. This indicates a relative limited development of new back WMSDs between 1987 and 1990. A European report of 2005 described relatively similar 1-year incidence ratios of low back WMSDs that ranged between 0.12 and 0.29 for Austria and 0.28 for the Czech Republic [58]. These relatively low incidence values and high prevalence values indicate a limited onset of back WMSDs with long-term complaints indicating the need for effective prevention strategies. Nevertheless, studies researching incidence of (lower) back WMSDs in secondary industries of twenty-first century Europe are scarce and results should be interpreted with caution.

The neck, shoulder, and neck/shoulder subgroup obtained the highest prevalence values of upper limb WMSDs with 51, 50 and 54%, respectively. These findings are in line with the work of Buckle and Devereux [59]. No clear trends in prevalence data could be observed in studies reporting high psychosocial stress or the presence of biomechanical risk factors. Although the exact reason for this observation remains speculative, it would not be surprising to find the cause in the known high heterogeneity. Also no discrepancies were found between studies that obtained prevalence data through validated questionnaires, non-validated questionnaires or expert assessment. An exception was the neck subgroup where the NMQ tends to result in higher prevalence values. It is possible that the thoroughness of this questionnaire led to this observation. However, since only one study in this subgroup refrained from using the NMQ [21], under-representation of other questionnaires could also have attributed to finding this trend. The need for effective prevention strategies is again highlighted in these subgroups. Implementation of new robotic devices to optimise employee’s ergonomics and therefore reduce risks for developing WMSDs form an extremely promising strategy to combat WMSDs in the future [12, 13].

Incidence of upper limb WMSDs was researched the most. Between 1987 and 1990, incidence ratios of 0.13 were reported (indicating that over the course of 1 year 13 out of 100 persons reported new upper limb WMSDs) [49]. Later from 1996 to 1997 and from 1997 to 2000, incidence ratios of 0.26 and 0.8 were obtained, respectively [48]. These results carefully suggest that the occurrence of new upper limb WMSDs decreases over time in secondary industries. However, some caution when interpreting these results is advised since no other studies were included to support these statements. Monaco et al. [50] reported an incidence ratio of 0.04 between 2012 and 2015, implying a further decline of upper limb WMSDs. This is contradicted by data provided by EU-OSHA stating that a significant reduction in upper limb WMSDs incidence data is still absent over the beginning course of the twenty-first century [58]. Lower incidence ratios are consistently reported compared to their prevalence values, indicating that - although the onset of new WMSDs is relatively limited - employees experience long-term chronic disorders, leading high prevalence numbers. This hypothesis is strengthened by the finding of Gold et al. [60] who states that employees affected by upper limb WMSDs were still experiencing the same disorder at least 1 year later.

More specific examples of chronic WMSDs were research by Leclerc et al. [52] who reported incidence ratios of 0.12, 0.13 and 0.06 for carpal tunnel syndrome, lateral epicondylitis and wrist tendinopathy, respectively. In accordance with Leclerc et al. [52], other included studies found similar incidence ratios of 0.14 and 0.12 for carpal tunnel syndrome, 0.02 for lateral epicondylitis and 0.03 for wrist tendinopathy [53, 54]. Notably, incidence values for these specific WMSDs tend to have higher consensus compared to other WMSDs. Although not similar in all variables, both Leclerc et al. [52] and Roquelaure et al. [54] investigated WMSDs during the same period (between 1996 and 1997) and worked with occupational physicians to obtain incidence data. In addition, carpal tunnel syndrome and lateral epicondylitis are two well-defined diagnoses compared to vague WMSD complaints in a nondelineated anatomical area. This clearly emphasizes the need for more precise definitions of WMSDs as well as standardized study procedures. In contrast to upper limb incidence data, there is a noticeable lack of information regarding neck WMSDs of industrial workers of Europe.

Lower limb WMSDs were less prevalent than back or upper limb WMSDs with 29, 11, 33, 17% obtained for the leg, hip, knee, ankle/feet WMSDs subgroups, respectively. EU-OSHA reported a similar trend with a 12-month prevalence of 29 and 30% lower limb WMSDs in 2010 and 2015, respectively [58]. It is possible that there are fewer risk factors present in an industrial setting for developing lower limb WMSDs compared to risk factors for upper limb WMSDs [2]. However, it is noteworthy that epidemiological research or risk assessments regarding lower limbs is generally under-represented which could mediate this discrepancy. Some contradicting trends in prevalence of WMSDs were found in lower limb subgroups regarding the examination method and the presence of biomechanical and/or psychosocial risk factors. Since prevalence values of WMSDs are not only influenced by abovementioned factors, it is plausible that the high heterogeneity, contributed to this absence of homogenous trends in included subgroups.

Three studies reported incidence rates of lower limb WMSDs, however overall incidence ratios could not be calculated due to the lack of sample size information [44‐46]. Chen et al. [44] reported incidence ratios of 0.0121 (indicating that over the course of 1 year 1 out of 100 persons reported new lower limb WMSDs), 0.000049 and 0.0317 in the food, chemical and metallic industry, respectively. An incidence ratio of 0.000028 was obtained by cherry et al. [45] in the metallic industry. Substantially less information regarding lower limb incidence values is available in European reports and scientific literature to support this research. However, the trend of relatively low incidence values and higher prevalence values again indicate that recurrence of lower limb WMSDs is common among industrial workers. Insufficient time to heal or rehabilitation could contribute to these long-term disorders. Strategies that temporarily relieve the employee from hard labour and effective rehabilitation could benefit the healing process and limit recurrence of WMSDs.

It is evident that with the high obtained heterogeneity in the majority of subgroups, the included studies did not utilise uniformity in methodology and reporting strategies. Often, lack of information or discrepancies regarding demographic characteristics of included sample sizes, performed tasks, examining methods and examining periods challenged pooling and interpretation of obtained results. This resulted for the majority of subgroups in a limited number of included studies and considerable heterogeneity that could not be further investigated and challenges the interpretation of obtained results. In order to improve uniformity and generalisation of results, a criteria document for physical examination and standardized questionnaires (e.g. the NMQ) could be used. Further, assessment of risk of bias was performed with the tool as described by Hoy et al. [18] which was the most applicable for included studies. Kane and Shamliyan [61] previously argued that uniformity of this assessment tool is challenged by unclearly defined constructs. Therefore, guidelines for each item of the assessment tool were subjectively expanded and adjusted to this research’s specific subject through discussion between authors. This self-interpreted variant of the initial definitions provided by Hoy et al. [18] could have influenced the risk assessment scores. Finally, only a limited number of studies reporting incidence values could be included due to scarcity in literature. This inhibited profound quantitative calculations of incidence ratios and highlights the need for more qualitative incidence research. Nevertheless, besides the high variability, the obtained high prevalence values, especially in food industries, indicate the need for effective rehabilitation and prevention strategies tailored to the unique characteristics of each industry type.