Exposure to hand-arm vibrations in the workplace and the occurrence of hand-arm vibration syndrome, Dupuytren’s contracture, and hypothenar hammer syndrome: a systematic review and meta-analysis

Abstract This study provides an overview of the relationships between exposure to work-related hand-arm vibration and the occurrence of pre-defined disorders of the hands. We searched Medline, Embase, Web of Science, Cochrane Central, and PsycINFO for cross-sectional and longitudinal studies on the association between work-related vibration exposure and the occurrence of hand-arm vibration syndrome (including vibration-induced white finger), Dupuytren’s contracture, or hypothenar hammer syndrome. We used a 16-item checklist for assessing the risk of bias. We present results narratively, and we conducted random effects meta-analyses if possible. We included 10 studies with more than 24,381 participants. Our results showed statistically significant associations between the exposure to hand-arm vibrations and the occurrence of the selected disorders, with pooled odds ratios ranging between 1.35 (95% CI: 1.28 to 2.80) and 3.43 (95% CI: 2.10 to 5.59). Considerable between-study heterogeneity was observed. Our analyses show that exposure to vibrating tools at work is associated with an increased risk for the occurrence of selected disorders of the hands. Due to the majority of studies being cross-sectional, no firm conclusion is possible regarding causal relationships between vibration exposure and disorder occurrence. Future research should specifically address whether reducing exposure to hand-held vibrating tools at work reduces the incidence of the disorders of the hands investigated in this systematic review.


Introduction
Musculoskeletal disorders of the hands and wrists are highly prevalent in many occupations. There is an indication that the use of hand-held vibrating tools is related to adverse effects on the hands and wrists of workers, including hand-arm vibration syndrome (HAVS, which includes vibration-induced white finger or VWF), Dupuytren's contracture (DC), and hypothenar hammer syndrome (HHS) (Descatha et al. 2011;Nilsson et al. 2017;Mathieu et al. 2020). HAVS is a disorder of the hand affecting mostly the fingers and is composed of vascular, neurological, and muscular components (Burstr€ om et al. 1998;Ye et al. 2015). VWF presents as deteriorated peripheral blood circulation and may encompass disturbed function of nerves, soft tissue, and bones in the hand (Cherniack 1990) with symptoms such as numbness and tingling in the fingers, loss of hand or arm strength, whitening of the fingertips, and problems with precision finger tasks. DC is a slowly progressive and irreversible disabling flexion of the fingers (Picardo and Khan 2012) which results in deteriorated hand grip function. Finally, HHS is a condition of the hand with reduced blood flow to the hand and fingers (Cooke 2003).
Over the years, hand-held tools have been improved, with some attenuation of the vibrations transferred to fingers, hands, and arms, which may have resulted in decreases in the risk of developing vibration-related disorders among workers (Jing et al. 2019). Despite this, research indicates that vibration-related disorders are still a common occupational disorder associated with exposure to hand-arm vibration (Nilsson et al. 2017;Mathieu et al. 2020). In this study, we conducted an up-to-date systematic review of the literature on the relationship between hand-arm vibrations and the occurrence of any of the aforementioned disorders of the hands. In addition, we evaluated whether psychosocial hazards (e.g., psychological job demand, social support, decision latitude) are associated with the occurrence of the pre-defined disorders. With the updated knowledge, we can further motivate and guide improvements in the working conditions, and encourage the development of improved working tools designed to prevent vibration-induced hand-wrist disorders in the work environment.

Methods
This review is part of a larger research project on workrelated physical and psychosocial hazards associated with the development of specific musculoskeletal disorders of the upper and lower limbs. The protocol for this project was registered in PROSPERO (PROSPERO 2020; Koes et al. 2020). Reporting of this systematic review with meta-analysis adheres to the 2020 PRISMA reporting recommendations (Page et al. 2021).

Study selection
Medline, Embase, Web of Science Core Collection, Cochrane Central, and PsycINFO were searched from inception to January 20, 2020, after a pilot search had been conducted in December 2019, by combining carefully selected text words and controlled vocabulary representing (1) musculoskeletal disorders of the shoulder, elbow, hand/wrist, knee, and foot, and (2) exposure to work-related physical and psychosocial risk factors. The full search strategy was developed by a medical librarian and is provided in the Supplemental Materials document.
Duplicate records were removed. Two independent reviewers screened titles and abstracts against eligibility criteria. The full texts of potentially eligible articles were then evaluated by two reviewers independently; in case of disagreements a third reviewer made the final decision about inclusion. Study selection was done using the web-based systematic review tool Covidence.

Study eligibility criteria
We included studies that reported sufficient statistical data to calculate associations between work-related vibration exposure (reported on at least two levels of exposure, i.e., presence or absence) or psychosocial hazards, and the occurrence of HAVS, VWF, DC, or HHS. Psychosocial hazards consisted of social support, job demands, job control, decision latitude, job satisfaction, job security, time pressure, periodic interruptions, and job-related psychosocial distress. We included studies considering a naturalistic workplace or clinical setting. We excluded studies that reported only individual symptoms (e.g., pain, stiffness) instead of disorders or syndromes, studies that used surgery as an outcome, studies defining exposures based on job titles only, and laboratory experiments. Since the main goal was to identify risk factors, we aimed to focus our analyses and conclusions on data from longitudinal studies. However, since we expected to identify only a small number of longitudinal studies, cross-sectional studies were included. If cross-sectional and longitudinal data were published based on the same data, results from the longitudinal analyses were reported.
We included each dataset only once. In cases where multiple publications used the same dataset, we included the most recent publication and consulted previous publications in case of missing data only.

Outcomes
The occurrence of HAVS (including VWF), DC, and HHS was the primary outcome, as reported by the study participants based on physician diagnosis.

Data extraction
We extracted characteristics of the study samples (e.g., mean age of the study sample, sex, type of occupation), outcome definition and measurement (pre-defined as HAVS (including VWF), DC, and HHS), type of risk factor (i.e., vibration exposure and exposure to psychosocial hazards), and the study (e.g., study design, year of publication, adjustment for confounders). We extracted the number of cases, controls, and the number of study participants with and without one of the pre-defined disorders, depending on the exposure to work-related hand-arm vibration or psychosocial hazards. In studies with more than two levels of exposure, we used the lowest level of exposure as a reference.
Odds ratios (ORs) and other effect size measures (e.g., hazard ratios, risk ratios) were extracted from the included studies in addition to the numbers of cases and non-cases which were exposed or not exposed. We used adjusted ORs from multiple/multivariable analyses if no selection of significant predictors was applied by the authors. If only significant predictors were kept in the final multivariable model, we extracted data from bivariate analyses.
Data extraction was done using an Excel spreadsheet, which included descriptive details for coding study information. We used Excel for data extraction because the online tool implemented in Covidence is focused on data extraction from randomized controlled trials and was considered inappropriate in our review. Data were extracted by two independent researchers (AB and HG).

Risk of bias assessment and certainty of the evidence
The risk of bias in the eligible studies was assessed using a list of 16 quality criteria used in previous systematic reviews of our research group to allow for comparability with previous work (van Rijn et al. 2009aRijn et al. , 2009bvan Rijn et al. 2010;van der Molen et al. 2017). The list included 16 items covering five main risks of bias domains: (1) study population, (2) assessment of exposure, (3) assessment of outcome, (4) study design and analysis, and (5) data analysis and presentation. Each item was rated as "high", "low", or "unclear", and criteria on how to score each item were defined a priori. The content of the quality assessment tool corresponds to recently published tools for assessing the quality of observational studies (Bero et al. 2018). The items relating to the second domain (i.e., assessment of exposure) were rated separately in relation to physical and psychosocial exposures. The risk of bias assessments contributed to the evaluation of overall certainty in the findings, which was evaluated using the GRADE framework (Huguet et al. 2013).
By the most recent guidelines for systematic reviews (Higgins et al. 2019), no cutoff was used to differentiate studies of high or low quality. Two reviewers assessed study quality independently (pairs of two among HG, KS, EM, JJ, RGE, RvR, and BK). Disagreements were resolved by a third reviewer (AC). One author conducted the GRADE assessment (HG).

Strategy for data synthesis
We summarized the results in tables including descriptive information about the included studies and ORs for the association between work-related hand-arm vibration and psychosocial hazards, and the occurrence of HAVS (including VWF), DC, and HHS.
We extracted or calculated ORs and corresponding 95% confidence intervals (CIs) for all studies. If more than one level of exposure was reported, we extracted all results to allow for an evaluation of exposureresponse relationships between exposure and health outcome. Then we calculated logORs and corresponding logCIs as well as standard errors to be able to conduct a meta-analysis. The meta-analysis was stratified for the type of disorder. If the exposure was reported as mean and standard deviation, we calculated a standardized mean difference as effect size and converted this into OR with CI using the formulas provided by Lipsey and Wilson (2001).
The meta-analysis was based on a random-effects model according to the method of DerSimonian and Laird, and the heterogeneity was estimated based on the Mantel-Haenszel model. Tau 2 was calculated as an indicator of heterogeneity between the true effects of individual studies (Higgins 2008), and I 2 was calculated to indicate the percentage of observed heterogeneity which is due to differences in true effects of individual studies rather than on random error (Higgins et al. 2003). This can roughly be interpreted as follows: 0%-40%: might not be important; 30-60%: may represent moderate heterogeneity; 50-90%: may represent substantial heterogeneity; 75-100%: considerable heterogeneity (Borenstein et al. 2011).

Characteristics of the included studies
The overall literature search for diverse musculoskeletal disorders of the upper and lower limbs resulted in a total of 8,264 records after the removal of duplicates ( Figure 1).
After the independent duplicate screening of 956 full-text articles, we identified 10 studies that fulfilled the inclusion criteria of the present review (Bovenzi 1994;Virokannas et al. 1995;Lucas et al. 2008;Bovenzi et al. 2011;Descatha et al. 2012;Scharnbacher et al. 2013;Descatha et al. 2014;Palmer et al. 2014;Bast-Pettersen et al. 2017;Haines et al. 2017). The screening process and reasons for the exclusion of full-text articles are documented in Figure 1.
Four of the included studies reported the occurrence of HAVS (including VWF) as an outcome, six reported the occurrence of DC, and one reported the occurrence of HHS. Six studies were conducted cross-sectionally, two were case-control studies, and two were prospective cohort studies (one with a 3-year follow-up, and one with a 5-year follow-up). Only one study reported the incidence of the selected disorders of the hands, while the remaining studies reported the prevalence of one or more of the specified hand disorders. Vibration exposure was assessed in different ways, including the duration of vibrating tools per day, years of exposure to vibration tool use, average daily vibration tool use, and cumulative exposure (hours Ã m/s 2 ). All studies relied on exposure assessments that provided continuous data. However, in three studies, the exposure was analyzed as a continuous variable, while the exposure was categorized into two to four categories of exposure severity in seven studies. None of the included studies reported an investigation of psychosocial hazards in addition to exposure to work-related hand-arm vibration.
In total, the 10 studies included more than 24,381 participants. The exact number of participants remains unclear as one study did not report the exact number of participants. Most study samples consisted of men only, four studies did not report sex, and two studies included women, resulting in a total of 3,698 female workers represented in the present systematic review. The mean age of participants in the 10 studies ranged from 38-67 years. Three of the included studies were conducted in France, two in Italy, and one study each was conducted in Norway, Finland, Great Britain, Canada, and Germany. Three studies specified the study population concerning the type of work, including stone, road maintenance, and railway maintenance workers. Three studies described the study population as workers, two studies described the population as patients and controls, and one study described the study population as vibration-exposed workers and French civil servants, respectively. Table 1 displays relevant characteristics of the included studies and characteristics of the exposure assessments. More details can be found in Supplemental Table 1.
Association between exposure to vibration at work and disorders of the hands Of the 10 included studies, nine consistently showed that intense and long-lasting exposure to hand-arm vibration was associated with the occurrence of HAVS (including VWF), DC, and HHS (Supplemental Table 2). All four studies reporting associations between vibration exposure and HAVS (including VWF) demonstrated significant associations with higher levels of exposure, while five out of six studies reported significant associations between vibration exposure and DC, and one study reported significant associations between exposure to vibrating tool use and using the hand as a hammer and HHS. However, with lower levels of vibration exposure, the pattern was less clear, with mixed results across studies (Supplemental Table 2).
Our meta-analysis confirmed the association between intense or long-lasting hand-arm vibration exposure and selected disorders of the hands (Figure 2), with ORs between the lowest and highest level of exposure ranging between 0.93 and 3.43 in individual studies, and with pooled estimates ranging between 1.35 (95% CI: 1.12-1.63) and 3.43 (95% CI: 2.10-5.59).
However, the magnitude of the association varied considerably across studies as demonstrated by large between-study heterogeneity (I 2 ¼ 79% for HAVS including VWF; I 2 ¼ 83% for DC). Due to the small number of studies, we did not conduct sensitivity analyses or exploratory analyses aimed at explaining the observed between-study heterogeneity.
The certainty of the available evidence was considered low for all three disorders (see Supplemental  Table 3). We considered the risk of reporting or publication bias as moderate, mainly because we expect that not all results are reported in the published reports (i.e., due to selective reporting).

Risk of bias and generalizability of findings
Four studies rated the highest number of items as low risk of bias (7 or 8 of the 16 items), while one study rated only 1 of the 16 items as low risk. Indications of high risk of bias showed a range from nine items rated the high risk of bias in one study, to three items rated the high risk of bias in three studies (Supplemental Table 3).
The most relevant quality-related issues were a lack of longitudinal design in all but two studies and a lack of blinding regarding outcome status in the assessments of exposure status in six studies. In addition, adequate blinding regarding exposure status in the outcome assessment was reported in only one study. Four of the studies reported a low number of cases (below 50) while one study did not specify the number of cases and controls, and none of the studies provided information on potential withdrawals from the study. On the positive side, in all but one study, the results were presented adequately and in all but two studies, the outcome was defined adequately (see Supplemental Table  3). The certainty of the evidence was considered low in all analyses (see Supplemental Table 4).
In addition to the described issues regarding internal validity, the external validity may be limited to male populations because most of the samples in the included studies included only men.

Main findings
Our systematic review and meta-analysis consistently showed an association between exposure to intense (Bovenzi 1994), or long-lasting vibration exposure (Descatha et al. 2012) at work and the occurrence of HAVS (including VWF), DC, and HHS. The associations between the lowest and highest level of exposure  (Continued) varied between ORs of 0.93 and 3.43. The exact magnitude of the association remains unclear because of considerable heterogeneity among estimates of the association reported in individual studies (see analyses on HAVS (including VWF) and DC as outcomes), and a scarcity of available studies on HHS. None of the included studies reported associations between the selected disorders and psychosocial hazards. The observed heterogeneity may be explained by methodological differences among studies as well as differences in the reported analyses. For instance, the study sample varied across the included studies, the definitions and measurements of hand-arm vibration exposure varied, and in terms of the reported analyses, sometimes analyses included potential covariates in multivariable analyses, while other studies reported bivariate analyses only. In addition, the reporting of statistical analyses appeared confusing in some articles about the terminology used. In fact, in all studies, important information regarding the methodological procedures was lacking. The observation of deficiencies in the reporting of statistical analyses in biobehavioral research is consistent with previous literature (Freedland et al. 2009;Hidalgo and Goodman 2013). Unfortunately, the small number of included studies per disorder prevented us from further analyzing and exploring possible sources of heterogeneity statistically by conducting metaregressions or subgroup and sensitivity analyses.
Concerning the assessed indicators of risk of bias for evaluating the quality of observational studies, our findings show a mixed picture, revealing that most studies demonstrated some risk of bias that prevented strong conclusions. Unfortunately, most of the included studies reported that evaluators were not blinded for outcome status or exposure status. Long-term exposure to vibrations often involves a mixture of self-reported data on tools used as well as working hours with tools and objective measures of vibration during the use of the tools. Non-blinded assessments pose a considerable risk to the validity of observer ratings. Future studies should therefore carefully address the issue of blinding in the assessments of exposure as well as outcomes, to allow for more valid conclusions. The certainty of the evidence was considered low in all analyses.

Comparison with previous literature
The overall result of a positive association between exposure to hand-arm vibration and HAVS (including VWF) and DC confirms and extends findings from previous studies (Descatha et al. 2011; Nilsson et al.  (3) never; yes, but not daily; yes, daily 2017; Mathieu et al. 2020), which strengthens possible conclusions from the existing body of evidence. Previous meta-analyses reported HAV to be associated with an increased risk for the occurrence of Dupuytren's disease (Descatha et al. 2011;Mathieu et al. 2020), Raynaud's phenomenon, neurosensory injuries, and carpal tunnel syndrome (Nilsson et al. 2017). When comparing our findings with these previously published systematic reviews with meta-analysis, the results show similar patterns of findings. However, previous publications did not report large amounts of heterogeneity in their analyses. The occurrence of between-study heterogeneity in meta-analyses indicates that a larger amount of variation was seen between results from individual studies than would be expected by chance (Higgins et al. 2003). Typically, this observation indicates that characteristics of the included studies are associated with larger or smaller effect sizes. The overall effect size resulting from the meta-analytic summary of findings needs to be interpreted with great caution, in such a case.
In contrast with the previous publications, we refrain from interpreting the magnitude of associations between HAV and the occurrence of disorders of the hands, given the considerable amount of observed heterogeneity and the low certainty of the summarized evidence. In our view, more rigorous studies with strong study designs (e.g., using prospective cohort designs and investigating the incidence rather than the prevalence of disorders) and more Figure 2. Forest plot of the association between highest level of exposure and the occurrence of HAVS (including VWF), DC, and HHS (logOR with 95%CI). CI ¼ confidence interval; OR ¼ odds ratio; DC ¼ Dupuytren's contracture; HAVS & VWS ¼ Hand-arm vibration syndrome & vibration induced white finger; HHS ¼ Hypothenar hammer syndrome; I-squared ¼ indicator of between-study heterogeneity; p-value ¼ indicates statistical significance level (a missing I-squared and p-value indicate that no meta-analysis was calculated. standardized approaches to analyzing data (e.g., by reporting results based on continuous data) are needed to allow strong conclusions regarding the exact magnitude of the association between the exposure to HAV and the occurrence or development of hand disorders.

Limitations
First, we focused only on the predefined work-related exposure to hand-arm vibration. If not accounted for by the authors using adequate statistical methods, we cannot rule out confounding or interactions with other risk factors (e.g., age), which were not taken into consideration. Second, we cannot rule out that we missed relevant studies. However, given our broad and systematic literature search in medical and psychological literature databases and a large number of screened articles, we consider the risk for the occurrence of selection bias according to our inclusion criteria to be small. Third, given the different methods for assessment and reporting of hand-arm vibration exposure, as well as the observed variation in the used cutoffs for categorizing continuous exposure assessments, we were unable to analyze exposure-response associations in more detail, for instance using metaanalytic techniques. To enhance the ability to conduct future meta-analyses, epidemiological studies should analyze and report continuous data for analyzing exposure-outcome associations, by best practices. Accordingly, the use of cutoff values of exposure should be applied with great caution, particularly because such values are often arbitrarily set and are not derived from a true physiological dose-response pattern. Fourth, we consider the quality of the included studies to be low, with all studies being at a high risk of bias in at least 2 of the 5 assessed domains, and at least 3 of the 16 items. An important limitation of the available evidence is the limited number of studies (2) with a longitudinal design. Therefore, only preliminary conclusions regarding causality are possible based on the available evidence. Fifth, the included studies predominantly reported associations between exposure and the prevalence of one of the relevant disorders but no associations with the incidences of disorders. However, recording only the prevalence of events prevents the observance of the natural course of disorders, and information is lacking about how many new cases occurred or how many cases were remitted during the time of observation. Therefore, it is not possible to conclude causal relationships between the observed exposures and disorders. Sixth, we consider the risk for the occurrence of reporting or publication bias as moderate because we assume that not all available results were reported in the published study reports. For instance, only 3 of the 10 included studies reported findings based on continuous data while all studies assessed vibration exposure on a continuous scale. Accordingly, it is possible that non-significant results were not included in the published study reports. Finally, it is important to consider that variation across the different studies' worker populations, work characteristics, methodology, as well as assessments and analyses, limits direct comparison across studies.
Interpretations based on such comparisons should be done with caution.

Conclusions
Our analyses confirm the association between exposure to intense or long-lasting vibration at work and the occurrence of HAVS (including VWF), DC, and HHS, with ORs ranging between 0.93 and 3.43. Due to variation among effect sizes from individual studies, the magnitude of the association remains unclear. Further, since the majority of studies were cross-sectional, no conclusions are possible regarding causal relationships between hand-arm vibration exposure and the occurrence of the selected hand-wrist disorders. However, the two longitudinal studies both suggest that hand-arm vibration exposure at an earlier time point is associated with the occurrence of HAVS and DC at a later time point. Our analyses thus confirm that exposure to hand-held vibrating tools at work is negatively associated with worker health.
We conclude that despite the uncertainty of estimating the exact magnitude of the association between HAV and disorders of the hands, we see a consistent reporting of significant associations across individual studies and meta-analyses. With regards to the implications of our findings for practice and policy, our results suggest that future research should specifically address whether reducing the exposure to hand-held vibrating tools at work will also contribute to reducing the incidence of disorders of the hands.