Letter to the Editor: Landsbergis et al (2019) Titled “Work Exposures and Musculoskeletal Disorders Among Railroad Maintenance-of-Way Workers”

Readers are invited to submit letters for publication in this department. Submit letters online at http://joem.edmgr.com. Choose “Submit New Manuscript.” A signed copyright assignment and financial disclosure form must be submitted with the letter. Form available at www.joem.org under Author and Reviewer information. To the Editor: We reviewed with interest the report of Landsbergis et al (2019) titled “Work Exposures and Musculoskeletal Disorders Among Railroad Maintenance-of-Way Workers,” with partial support from the Association of American Railroads. We were initially excited to read a publication addressing potential hazards this population of workers may face. We especially commend the Brotherhood of Maintenance of Way Employees Division and their efforts toward the health and safety of their membership. However, as our interest is the pursuit of workplace health matters, we must express concerns when we find research efforts that appear to lack appropriate scientific rigor. We appreciate the opportunity to discuss flaws in the Landsbergis et al (2019) article, some of which appear to be severe enough to invalidate the results, individually and especially in aggregate. Our discussion is provided below in the sections of Study Design, Study Survey, Measures of Exposure and Available Research, Health Outcome Data, and Statistical Methodology. We previously obtained IRB approval and requested the de-identified database used in this article, to independently assess our concerns, but our request was declined. Thus, we were unable to otherwise dispense with some items of concern, which necessitated this lengthier critique. STUDY DESIGN The authors purport to have used a cross-sectional study “to measure musculoskeletal disorders and occupational risk factors among railroad maintenance-of-way (MOW) workers.”1 by administering a survey of workers located throughout the US. Thus, at best, it is limited to providing estimates of the prevalence of the population characteristic of interest.2,3 The survey relied on survey responders’ recall of discomfort (with the exception of two questions regarding recall of a doctor's diagnosis of carpal tunnel syndrome (CTS) or a “back problem”), and solely relied on subjective recall of exposure to workplace factors. No vibration emissions of tools, equipment, or vehicles were documented and/or integrated in the analyses for hand-arm vibration (HAV) or whole-body vibration (WBV) exposure assessment, but rather it appears, the authors arbitrarily assigned the emission levels of these workplace exposures. The authors’ stated hypothesis was, “the MOW workforce will have a higher prevalence of hazardous work exposures and a higher prevalence of musculoskeletal disorders (MSDs) than the US workforce as a whole.”1 The objective and hypothesis were poorly defined, lacking both specificity of exposure and outcomes. There are generally 5 categories of hazardous work exposures: chemical, biological, physical, ergonomic, and psychological. Among chemical hazards, there are more than 400 hazardous substances where OSHA currently has rules to limit exposures.4 The World Health Organization defines over 150 distinct MSDs.4 The generalization of so many categories of hazards and use of undefined MSDs created potential multiple comparison problems that were not addressed either qualitatively or apparently statistically.5–10 STUDY SURVEY Subjective Recall Data collection by Landsbergis et al (2019) relied solely on responders’ recall of both discomfort and perceived exposure to workplace factors. Only two questions on the entire survey were regarding if the participant recalled a doctor providing them a diagnosis (ie, CTS). Many studies have concluded that self-reporting of exposure to workplace factors is not accurate.11–17 Landsbergis et al (2019) cited Stock et al (2005) as support for their data collection using subjective recall through survey questions. Stock et al (2005), in their review of the literature, examined previous studies’ comparison of survey questions on exposure to physical workplace factors and referent measures, including direct measures. For the issue of validity, only two papers met their inclusion criteria. Examples of their findings included that validity was “poor” (eg, distance walked), “very poor” (eg, kneeling or squatting), “poor to fair” (eg, repetitive movements). In contrast with the representations of the work by Landsbergis et al (2019), Stock et al (2005) concluded that, “An interdisciplinary approach to measuring workload exposure is particularly relevant to specific industry-based studies or studies of individual workplaces aimed at identifying risk factors in order to intervene and transform work conditions.”18 Response Rate and Selection Bias The authors reported that their response rate for their survey was 12% (4816 respondents from 38,555 invited is 12.5%) of the available MOW workers (active and retired) US population. However, the authors also stated that, “respondents could skip questions and could submit the survey with any degree of completion.” An example of this concern is shown in the survey question “How often does your job involve repeated lifting, pushing, pulling, or bending?” answered by 2741 respondents, for a question-specific response rate of 7.1% (or only 57% of survey respondents). A minimum 70% response rate, 10-fold greater, was one of the four quality criteria in the well-known NIOSH publication for quality inclusion criteria for epidemiological studies of musculoskeletal disorders.19 Aigner et al (2018) reported that low response rates (ie, the percentage of those responding to a study out of all contacts) contribute to the systemic error of selection bias, and systematic selection bias error grows stronger as the response rate decreases.20 Low response rates contribute to “sample selection bias,” whereby individuals who do not participate are collectively, systematically distinct from those who do participate.21,22 Liu and Wronski (2018) reviewed the results of 25,080 actual web-based surveys and found that the completion (ie, response) rate was influenced by the survey design, including the duration, the structure of the questions, and the availability of progress (ie, the web-based survey telling the user how many questions are left or some countdown of the expected period to completion).23 Identifying one's gender was a question that appeared late in the Landsbergis et al (2019) 16-page survey, which received a response rate of only 8.1%, or that 35% of respondents did not answer this basic question and/or had quit the survey entirely by this point. Indeed, the authors state that, “the response rate tended to decline in later pages of the survey.”1 However, a careful review of the survey instrument and the apparent response rates in the tables in the manuscript suggests that most of the attrition, possibly one-third of the few respondents, may have occurred quite soon into the survey, apparently and critically at the point where the first complex exposure question was asked. At and beyond that point, the participation rate appears to only be approximately 8%, which is apparently this study's effective response rate. Which participants would drop out at that point and would they be effectively missing at random or non-random? It would seem more likely that people without problems were more likely to drop out, likely producing an even greater bias than that already presented by a “12%” participation rate. These findings should raise marked concerns about the validity of the study's residual participant responses and resultant data analyses. This study documented an extreme degree of self-selection bias. Landsbergis et al (2019) attempted to investigate selection bias through the administering of a much shorter 10-question survey. The authors relied on BMWED Health and Safety Committee members to contact 395 identified non-respondents, of which 135 completed the phone-administered survey, a subject pool of (139/38,555) 0.35%. Supplemental digital content describes 10 of the variables that were asked of the non-respondents as compared to the study participants. Of those 10 variables, 8 were statistically significantly different, with all but one having a P-value < 0.001, and one was approaching statistical significance (P = 0.095). These indicate that the sampled non-respondents are statistically markedly different from the study participants, thus helping quantify the extreme degree of self-selection bias in this manuscript. The probability that the sampled non-respondents and the study participants are drawn from the same population is essentially impossible, which can be roughly estimated by multiplying the P-values of the 10 variables that are compared between the two populations. Validity of Survey Questions The authors report that their survey questions were based on several other “validated questionnaires”1 including, “the 2015 U.S. NHIS-OHS,”24,25 the “Nordic Musculoskeletal Questionnaire,”26–28 the “Medical Research Council (MRC) National Survey of Health and Vibration in the UK,”29 a “European collaborative study of vibration related health risks,”30 and “previous studies of transportation workers.”1,31–33 Close inspection of the questions in the Landsbergis et al (2019) survey reveals that few of their questions matched the corresponding source for either the wording of the actual question or the provided scale for response, or both as is required for claims of validity. MEASURES OF EXPOSURE AND AVAILABLE RESEARCH MOW Exposure Research Critically, Landsbergis et al (2019) failed to present or discuss objective exposure data that are publicly available regarding MOW workers’ exposures to workplace factors. Specifically, some of us (Weames et al [2017a] and Weames et al [2017b]) have published objective data for average full-shift workplace exposure for MOW section maintainers (average objective exposure data based on 14 full-shift data collection days in 10 States and spanning 2003 to 2011) and MOW track welders (average based on seven full-shift data collection days in four states and spanning 2004 to 2013, respectively).34,35 Section maintainers work in gangs of three to four workers and effect localized track repairs. Track welders work in gangs of two workers and are trained to restore worn steel that typically occurs at switch points and frogs (a track appliance that is part of a switch that allows trains to safely roll from one track to another) as well as weld pieces of rail together. Section maintainers operate out of a basic track maintenance forces work vehicle while track welders generally use an industrialized-looking delivery truck. Both types of vehicles can operate on roadways and on railroad tracks. There are several different power tools used by these MOW workers including the hydraulic spiker, hydraulic spike puller, hydraulic tamper, impact wrench, rail saw, rail grinders, and chipping hammers. Power tool operation is the source of HAV for MOW section maintainers and track welders. The published research of Weames et al (2017a) quantified that the section maintainer uses power tools (all various power tools combined) over the duration of their shift for only 6.9 minutes and 7.1 minutes, on average, for the left and right hand, respectively. Weames et al (2017b) quantified that the track welder uses power tools over the duration of their shift for 22.4 and 22.6 minutes, on average, for the left and right hand, respectively. Weames et al (2017a and b) defined power tool use as manipulating and triggering the power tool at the site of its use (for example, setting up a rail saw and actively cutting the rail). Thus, actual HAV exposures can be somewhat less than the results reported. Thus, the results of Weames et al (2017a) and Weames et al (2017b) quantified that there is not an increased risk for the development of MSD adverse health effects for these MOW workers. Critically, Landsbergis et al (2019) failed to note that the published research demonstrates that such brief exposures to HAV, regardless of vibration characteristics such as amplitude, does not show associations with any MSDs including vibration white finger. The Weames et al (2017a) and Weames et al (2017b) objectively measured exposures would, according to Landsbergis et al (2019), be categorized as occurring “rarely.” Two other publications, Johanning et al (2020) and Landsbergis et al (2020) are based on the exact same survey data as Landsbergis et al (2019).1,36 In contrast with the objectively measured exposure data from many workers, it is reported in Johanning et al (2020) that approximately half the power tools in the survey are each individually reported by their survey respondents as being “always” (8 to 10 hours/day) or “often” (4 to 6 hours/day) used daily according to >50% of the respondents. These results are incompatible with the objectively measured findings of Weames et al (2017a and b) presented in Tables 1 and 2. Griffin and Bovenzi (2007) provided specific warnings against sources of error if researchers pursue subjective, self-reported estimations of HAV exposure duration, rather than reliance on objective measurement.30 Within the data relied upon by Landsbergis et al (2019), there was a failure to recognize that for any single respondent it is not mathematically possible for any combination of two or more of nine different power tools to be used “always” or “often” daily. In fact, as MOW work is by its nature highly variable depending on the type of repair to be done, it is impossible for any MOW worker to use any one power tool “always” or “often” daily. Tables 1 and 2 from Weames et al (2017a and b) show the percent of the work shift workers are exposed to workplace factors for hand activity for section maintainers and track welders, respectively. These measurements are made from full-shift video using standard work sampling methodologies (Niebel, 1982)—14 shifts for section maintainers, 7 shifts for track welders.37 For both crafts of MOW workers, the typical workday includes general activities at the beginning and end of the shift (warm-up time, safety briefing, and job briefing), breaks, travel, and no work activity time (waiting for access to the track, otherwise known as track and time). In total, performance of section maintainer work constitutes 17.4% and 18.8% of the work shift for the left and right hands, respectively. Power tool use makes up 1.4% and 1.5% of the work shift for the left and right hands, respectively for section maintainers. For track welders, performance of track welder work constitutes 19.6% and 22.4% of the work shift. Power tool use makes up 4.7% and 4.7% of the work shift for the left and right hands, respectively for track welders. The report by Landsbergis et al and their survey failed to address overtime. However, overtime is typically a function of idle time and/or waiting for safe access to the railroad track. Thus, overtime does not materially increase tool use duration. Additionally, overtime does not typically happen on a daily basis. The Weames et al (2017a and b) articles included shift durations up to 10 hours. The exposure durations reported by Landsbergis et al remain impossible even in light of potential overtime work. Weames et al (2017a) and Weames et al (2017b) analyzed distal upper extremity morbidity for the section maintainer and the track welder, through the application of the Strain Index.38–40 The Strain Index is a validated semi-quantitative method that is a calculation of risk that includes six components of intensity of exertion, repetitions/minute, percent duration of exertion, hand/wrist posture, speed of work, and duration of exposure/shift. The analyses using the Strain Index found that the section maintainer and track welder jobs were not hazardous. These data were, again, ignored by Landsbergis et al (2019). MOW work equipment is used in track maintenance and repair and involves various types of operated machinery, mostly with the operator seated on-board the equipment, a source of whole-body vibration that Landsbergis et al (2019) assert to have evaluated through their survey. Qualitatively, MOW work equipment is quite heavy, the rail are and the equipment less than and while in work at less than a A MOW work equipment operator while on-board their work equipment, will generally of travel, and each with exposure time and of even the same with MOW work equipment, particularly if any adverse health effect is expected to be and MOW work equipment has evaluated for exposure in exposure that are below the of the Health which effects have not documented and/or objectively of and probability of an adverse health of Thus, the objective vibration for MOW work are or those for in a and these data were all not included by Landsbergis et al To our concerns regarding the data of Landsbergis et al (2019), the data are and Additionally, there was no on available objective data by which have shown different exposures and MSD risk than in Landsbergis et al (2019). The study of Landsbergis et al (2019) does not use any a the required to a generally or a used in epidemiological Landsbergis et al results of is mostly a discomfort survey in a percentage of MOW workers The authors did to that respondents the population of MOW they failed and found marked in of 10 A health concern is the lack of for or known are study flaws without of any other Landsbergis et al (2019) only and as However, there are many each with its of that markedly by For for example, job and other factors are known risk factors for risk and Statistical is a in any and should the of a statistical with a power of that and a of the analyses including from the analyses with The report by Landsbergis et al (2019) questions about the of the statistical analyses Examples and of the data analyses including missing an of statistical a lack of statistical of and for potential and no between specific a statistical analyses and analyses. However, of the potential for of the of study data is to without access to the The Landsbergis et al (2019) does not to assess the scientific approach used in their Statistical a in to associations and between exposures and potential outcomes. For research to be or it must have appropriate study design, be and analyzed study and can often in which impact the quality of Examples the use of (eg, participant lack of objective measures, and In most can be identified with of a and Landsbergis et al (2019) provided some for their collection of the lack of makes it impossible to the analyses There are several where data are without an for The of data is and concerns about of data and how missing which are documented to marked of individual data in this were were that they could questions and the of those are and of missing data are especially when they may have typically approximately one-third of the data for any and not so the of The many statistical without how many other were that there may be multiple comparison There are no for multiple that were reported. There are concerns about the potential for results when multiple are The development of a statistical to that the specific statistical made and results, are influenced by factors including quality of appropriate and of statistical of the and of the the type of statistical to use in the is by the type of data and the study question being There is no or of a statistical in the Landsbergis et al (2019) The of the analyses was and it is if the statistical were Examples of this how many were how missing data were or how were The results of the statistical to the study Landsbergis et al (2019) only describes three and No other of potential is The method for for is also not to a where it would be independently statistical and regarding for are not of variables shown to be for various was not thus providing source of potential In Landsbergis et al (2019) purport to have identified risk factors for various MSDs among MOW workers. However, there are problems with the research which low effective response rate of as compared to the reported rate of of either questions and/or of validated items in validity. which do not risk factors. of known specific to each Only subjective data for all exposures and were of known objective exposure The known objective exposure data be with the data from Landsbergis et al (2019). response in between respondents and the of exposure durations are We have this to the regarding multiple of validity and of the Landsbergis et al (2019) We the Brotherhood of Maintenance of Way Employees Division and their efforts toward the health and safety of their membership. We with their to and quantify potential hazards that MOW workers face. However, the data the Landsbergis et al (2019) do not support the to MSDs and occupational risk among MOW workers.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.