Reviewed for the IDSP by
Susan R Stock MD MSc FRSC
April 1994
This review is an extremely thorough and rigorous one and the most complete review of this topic I have seen to date. It is apparent that much thought and work has gone into it. The approach taken is that of a systematic overview and it has included all the steps recommended for such an overview.
Completeness of literature search
Literature search methods were clearly spelled out Significant effort went into identifying as much of the relevant published literature as possible and the literature search appears quite comprehensive. It might have been useful to include a search of CIS, the ILO occupational health bibliographic database, as it includes a wide range of European studies that often are not included in Medline and contains unpublished PhD theses and public agency reports. Criteria for initial identification of relevant studies and for selection of studies to be included in the detailed review were stated. The criteria for the latter could have been more specific.
Primary question stated
A primary question for the systematic review of the epidemiologic literature was identified, ie "Is Dupuytren's contracture associated with manual work (possibly involving repetitive tasks and/or vibration; not involving a single injury')?
Validity study of the studies reviewed
The criteria questions used are appropriate and quite pertinent. The specific wording of the choices for scoring are somewhat ambiguous and could be improved. This likely would have improved the inter-rater reliability which was, at best, moderate. The decision to calculate odds ratios and confidence intervals for each of the studies was particularly useful for comparing strength of association or statistical significance.
Conclusions of the review
My interpretation of the results of this overview of manual work/vibration studies is somewhat different than the author's. Two of the studies rated 14/22, a moderate rating, and met at least the minimum standards for the criteria posing the greatest threat to validity. These studies are of moderately good quality and the only two upon which significant weight can be put. Both demonstrated a significant relationship between manual work or vibration and Dupuytren's contracture. It is only the much poorer studies which had negative conclusions. While these two studies certainly have methodologic weaknesses, and therefore their conclusions are not definitive, they do suggest a relationship between this disorder and biomechanical factors at work. The strong dose response relationships noted by Cocco for vibration is also very suggestive that a causal relationship may well exist. This relationship will need to be confirmed in better designed future studies as suggested in Dr Liss' recommendations.
Other comments
In order to sort out whether Dupuytren's contracture is associated with biomechanical factors at work, it is important to identify specific ergonomic factors that may contribute to the development of this disorder rather than speak of "manual work'' in a very vague and general sense.
It has been my understanding that one of the main mechanism hypothesized for the work-related development of this disorder is mechanical rubbing of tools or parts in the palm of the hand particularly when these tools or parts must be gripped very tightly. (Vibration, or the use of vibrating tools, is the other main mechanism hypothesized.)
This certainly has been my anecdotal experience. In 1988 I participated in the physical examination of some 200 meat processing workers in a huge US meat packing plant. We noted a high number of cases of Dupuytren's contracture and trigger finger in those who had to grip meat hooks to hold carcasses in place while they cut or otherwise manipulated them. These workers appeared to be exerting very high grip forces for most of their work day. This observation fits with Skoog's 1957 observation that the ulnar part of the aponeurosis is where the greatest strain occurs in grasping and passive support (Liss, p. 19) and with Hueston's 1987 observation that elderly women may develop Dupuytren's contracture after wearing a Colles' fracture cast (Liss, p. 25).
Future studies will need to test these hypotheses and ensure that exposures are measured in such a way that these factors are taken into account while controlling for relevant confounders and risk modifiers. ''Manual work's per se may not be relevant but rather work tasks that involve continuous and/or forceful grasping or gripping or exposure to vibrating tools.
Andreas Maetzel, MD
Ontario Workers' Compensation Institute -
Date: 6 April 1994
The purpose of this review is to critically assess Garry M. Liss's paper "Occupation and Dupuytren's Contracture" (referred to hereafter as the "paper").
In his paper Gary M Liss conducts a systematic review of the scientific literature on Dupuytren's Contracture (DC) with the goal of synthesizing evidence concerning two questions about the disease:
1) Is Dupuytren's Contracture (DC) associated with or caused by manual work particularly repetitive manual work (including vibration), and
2) Is DC associated with or caused by a single injury
To be useful, any review of scientific literature on a particular issue must involve a thorough compilation of the available literature, a narrowing down of that literature to that which is applicable, an assessment of the methodologies employed, and finally a synthesis of the findings. Reviews are an important source of information and for this reason much attention has been focussed on improving the scientific standards of reviews.
Criteria for assessing the quality of review articles have been published, [Oxman & Guyatt, 1988] and validated [Oxman et al., 1991], and I intend to apply the 11 criteria outlined by the McMaster group to evaluate the paper. The criteria have been designed to maximize reproducibility and ensure validity. They serve to discriminate among quality characteristics of reviews. They are as follows:
1) Were the search methods reported?
2) Were comprehensive search methods used to locate relevant studies?
3) Were explicit methods used to determine which articles to include in the review?
4) Was selection bias avoided?
5) Were validity criteria reported?
6) Was validity assessed appropriately?
7) Was the assessment of the primary studies reproducible and free from bias?
8) Were the methods used to combine studies reported?
9) Were the findings of the primary studies combined appropriately?
10) Were the conclusions supported by the reported data?
11) What was the overall scientific quality of the overview?
For each criterion the paper will be rated on the following seven-point scale:
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
where my rating will be represented by a grey shading of the respective value.II) The organisation of the paper
The paper is divided into 8 parts. Parts 1 through 5 are narrative in nature and ,generally relate to pathogenic, diagnostic and epidemiological concepts regarding DC. Parts 6 through 8 focus- on the goals of the paper as these relate to the review of the evidence concerning the relationship between manual work/single injury and DC.
III) Application of the McMaster Criteria
1) Were the search methods reported?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
Explicit specifications of exposure, outcome and populations are essential and have to be incorporated into the search methods. Using the keywords occupation, Dupuytren's Contracture, cumulative trauma, and acute injury the author has done a good job incorporating exposure and outcome into his search methods. A more comprehensive search incorporating text- and/or keywords like work, agricultural worker's diseases, injury, trauma could have been used to broaden the search, since the author points out the confusion about these terms on page 13. However, it is questionable if a broader search had yielded more articles due to the paucity of the published research on this topic.2) Were comprehensive search methods used to locate relevant studies?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
The author has undertaken a very comprehensive literature search by consulting MEDLINE as well as NIOSHTIC databases, by reviewing occupational medicine texts and books on Dupuytren's contracture, and by examining the references provided in the studies. Index Medicus has been searched by hand for the years 1990 through 1993. This demonstrates the author's awareness of the limitations of electronic literature searches and guarantees sufficient comprehensiveness of the literature search. Some might suggest that an effort be made to search unpublished literature. However this is very difficult to do and its value is debated in the scientific community.
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
3) Were explicit methods used to determine which articles to include in the review?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
While the author makes every effort to select studies that employ specific and detailed exposure measurements, he also acknowledges the limitations of being overly explicit. Occupational exposure is often only broadly defined. For example, occupational exposure will usually be measured as a function of job title rather than with a detailed reference to specific body movements and actions. Even large and well-designed studies like the Framingham Study or the National Health and Nutrition Examination Survey (NHANES) limit the definition of occupational exposure to job title. Upon reviewing the available evidence one is unfortunately restricted to such simple exposure definitions which in fact limit the conclusions that can be drawn from the available research.Although it might be suspected that a disease like DC might be better defined for epidemiologic studies, the author demonstrated the inconsistency in currently available disease classifications. Therefore I conclude that the author used inclusion criteria which were well suited to the existing knowledge.
4) Was selection bias avoided?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
The author specified how primary articles were chosen. The author however did not test whether the inclusion criteria yielded reliable results. He could have done this in the same way he assessed validity by choosing a second examiner applying the inclusion criteria without knowledge of the authors results. Had this approach been used, a statistical measure of agreement could have been reported. This procedure would have added additional credibility to the work, by serving as a check against bias.5) Were validity criteria reported?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
This part of the review is very crucial since it relates to the validity of the findings and the generalizability of the results. The author uses criteria developed by Susan Stock in her review on workplace ergonomic factors and the development of musculoskeletal disorders of the neck and upper limbs. The validity criteria used by the author satisfy fundamental requirements concerning the judgment of epidemiological studies. They are supported by modern textbooks [Rothman, 1987] and also other authors who synthesized evidence from nonexperimental studies [Esdaile and Horwitz, 1986; Stock, 1991]. In particular the following characteristics are required (and were present in the paper):a) Avoidance of selection bias
b) Unbiased exposure ascertainment
c) Unbiased outcome ascertainment
d) Control for obvious confounders
e) Adequate follow-up of all those enrolled or satisfactory participation by cases and controls
6)Was validity assessed appropriately?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
" Authors will come to correct conclusions only if they accurately assess the validity of the primary studies on which the review is based. If all the studies have basic flaws their conclusions may be questionable even if their results are comparable [Oxman & Guyatt, 1988]".It is obvious that a study with major flaws in all of the mentioned qhality _characteristics cannot be considered as having valid results. It is nevertheless a matter of choice where to put the threshold for non-consideration. Susan Stock for example in her article [Stock, 1991] concludes that "Prior to assessment it was decided that, if any study had major flaws, (i.e., score of 1 on the validity assessment tool) in any of these areas (comparability of the groups and validity of exposure and outcome measures), the validity of the results of that study would be seriously compromised, and the study would not be included in any aggregate analysis" [Stock, 1991 p.91]. When considering whether to exclude any study with a major flaw in one or more of the criteria, one ought to do so if it can be proven that this or these flaws seriously compromise the conclusion. The author conducted a very thorough assessment and has done a good job, given the quality of the studies he had to work with.
7) Was the assessment of the primary studies reproducible and free from bias?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
"Ideally the potential primary studies should be assessed for inclusion by at least two authors, each blind to the other's decision, and the extent of agreement should be recorded. Reproducibility should be quantified with a statistical measure that quantitates agreement ... such as an intraclass correlation coefficient or a kappa statistic" [Oxman & Guyatt, 1988].The author follows high quality standards to ensure reproducibility of the assessments. This is demonstrated by the high agreement among the two observers. However, validity assessments are lacking for the three "non-english" studies; therefore it is difficult to say that the assessment was free from bias and it would have been desirable to see also the results for those three studies.
8) Were the methods used to combine studies reported?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
In order to combine results from epidemiologic studies several conditions must be met. Exposure categories must be identically defined for the results to be combined across studies. Otherwise, without a common definition, the combined measure will have no real meaning. Although there are methods to adjust odd's ratios coming from studies with different exposure measurements, these adjustments only can take place if the correction factor is known from previous studies. Clearly the paucity of available studies makes any adjustment of odd's ratios impossible. The author correctly avoided combining the results of the studies because of the major differences in study design, and exposure and disease classification.9) Were the findings of the primary studies combined appropriately?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
As mentioned above, it was not appropriate to combine the results of the studies.Although it does not fall under the McMaster criteria it is worth noting that the author is the only person to extract the results and calculate odd's ratios from the raw data provided in the literature. This step of data-extraction is very prone to error and I strongly recommend that data extraction and calculation be performed by a second examiner and that each examiner possibly do this procedure twice.
10) Were the conclusions supported by the reported data?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
Because the results can't be statistically combined the author should state explicitly the basis for his conclusions.In his conclusion the author discusses all studies included in the review. However, as the author points out, from table 7 it is obvious that 4 of the 6 studies lack even basic quality standards to merit further consideration. Strictly speaking, these studies should be excluded from the analysis.
The characteristics of the remaining studies are only briefly discussed. In my view, this discussion should have been more explicit. For example, how large is the influence of existing bias for the reported results? Do the biases lead to over- or underestimation of the odd's ratios?
For example, the study by Bennet may suffer from selection bias since it used different populations for index- and comparison-group, whereas the study by Thomas & Clarke does not. Even though the bottomline of both studies shows that there is a positive association.
Both lacked blinding of the examiners for outcome ascertainment. This is a problem if DC is defined by the presence of nodules. DC was defined this way in both studies whereas it was defined as presence of contracture in the study by Cocco. The effect of unblinded observers probably will be negligible in this latter study.
All studies, even those of good quality lacked blinding to case status during determination of exposure. This is a considerable problem if determination of exposure is susceptible to recall-bias but is probably less relevant if exposure is restricted to job title.
Confounding factors which are associated both with exposure as well as disease have been considered in the validity assessment. The assessed studies did adjust for age and gender and it is debatable whether adjustment for the other confounders mentioned in the validity criteria would have had a big impact on the final result, e.g., ethnic origin won't be of relevance in the study by Bennet and it is debatable whether measurement and adjustment for family history of DC would have yielded more precise results, since measurement of family history as such is prone to error.
In my opinion the biases of the better rated studies probably would have led to overestimation. How much is difficult to say since a high standard study is unavailable. The author considers the guidelines for assessing the strength of a causal inference:
1)Is the temporal relation correct?
2)Is the evidence strong?
3)Is the association strong?
4)Is there consistency between studies?
5) Is there a dose-response relation?
6) Is there indirect evidence that supports the inference?
7) Have the plausible competing hypotheses been ruled out?
The author correctly concludes that there is insufficient evidence to definitely say that DC is a disease related to manual work. The results nevertheless strongly suggest a possible association although this remains to be confirmed by future studies.
What was the overall scientific quality of the overview?
1 2 3 4 5 6 7
Extensive Flaws Major Flaws Minor Flaws Minimal Flaws
(Very Poor) (Poor) (Good) (Exemplary)
The author selected the relevant research and thoroughly presented the validity of the research that was reviewed.In the initial sections of his work the author describes the disease, and discusses epidemiological characteristics and pathogenetic concepts.
These sections, in my mind, provide relevant information to the reader who needs to be informed about the disease, its epidemiology and the common pitfalls when studying a disease with a probable multifactorial etiology. The author correctly points out the numerous difficulties encountered in the study of DC, e.g., the absence of diagnostic criteria, the presence of disorders which mimic DC, and the difficulties in establishing a hereditary disease pattern. These sections of the paper however do not contribute to the analysis of the special question he wants to address.
The discussion about single injury and DC reported in section 7, if considered under the same rigorous criteria as the review of the evidence about the relatedness of DC to manual work, is rather narrative. There seems to be no scientific evidence with which to make a valid conclusion. The discussions in different textbooks are narrative in nature and dominated by beliefs and opinions. The diverging opinions among experts are also reflected in the information presented in the table on page 17. Therefore the guidelines forwarded by McFarlane and Shum should be welcomed but certainly would be improved if consensus existed among experts, otherwise they will stay as arbitrary as they appear. A more thorough data-extraction also would have added to the quality of the paper. Finally, the references Thomas & Clarke, and Herzog should be added to the reference section of the paper.
IV) Commentary
I found the paper useful as a review of DC's relatedness to work. In this regard I believe that the author thoroughly assembled and analysed the available literature according to the issues addressed. The scientific methods were properly employed.
My only criticism of the paper relates to its organization. In my view the lead-in to the paper's main purpose is inordinately lengthy. The paper really only sets out to fulfill its goals in chapter 6 through 8 and the tables in the appendix, which are very useful.
V) References
1. Esdaile JM, Horwitz RI. Observational studies of cause effect relationships: an analysis of methodologic problems as illustrated by the conflicting data for the role of oral contraceptives in the etiology of rheumatoid arthritis. Journal of Chronic Diseases. 1986;39:841-852.
2. Oxman AD, Guyatt GH, Singer J et al. Agreement among reviewers of review articles. J Clin Epidemiol. 1991;44:91-98.
3. Oxman AD, Guyatt GH. Guidelines for reading literature reviews. CMAJ. 1988;138(83:697-703.
4. Rothman KJ. Modern Epidemiology. Boston/Toronto: Little,Brown and Company; 1986.
5. Stock SR Workplace ergonomic factors and the development of musculoskeletal disorders of the neck and upper limbs: a meta-analysis. Am J Ind Med. 1991;19:87-107.