Work-related outcomes in randomized, double blind, placebo-controlled trials in osteoarthritis – are they adequately reported in journal publications? A systematic review

Osteoarthritis (OA) is a very common condition among the aging population with a high incidence at age 50 years and older [1]. The prevalence estimates of OA vary depending on the specific joint that is affected and the characteristics of the population under study [2, 3]. A study by the World Health Organization (WHO) Scientific Group on Rheumatic Diseases estimated that 10% of the world’s population who are 60 years or older are affected by some form of OA [4]. Based on x-ray imaging, the prevalence of knee OA among an adult cohort aged 56–84 years in Malmo, Sweden, was about 25%, while symptomatic knee OA had a prevalence of 15% [5]. Of the 978 subjects in the Framingham Study Community [6], it was estimated that hip OA was present in 20% of the cohort based on imaging and about 4% had symptomatic OA.

OA is typically characterized by pain, impairment of mobility and stiffness [7]. The pain is related to loss of cartilage, which may lead to joint failure [8, 9]. Factors identified that are associated with OA include inflammation, obesity, weak muscles, cartilage defects and genetic factors [9]. Joints commonly affected by OA include the knee, hips, hand and wrist. Osteoarthritic pain and other accompanying symptoms may lead to a reduction in patients’ quality of life and may also interfere with productivity at the workplace [9]. Indeed, it has been shown that moderate to severe chronic painful conditions may lead to lack of productivity at work [10], work absenteeism as well as substantial financial costs imposed on the healthcare system [10, 11].

For the present review, we refer to outcomes that speak to a person’s ability to work or to productivity at work as ‘work-related outcomes’. Data on such work-related outcomes, which are commonly captured as time lost from work or interference with work ability, are often collected in pain trials – typically as answers to component questions of commonly used questionnaires. However, these work-related outcomes are only infrequently reported in publications on such pain trials [12]. How work-related outcomes in trials on OA specifically have been reported in the scientific literature is not known; this is the subject of the present review.

In this systematic review, we therefore investigated the reporting of work-related outcomes in randomized, double blind, placebo-controlled trials on the treatment of hand, hip or knee OA. Additionally, we had aimed to analyze pain intensity as an outcome in OA trials and to link this pain intensity to work-related outcomes informative of work ability and productivity; however, the data we found did not enable us to conduct dependable analyses on these issues. While there is evidence that work-related outcomes are not frequently reported in journal article abstracts [12], they could conceivably still be reported more often in the full journal articles.

In systematic reviews, electronic database searching involves the use of search terms to identify different concepts and combination of these different concepts. This is followed by screening of the search results based on title and abstract, then full text articles are further assessed, and relevant articles that satisfy the eligibility criteria are included in the review. One major objective of systematic reviews is to capture all (or as large a proportion as possible) of the relevant publications regarding a subject through the database search [13, 14].

The aim of the present publication was to conduct a systematic review employing two distinct search strategies: one where work-related terms were included in the database search string (A) and one where this was not the case and work-related outcomes were identified instead by searches of full text Portable Document Formats (PDFs) (B). Search strategy A would conventionally be used and would only identify publications where work-related terms were mentioned in the title or abstract. Search strategy B presents the innovation of full text PDF searching and would identify publications were work-related terms were reported in the full text, regardless of whether they are mentioned in the title and abstract or not. We hypothesize that search strategy B identifies more relevant publications than search strategy A.

Search strategy A yielded 50 hits combined for hand, hip or knee OA, all mentioning some work-related concept in the title or the abstract (Fig. 1). Full text PDFs were not available for 12 papers (24% of the total) and these were therefore excluded. Of the remaining 38 full text PDFs that were screened, only two satisfied our inclusion criteria as shown in Fig. 2.

Search strategy B generated 986 articles of which there were no duplicates. Of these, 201 records (20% of the total) were excluded because the full text PDFs were not available to us. The remaining 785 full-text articles were screened by electronically searching the PDFs for the work-related terms and verifying the other study characteristics needed for inclusion. This resulted in 722 articles being excluded mainly because the articles did not mention any work-related concept or the articles were not randomized, double blind, placebo-controlled trials. The remaining 63 articles, which mentioned some work-related concept, were further screened to determine which articles actually evaluated and reported work-related concepts associated with OA in the respective study. As shown in Fig. 3, 51 studies were excluded at this stage resulting in 12 studies being retained. Furthermore, two studies [15, 16] did not report on pain intensity in relation to the reported work-related outcomes and were therefore excluded, leading to ten studies that were included in this systematic review. As a tool for assessing the degree of pain associated with the respective OA, six of the included studies [17‐22] employed a visual analogue scale (VAS). These studies, however, varied in the work-related outcome measures reported. The work-related outcome measures reported included resumption of work, number of days on sick leave, work performance, pain interference with work and work limitation. Of the four remaining included studies, one study [23] assessed pain intensity using a self-administered, validated questionnaire adopted from the McGill pain questionnaire, the SF-36 Health Survey, and the Medical College of Wisconsin noncancer pain questionnaire. Two of the studies [24, 25] employed the Brief Pain Inventory to assess pain, while one study [26] used a five-point pain scale.

Further comparison between the hits identified with search strategies A and B showed that the ten eligible articles that resulted from search strategy B also included the two eligible articles obtained from search strategy A. Additionally, all the available 38 full text PDFs resulting from search strategy A were also captured by results from search strategy B.

Table 1 lists the 51 excluded articles assessed as full text PDFs and the reasons for their exclusion. Of these 51 excluded articles, five concerned hand OA, 13 hip OA, and 33 knee OA.

Although 18 of the 51 excluded studies mentioned some work-related concept in the ‘Introduction’ section of the respective article, they subsequently did not report on any work-related outcome measure of the study participants. The work-related concepts mentioned in the ‘Introduction’ section of the respective article included work disability, inability to work, absenteeism from work, reduced participation in work, decreased work performance, lost productive work time and cost associated with lost work time. None of these work-related concepts were assessed in the participants of the respective studies.

Employment or job status was provided as part of the study participants’ baseline characteristics in 12 studies. The authors of these studies, however, did not report further on any work-related outcomes of the trials. Although some of the excluded studies made mention of some work-related concepts in the ‘Discussion’ and other sections of the articles (21 articles), these were, again, not directly related to the study participants.

Overall, 12 studies reported on the impact of OA and treatments for OA on work-related outcomes related to the study participants (Table 2). One study was on hand OA, two were on hip OA and nine were on knee OA. The particular type of work-related outcomes reported included ‘number of days on sick leave’ (one study) [18], decreased work performance (three studies) [22], limitation of work (four studies) [20‐22], pain interference with normal work (five studies) [20‐22], and improvement in normal work following intervention (two studies) [22]. In one study [17], resumption of work following therapeutic intervention for hand OA was reported. The ten studies ultimately included in our review (Table 2) reported pain intensity as well as work interference or time off work. The treatments investigated in the ten included studies varied and included naproxen, acetaminophen, celecoxib, rofecoxib, tramadol, duloxetine, soy protein, avocado/soybean unsaponifiables and glycosaminoglycan polysulphate.

These ten included articles demonstrated a considerable degree of variability in their reporting of both the work-related outcome measure and pain intensity. This prevented further analysis to determine the correlation between pain intensity and work-related outcome measures, which we had originally planned to do.

In summary, out of the 785 articles that could have reported on work-related outcomes in relation to pain intensity, only 1.3% (10/785) did so.

OA can place a substantial burden on an individual’s life and may lead to reduced productivity at work, interference with work, absenteeism and cost associated with lost work time [27, 28]. Trials on therapies for OA rarely consider the impact of the condition on work-related outcomes. The present systematic review aimed to evaluate the extent to which these work-related outcomes are reported in publications on randomized, double blind, placebo-controlled trials of hand, hip or knee OA. Our systematic review showed that only 1.3% (10/785) of full text articles reported on both, pain intensity (measured mostly by the visual analogue scale) and work interference or time off work.

This finding is consistent with other research, which demonstrated that work-related outcomes in pain trials are infrequently reported based on a database search of abstracts [12]. Chronic pain has been shown to negatively impact work-related outcomes by interfering with productivity and performance at the workplace [29], but this impact is often not captured in the outcomes reported in publications on intervention trials.

In the present review, two different search strategies, strategy A and strategy B, were compared. As described, strategy A followed the conventional search approach, which yielded two articles out of 50 possible identified records, which satisfied our inclusion criteria. By contrast, strategy B, which involved full text PDF searching, generated 986 records out of which ten articles were ultimately eligible. This demonstrates that full text PDF searching may be a useful strategy in circumstances when outcomes of interest are expected to not be reported in abstracts.

The ten included articles demonstrated a high degree of heterogeneity in the reporting of the pain assessment scale employed and the work-related outcome measures reported. Thus, it was not possible to perform any further analysis to determine the association between the pain intensity and work-related outcome measures, which we had originally planned to do. If future studies on OA were to follow similar approaches to assess pain intensity and work-related outcomes, this would enable meta-analysis of data from different studies. Standardization of the outcomes reported would therefore be of benefit.

Although some of the publications collected data on job or employment status of the study participants, the impact on work ability of the pain resulting from OA, and the treatments studied, is most often not reported. We suggest that future trials should put more emphasis on reporting such work-related outcomes in journal publications.

Our findings are limited by the fact that the search was conducted in only one database; potentially relevant articles may therefore have been missed. Stevinson et al. [13] demonstrated that searching only one database was not adequate to capture all possible publications on a given subject. Nonetheless, for our purposes in this systematic review, the results allow us to estimate the extent of under-reporting of work-related outcomes in randomized, double blind, placebo-controlled trials in OA, which is substantial.

Since reduction in pain intensity may lead to an improvement in an individual’s work ability [12], we recommend that treatment trials of OA should routinely report on the work-related outcomes of the study participants. This is in keeping with a recent publication to define an international standard set of outcome measures for patients with hip or knee osteoarthritis, which recommended the reporting of a patient’s work status at baseline and a regular follow-up to evaluate the impact of OA [30]. The recommendations also included various standardized instruments for evaluating pain and impact on overall quality of life. Further, a recent review suggested that core outcomes in OA trials should include time off work (sick leave), employment status, work productivity and work interference [31]. Although most of the included articles in the present review measured either one or more of the above work-related outcomes, the reporting of a particular outcome was inconsistent between the studies. As mentioned above, a standard for work-related outcome measures and reporting that adheres to the same would be goals for the future.

Work-related outcomes are inadequately reported in journal publication on randomized, double blind, placebo-controlled trials of hand, hip and knee OA. This is an area where future studies can do better.