A systematic review aims to comprehensively compile and synthesize evidence to answer an explicit research question using reproducible methods, and is often applied to randomized controlled trials. However, when the question of interest concerns etiology, randomized controlled trials are not only problematic to implement, they are often unethical (16). The study of etiology is therefore typically conducted using observational designs such as cross-sectional studies, case–control studies, or cohort studies, with systematic reviews increasingly being used to summarize these results (16). This systematic review of observational studies asks the question, “Is elevated general or regional adiposity related to tendon injury?”
Our outcomes of interest were 1) a clinical diagnosis of tendinopathy or tendon rupture and 2) abnormal tendon imaging on ultrasound or magnetic resonance imaging (MRI). The exposures of interest were 1) increased overall adiposity and 2) altered body fat distribution. Population-based studies typically use body mass index (BMI) to measure adiposity, and the World Health Organization defines overweight as a BMI between 25.0 and 29.9 kg/m2 and obesity as a BMI of ≥30 kg/m2 (17). Anthropometric measures such as waist circumference and waist to hip ratio are often used to quantify body fat distribution. Body composition can also be quantified using imaging methods. Modalities that distinguish fat from bone and lean tissue include MRI, computed tomography (CT), and dual x-ray absorptiometry (DXA). For reasons of expense, time, access, and radiation exposure (CT and DXA), these techniques are often limited to smaller studies. For clarity, we used adiposity throughout this report to describe increased adiposity or altered body fat distribution measured by any of these methods.
Identification and selection of the studies.
One author (JEG) conducted an electronic literature search in March 2007 using multiple databases (Allied and Complementary Medicine [1985 to March 2007], Biological Abstracts [1985 to February 2007], CINAHL [1982 to March week 1, 2007], Current Contents [week 27, 1993 to week 12, 2007], EMBase [1988 to week 10, 2007], Medline [1950 to February week 4, 2007], SPORTDiscus [1830 to November 2006], and Web of Science [1945 to March 13, 2007]) to identify studies examining associations between adiposity and tendon injury in humans. No limitation was applied regarding publication year, although only studies published in English were considered. The medical subject headings used included Tendons, Tendon Injuries, Tendinopathy, Tendinitis, Adiposity, Obesity, Body Mass Index, Anthropometry, Body Fat Distribution, and Adipose Tissue. Plain text searching was also used in combination with wildcards and truncation (18). Terms used for plain text searching included tend#nopath$, tend#nos#s, achil?odynia, obesity, adiposity, body composition, fat distribution, anthropometry, waist circumference, hip circumference, (waist adj2 hip) AND ratio, (abdominal OR visceral) adj5 adipose (for full details, see Supplemental Appendix A, available in the online version of this article at http://www3.interscience.wiley.com/journal/77005015/home). Electronic searches with broad scope (tendon AND obesity, tendon AND body composition, tendon AND adiposity) were conducted using the Google and Google Scholar search engines. The first 200 records for each search were screened for relevance by title.
Hand searching supplemented the electronic searches. Two authors (JEG, JLC) independently searched the reference lists of included articles for relevant studies. Additionally, 5 articles considered to be of key relevance (10, 11, 13, 14, 19) were identified on PubMed and the “Related Articles” function was selected. Both authors independently assessed the first 100 citations linked to each of the 5 key articles.
Records were imported into Endnote, version 9.0.0 for Macintosh (Thompson, Stamford, CT). Letters, conference proceedings, and duplicates were excluded. Two authors (JEG, JLC) then independently applied the predefined inclusion criteria to the titles and abstracts of the retrieved references. If both authors included the study or if there was disagreement, the full-text article was obtained and checked for eligibility. Disagreement regarding inclusion was resolved by discussion between the 2 authors.
To ensure the validity of the included data, only full and original articles from peer-reviewed journals were considered for inclusion. To be eligible, a study had to compare adiposity between individuals with and without tendon injury or examine the influence of adiposity on recovery from tendon injury. In both cases, the statistical methods had to be appropriate to the study design. Factors predicting successful outcome after surgical intervention for tendon injuries were excluded. Studies examining conservative management of tendon injury, however, were included if treatment was standardized. Carpal tunnel syndrome was excluded a priori because it differs from other tendon injuries in several fundamental aspects. Appropriate assessment of adiposity was considered to include height to weight ratios (e.g., BMI), waist circumference, waist to hip ratio, and adipose tissue volumes or cross-sectional areas as measured by CT, MRI, or DXA. Reporting body weight without adjusting for height is not a valid measure of adiposity and these studies were excluded.
The published version of each study was examined for methodologic quality using the checklist by Downs and Black (20). Two items suggested by the Centre for Reviews and Dissemination (21) were added. Two authors (JEG, MCA) marked each study, with a third author (JLC) making the final decision in the case of disagreement.
Data were extracted from each eligible study by 2 of the authors (JEG, MCA). The items were chosen for their ability to highlight the important aspects of study design, the demographics of the studied population, and the definition of disease. The methods used to measure adiposity were noted, as was the use of cutoffs (e.g., BMI >30 kg/m2). The association between adiposity and tendon injury was either recorded as an odds ratio or as an effect size (22), and the P value was recorded as calculated from presented data (23).
In circumstances where it was not possible to extract these data from the published manuscript, corresponding authors were contacted for the required information. Nine of the 14 corresponding authors contacted supplied the requested data, 2 stated that the data were no longer available, 2 responded but did not provide the requested data, and one author could not be contacted. In one case, all data were available except for the SD of BMI in the tendinopathy group (24). To allow interpretation, a conservative estimate (mean + 2 SDs) was calculated from the reported SD of BMI in the other studies. If more than one data point was missing, no values were substituted.
Because a single study may often report results for multiple groups (e.g., men with upper extremity tendinopathy, women with upper extremity tendinopathy, men with lower extremity tendinopathy, etc.), it was considered important to maintain these subpopulations for the purposes of this review. In contrast, where studies used multiple measures of adiposity (e.g., BMI, waist circumference, and waist to hip ratio) or examined the effect of different cutoffs (e.g., BMI >30 kg/m2 versus BMI >35 kg/m2) to report the results from a single subpopulation, it was important that these results were grouped together and only counted once. Similarly, where data for individuals with unilateral and bilateral tendinopathy were reported separately, online tools were used to combine the 2 groups (25). These tools allowed the input of means and SDs from 2 groups and then calculated the mean and SD of the 2 groups combined. This was performed for the variables of interest (i.e., age, BMI, waist circumference, and waist to hip ratio).
Any person participating in an included study was referred to as an individual. This term was further qualified according to whether they were an individual with a tendon injury (tendinopathy group) or an individual without a tendon injury (control group). Individuals recruited through their participation in organized sports were referred to as athletes, whereas those recruited through their place of employment were known as workers. Finally, those presenting to their health care practitioner for the management of tendon pain were referred to as patients. Of note is that athletes and workers (and those in the general community) often do not seek treatment for tendon pain and so, for example, a group of workers with tendon injuries was not always synonymous with a clinical patient group.
Where a study showed that individuals with a tendon injury had significantly greater levels of adiposity than controls, it was reported as a significant positive association. This term was also used when individuals had significantly greater abdominal adiposity (defined by elevated waist circumference or waist to hip ratio), because this distribution pattern is associated with increased incidence of CVD, chronic kidney disease, and type 2 DM (9). Conversely, when the opposite was shown (decreased general or abdominal adiposity), it was described as a significant negative association.
Sensitivity analyses were conducted based on 6 variables: sex (men, women, combined), population (athletes, patients, workers), injury location (upper extremity, lower extremity, not specified), adiposity measure (BMI, circumferences, other), study design (longitudinal, cross-sectional, case–control), and quality score (above middle rank, below middle rank). The effect of each of these grouping scenarios was tested with a chi-square test (SPSS, version 15.0.1 for Windows; SPSS, Chicago, IL). P values less than 0.05 were considered significant.