Abstract
Background The inclusion of sociodemographic and anthropometric variables in published clinical trials enables physical therapists to determine the applicability of trial results to patients in their clinics.
Objective The aim of this study was to examine the reporting of participant sociodemographic and anthropometric characteristics in published physical therapy–related clinical trials.
Design This was a retrospective review of clinical trials from 2 samples drawn from literature applicable to physical therapy.
Methods Two reviewers independently extracted data from a random sample of 152 clinical trials from the Physiotherapy Evidence Database (PEDro) and a purposive sample of 85 clinical trials published in the journal Physical Therapy (PTJ). A database containing the occurrence of sociodemographic (age, sex, race/ethnicity, level of education, marital status) and anthropometric variables (height, weight, body mass index) in each article was created to generate descriptive statistics about both samples.
Results Among the sociodemographic variables, at least 90% of articles reported the sex and age of trial participants. Additional sociodemographic characteristics that were reported in 20% to 26% of articles were participant level of education and participant race/ethnicity. The reporting of anthropometric data differed between the 2 samples, with body mass index being most commonly reported in the PEDro sample (48.0%) and weight being most commonly reported in the PTJ sample (38.8%).
Limitations Articles reviewed were limited by year of publication (from 2008 to 2012 for PTJ articles and 2010 for clinical trials from PEDro) and to English-language–only literature.
Conclusions The physical therapy literature would benefit from enhanced reporting requirements for both sociodemographic and anthropometric data about participants.
With the current emphasis on evidence-based practice, clinical trial outcomes play an important role in health care delivery. Well-designed and well-executed clinical trials form the basis of systematic reviews and clinical guidelines that, in turn, inform clinical practice. The results of clinical trials may not be applicable to patients who are underrepresented or not represented in the original study samples.1–3 A physical therapist with a clinical question is advised to use the evidence-based patient/intervention/comparison/outcome (PICO) format that is advocated in most textbooks on evidence-based practice and is used in the Linking Evidence And Practice (LEAP) series, published in the journal Physical Therapy (PTJ).4–7 The “P” component of PICO may be challenging if subpopulations are not defined,4 and, if relevant anthropometric or demographic details are not included, incorrect estimates of treatment effects may occur.4 Conversely, homogeneity of clinical trial participants in terms of sociodemographic background or anthropometric characteristics limits the generalizability of studies and forces the “P” to be more generic in nature. Thus, a therapist may need a more defined “P” in the PICO but not be able to find sufficient participant descriptors in the physical therapy clinical trial literature to answer a question related to a more restrictive “P.”
For example, a physical therapist may have a patient with neck pain for whom it appears manipulation is an appropriate intervention.8 Jonely and Scalzitti8 provide guidance for physical therapists through their LEAP series article. Does the evidence to support manipulation also support it in the case of a patient who is morbidly obese or a patient from a lower socioeconomic background? A physical descriptor such as “obesity” may be important to the patient description when a physical intervention is studied. Social and demographic characteristics that have been shown to be determinants of health, pain, and disability among people with musculoskeletal disorders also may be important elements of the patient description.9,10 The therapist in this instance would need to limit the “P” to the same one used in the article,8 which was a “36-year-old woman with neck pain.” Although anthropometric and demographic characteristics may be important factors in the decision-making process, subgroups created from these characteristics were not part of the original clinical trial data and were not part of the analysis for the Cochrane review that formed the basis of the LEAP article.11 Jonely and Scalzitti did specify “woman”; however, the Cochrane review on which this LEAP article was based11 provided no demographic details that would enable application specifically to women as a demographic subgroup.
In a few instances, subgroup analysis has been completed by sociodemographic factors in literature related to physical therapy. For patients with orthopedic diagnoses, the current literature suggests that racial/ethnic differences may affect outcomes.12–16 For example, Ibrahim et al16 found that African American patients had significantly higher rates of complications following knee arthroplasty than white patients, and rates of infection-related complications were higher in Hispanic patients than in white patients. These differences between racial/ethnic groups were not seen following hip arthroplasty. Differences in pain perception also have been found between men and women and among racial/ethnic groups.13,14,17 When examining performance on 4 commonly used clinical tests (Six-Minute Walk Test, Berg Balance Scale, Timed “Up & Go” Test, and comfortable- and fast-speed walking), Steffen et al18 found both age- and sex-related differences. Although their study was not a randomized controlled trial, it does help illustrate the need to differentiate results by age and sex.
When sociodemographic details of participants have been described in other medical fields, disparities in the inclusion and reporting of sociodemographic characteristics of sex, race, and ethnicity have been documented.1–3,19–23 Furthermore, the reporting of sociodemographic and anthropometric characteristics in published clinical trials is essential for physical therapists to determine the applicability of trial results to the patients in their clinics.
In 1993, the National Institutes of Health (NIH) Revitalization Act mandated standards for adequate representation of women and minority groups in all NIH-funded clinical research.24 The guidelines published in the Federal Register mandated the inclusion, analysis, and reporting of data on women, minorities, and their subpopulations “to ensure that all future NIH-supported biomedical and behavioral research involving human participants will be carried out in a manner sufficient to elicit information about individuals of both sexes and the diverse racial and ethnic groups and, in the case of clinical trials, to examine differential effects on such groups.”25 Although some progress has been made to include more women and minorities in clinical trials and to report sex- and race/ethnicity-specific results, these groups are still underrepresented, and only a minority of studies report sex and race/ethnicity-specific results.1–3,20,22,23,26–28 Enrollment and reporting are especially important in diseases and conditions with known health disparities.1
Understanding racial/ethnic variations related to health conditions is essential to meet the health care needs of an increasingly diverse population.1,2 Previous studies examining the inclusion and reporting of minority groups in clinical trials have focused on publications in medical journals,1,2,20–22,26–28 trials related to a specific diagnosis (eg, cancer),19 or trials related to a specific procedure (eg, vascular surgery).3 Little is known about the inclusion and reporting of this information in the literature related to physical therapist practice. The primary purpose of this study was to examine the reporting of participant sociodemographic and anthropometric characteristics in published clinical trials applicable to the practice of physical therapy. A secondary purpose was to determine if sample size or receipt of NIH funding influences reporting. We hypothesized that studies with larger samples may have sociodemographic diversity and that studies with NIH funding would, in compliance with the mandated standards,25 incorporate samples with more women and minorities.
Method
Data Sources
We developed 2 samples for this study: a randomly selected sample and a purposive sample (Fig. 1). The intent in having 2 samples was to develop one sample from PEDro-indexed articles that evaluated reporting of characteristics with a broad view based on literature that affects physical therapy but is interdisciplinary and another sample that enabled us to specifically look at literature directed toward physical therapists.
Flowchart of sample development. PEDro=Physiotherapy Evidence Database, PTJ=Physical Therapy journal.
To develop the randomly selected sample, we used the Physiotherapy Evidence Database (PEDro). We selected PEDro because of its strength as a bibliographic database that indexes citations to clinical trials in physical therapy from multiple journals and multiple countries.29 In May 2013, using Microsoft Excel 2010 software (Microsoft Corp, Redmond, Washington), we developed a database comprising the 760 articles indexed in PEDro. Inclusion criteria were: publication in 2010, publication type of clinical trial, English language, and a PEDro score of 5 or better. Conceptually, we believe that articles with a PEDro score of 5 or better are of higher quality, are more likely to be influential to the clinical community, and are more likely to be included in subsequent systematic reviews.30 We used articles published in 2010 because at the time our study was conducted, 2010 was the most current year with the greatest proportion of articles both reviewed and rated in the PEDro database. Mirroring the methodology of Pinto et al,31 a 20% sample (n=152) was then randomly selected from the database using the Excel 2010 random number function. Full-text copies were obtained of each article selected. Three of the original articles selected at random were replaced with the next randomly generated article—2 because we were unable to obtain copies of the original articles and 1 because it was retracted.
The purposive sample was developed by creating a database of all clinical trials published in PTJ from 2008 to 2012. The 5-year period was selected to maximize sample size. This sample represents a single influential journal likely to contain articles that were NIH funded, and we felt that including a sample of potentially NIH-funded articles would enable us to see the influence of NIH requirements on reporting of demographic information. Articles were identified through a MEDLINE search limited to clinical trials, time period, and journal name. The MEDLINE search produced 101 articles. Sixteen of these articles were dropped from analysis because they were nested studies or secondary analyses of trial data but not clinical trials themselves, resulting in a final sample of 85 articles. This sample of 85 articles overlapped with the PEDro sample in only 3 instances.
Data Extraction and Quality Assessment
The 2 samples of articles underwent a structured independent review by 2 researchers. Each researcher independently abstracted the following sociodemographic data from each article: number of male and female participants; mean or median age of participants; and reporting of participant age, sex, race, ethnicity, income or income status, education, marital status or habitation status, and employment status. To be counted as present in an article, a variable needed to be enumerated either in a table describing sample characteristics or in the body of the article. If the authors mentioned in the Method section that the data were collected but did not report the values for a variable, the variable was not counted as present. We used the characteristics listed above because we felt they best represented social determinants of health. If a study analyzed only women or any other demographic subgroup as the target population, we noted that during the abstraction process. Anthropometric data abstracted included participant height, weight, body mass index (BMI), percent body fat, and waist circumference. Additional data noted during the abstraction process included sociodemographics or anthropometrics beyond our original listing; the target population of the study; the location/country in which the study was undertaken; and, for the articles published in PTJ, the funding source, if any. The 2 researchers met to resolve any discrepancies in their abstraction work. As an assurance of the reliability of the extraction work, 75 articles randomly selected from the PEDro sample underwent an additional extraction by a third, independent researcher, who concurred exactly with the original work in 87% of the articles. The 10 articles in which there was a discrepancy in the data collected were readjudicated by the 2 researchers. The final adjudicated data were entered into the Excel 2010-generated database.
Data Synthesis and Analysis
Characteristics reported in the 2 samples were summarized using proportions and frequency tables. Pearson correlation coefficients were calculated to examine relationships between the number of participant characteristics reported and trial sample size.
Results
The majority of the PEDro sample of 152 articles were studies completed in Europe (37.5%) or North America (28.9%). See Table 1 for sample characteristics. The sum of the participants of all the studies totaled 27,231, with a range of 8 to 5,391 participants in each study. There was no relationship between the size of the study sample and the number of participant characteristics reported (r=.05, P=.513). The most commonly reported sociodemographic characteristics were age and sex, which were reported in 98.7% and 96.7% of studies, respectively (Tab. 2). The most commonly reported anthropometric characteristic was BMI, which was reported in 48% of the studies.
Characteristics of Sample Articlesa
Demographic and Anthropometric Reportinga
The PTJ sample of 85 articles included studies predominantly completed in North America (61.2%). Of these 85 articles, 18 (21.2%) had some form of funding from NIH. The number of participants represented by the PTJ sample totaled 6,892, with a range of 11 to 1,141 participants in each study. In the PTJ sample, age was consistently reported in all studies, and sex was reported in 94.1% of the studies. In terms of anthropometric data, both height and weight were reported more often than BMI. Among the 18 studies that had NIH funding, 10 reported only the sociodemographic characteristics of age and sex.
Discussion
Sociodemographic and anthropometric characteristics beyond age, sex, and BMI are poorly reported in the physical therapy–related clinical trial literature. The low level of reporting has implications for the generalizability of trial findings and limits the ability of physical therapists to determine the applicability of trial research to any social or demographic subgroup beyond those defined by age or sex. The deficiency of anthropometric characteristic reporting is of note because physical therapy trials often involve physical interventions, including exercise, which may have differential outcomes based on body characteristics of the participants.32 In terms of the generalizability of these trials, the national trend of increasingly overweight and obese individuals in the general population33 leads us to wish for better anthropometric reporting.
Reporting and possible inclusion of minorities by race and ethnicity are distinctly low in physical therapy–related clinical trials. In Figure 2, we propose a model by which the consistent reporting of sociodemographic and anthropometric characteristics is more likely to occur in physical therapy literature. The model is dependent upon a clear understanding of the demographic makeup of the researcher's target population, the development of an inclusive sampling frame, and the subsequent intentional enrollment of women and minorities in the sample. We can only hypothesize about enrollment and inclusion in our study because our analysis was limited to data reported.
Model to ensure inclusion of women and minorities and reporting of trial participant characteristics.
The small sample sizes we found in the reviewed articles may explain the lack of reporting of demographic information. Somerson et al1 found that studies with larger sample sizes were more likely to report race/ethnicity information. In our study, even the articles that address special populations, especially those that are related to minority groups, have samples that are very small in number. The lack of reporting persists whether the literature derives from North America, where NIH funding mandates inclusion, or from other geographic regions. In the PEDro sample, race/ethnicity was reported in 32% of the trials. In the PTJ sample, race/ethnicity was reported in 17% of the trials. Somerson et al1 found that only 20.3% of 158 orthopedic randomized clinical trials reported on race/ethnicity. Berger et al20 examined randomized clinical trials cited by the American Heart Association guidelines for cardiovascular disease and found race/ethnicity was reported in 35% of the trials. Hoel et al,3 in a study of vascular trials, found that only 21% reported on race/ethnicity. Clearly, the physical therapy clinical trial literature is in good company, but we are compelled to ask if this is the company we want to keep.
We acknowledge that our study does have some limitations. Although we were careful in our methods of trial identification, we limited our samples to those in the PEDro database and to only PTJ. We recognize that trials represent only a portion of the physical therapy–related literature and that there are studies of other types that may provide lower levels of evidence that were not included. In this study, we selected clinical trials because we believe they have the most influence on clinical decision making, as they are used in both systematic reviews and clinical guidelines. Finally, we limited our efforts to only the English-language literature.
Although the majority of the articles we analyzed did report on sex composition, subgroup analysis by sex was not necessarily included. Men and women should be analyzed separately unless no differences are found on the measured variables. Future investigations on reported characteristics of participants could include recording of subgroup analysis for sex and other variables.
We hope to promote change by drawing attention to the low reporting and low inclusion of sociodemographic variables (in particular, to race). Journal editors and peer reviewers have the potential to change both the rates of inclusion and the rates of reporting by requiring that these characteristics be present in submitted manuscripts, even when the numbers for some groups are zero. We encourage modifications to the instructions to authors in our scientific journals to highlight the importance of incorporating these characteristics in submissions. Two obvious reasons for not reporting sociodemographic and anthropometric characteristics are failure to collect the data and failure to enroll diverse participant pools. Study investigators should make efforts to enroll minorities and women in physical therapy–related clinical trials and to collect and report data for sociodemographic and anthropometric variables.
Footnotes
Both authors provided concept/idea/research design, writing, data collection and analysis, and project management. The authors thank Maggie Daingerfield, DPT, and Tiffany Chin for their assistance in data management and review.
- Received June 4, 2014.
- Accepted January 2, 2015.
- © 2015 American Physical Therapy Association