Advertisement

Many Randomized Trials of Physical Therapy Interventions Are Not Adequately Registered: A Survey of 200 Published Trials

  1. Rafael Zambelli Pinto,
  2. Mark R. Elkins,
  3. Anne M. Moseley,
  4. Catherine Sherrington,
  5. Robert D. Herbert,
  6. Christopher G. Maher,
  7. Paulo H. Ferreira and
  8. Manuela L. Ferreira
  1. R.Z. Pinto, MSc, The George Institute for Global Health and Sydney Medical School, University of Sydney, PO Box M201, Missenden Road, Sydney, New South Wales 2050, Australia.
  2. M.R. Elkins, PhD, Department of Respiratory Medicine, Royal Prince Alfred Hospital, Sydney, New South Wales, Australia, and The George Institute for Global Health and Sydney Medical School, University of Sydney.
  3. A.M. Moseley, PhD, The George Institute for Global Health and Sydney Medical School, University of Sydney.
  4. C. Sherrington, PhD, The George Institute for Global Health and Sydney Medical School, University of Sydney.
  5. R.D. Herbert, PhD, Neuroscience Research Australia, Sydney, New South Wales, Australia, and The George Institute for Global Health and Sydney Medical School, University of Sydney.
  6. C.G. Maher, PhD, The George Institute for Global Health and Sydney Medical School, University of Sydney, and EMGO+ Institute for Health and Care Research, VU University Medical Centre, Amsterdam, the Netherlands.
  7. P.H. Ferreira, PhD, Discipline of Physiotherapy, Faculty of Health Sciences, University of Sydney.
  8. M.L. Ferreira, PhD, The George Institute for Global Health and Sydney Medical School, University of Sydney.
  1. Address all correspondence to Mr Pinto at: rafaelzambelli{at}gmail.com.

Abstract

Background Clinical trial registration has several putative benefits: prevention of selective reporting, avoidance of duplication, encouragement of participation, and facilitation of reviews. Previous surveys suggest that most trials are registered. However, these surveys examined only trials in journals with high impact factors, which may bias the results.

Purpose This study examined the completeness of clinical trial registration and the extent of selective reporting of outcomes in a random sample of published randomized trials in physical therapy.

Data Sources This was a retrospective cohort study in which 200 randomized trials of physical therapy interventions were randomly selected from those published in 2009 and indexed in the Physiotherapy Evidence Database (PEDro), regardless of the publishing journal.

Data Extraction Evidence of registration was sought for each trial in the study, on clinical trial registers, and by contacting authors.

Data Synthesis The proportion of randomized trials that were registered was 67/200 (34%). This proportion was significantly lower than among the trials in journals with high impact factors, where the proportion was 75% (odds ratio=7.4, 95% confidence interval=2.6–21.4). Unambiguous primary outcomes (ie, method and time points of measurement clearly defined in the trial registry entry) were registered for 32 trials, and registration was adequate (ie, prospective with unambiguous primary outcomes) for 5/200 (2.5%) trials. Selective outcome reporting occurred in 23 (47%) of the 49 trials in which selective reporting was assessable.

Limitations The inclusion of only English-language trials prevents generalization of the results to non–English-language trials.

Conclusions Registration of randomized trials of physical therapy interventions is rarely adequate. Consequently, the putative benefits of registration are not being fully realized.

Interpretation of scientific research by clinicians, patients, and policy makers can be biased because completed research often is left unpublished.1 There also is a tendency for trials with statistically significant or positive findings to be published more frequently, and more quickly, than trials with negative or null findings.1,2 Both of these phenomena, selective submission and acceptance of reports based on the strength and direction of the trial results, are forms of scientific misconduct3 and have been termed “publication bias.”1 Recently, an additional problem within published studies has been identified: the selective reporting of outcomes.4,5 One strategy to deal with at least part of these problems is the prospective registration of trials at inception.6,7

Prospective registration of clinical trials also is believed to be advantageous to avoid unnecessary duplication of trials, facilitate participation in trials by volunteers, and facilitate the location of data for systematic reviews.8 The minimum amount of trial information required to consider a trial fully registered has been defined by the World Health Organization (WHO) as the WHO Trial Registration Data Set (Tab. 1).9 Although clinical trial registration was recommended as early as 1993,10 the first major register for health care trials did not become available until 1998,11 and registration did not become common until the International Committee of Medical Journal Editors initiated a policy making it mandatory in 2005.11 Soon after this, clinical trial registration was recommended by other research disciplines, including allied health,12,13 dentistry,14,15 and health informatics.16 However, no formal process has been established to monitor whether the published versions of trials match their registry entries. Therefore, it is important to assess whether clinical trial registration is being undertaken and, when it is undertaken, whether it prevents selective reporting.

View this table:
Table 1.

Summary of the World Health Organization (WHO) Trial Registration Data Set (Version 1.2.1)

Several authors have surveyed the prevalence of registration among clinical trials of medical interventions. In these surveys, the proportion of published clinical trials that were registered ranged from 59% to 97%.17,18 These data suggest it is important to examine the prevalence of registration within individual disciplines. These estimates of the prevalence of registration in medical journals may have been influenced by the use of only trials from journals with the highest impact factors.17,18 Such journals were among the first to require trial registration,11 and they publish trials of significantly higher quality.19 This sampling may tend to inflate the estimate of clinical trial registration among trials in these various medical disciplines. However, the effect of this potential source of bias has not been assessed objectively.

A recent systematic review concluded that discrepancies among registry entries, study protocols, and published reports are common.20 In the surveys included in the review, the proportion of trials with inconsistency in the primary outcomes between the registry entry or study protocol and the published report ranged from 30% to 67%.4,17,18,21–24 Again, some of these estimates also may be influenced by the fact that the surveys analyzed only trials from journals with a high impact factor.17,18

An analysis of clinical trial registration among published trials could avoid potential biases such as the one described above by using a representative sample of all randomized trials published in a particular discipline. This representative sampling would require that the trials were not chosen because they have been published in particular journals (such as those with high impact factors or those listed in MEDLINE). This sampling can be achieved for randomized trials of physical therapy interventions because the Physiotherapy Evidence Database (PEDro) (http://www.pedro.org.au/) indexes such trials regardless of whether they are registered, the journal in which they are published, the electronic databases in which they are indexed, or their methodological quality.25 Coverage of randomized trials in physical therapy indexed in PEDro is equivalent to or better than 7 other databases, including PubMed and CENTRAL.26,27 In December 2011, more than 16,500 randomized trials were indexed in PEDro.

The aim of this study was to examine the completeness of registration and the extent of selective reporting of outcomes in a sample of reports of randomized trials in physical therapy. Specifically, the study sought to determine: (1) the proportion of published randomized trials in physical therapy that are registered and, among these, the proportion that can be considered to have adequate registration; (2) whether the proportion of registered randomized trials is higher among trials published in journals with the highest impact factors; (3) whether randomized trials with certain characteristics are more likely to have been registered; (4) the proportion of trial registry entries that have ambiguous primary outcomes; (5) whether there is consistency between registry entries and published reports regarding the primary and nonprimary outcomes; and (6) the proportion of registered randomized trials that selectively report outcomes and whether selective reporting favors statistically significant results.

Method

Study Selection

We evaluated reports of randomized trials that were indexed in PEDro, written in English, and published in 2009. Trials were not excluded on the basis of whether they were described as a pilot trial or whether the study design was parallel or crossover. Twenty percent of these trial reports were randomly selected using the random number function in Microsoft Excel software (Microsoft Office 2007, Microsoft Corporation, Redmond, Washington), and the citation details were downloaded from PEDro. The PEDro score28,29 of methodological quality for each trial also was downloaded. Although the use of trial quality scores often is discouraged because the validity of their summary scores has been questioned,30 the PEDro score has been validated as a continuous measure of trial quality.28 Full-text copies of each trial report were retrieved. For each of these reports, one investigator extracted data on the sample size (ie, number of participants randomized), the country where the randomized trial was conducted, whether the trial was funded, and whether the trial was multi-centered or single-centered and downloaded the 2010 impact factor of its publishing journal.31 Trial reports also were categorized as published or not published in a specific physical therapy journal. All 64 member journals of the International Society of Physiotherapy Journal Editors32 were considered specific physical therapy journals. The PEDro codes regarding the subdiscipline in which each trial belongs also were downloaded. The 11 possible categories are indexed terms for searching the database (further details for each category are available at http://www.pedro.org.au/english/downloads/codes/). As each trial report can be assigned more than one code, one investigator was responsible for deciding the most appropriate category for those trial reports with more than one code.

The trial report was checked for a statement regarding registration or a registration number. If neither was identified, the corresponding author was sent an e-mail message requesting details of any registration. A reminder e-mail was sent 2 weeks later if necessary. In the absence of a reply, the following registers were searched: ClinicalTrials.gov, the International Standard Randomized Controlled Trial Number Register, the Australian New Zealand Clinical Trials Registry (because these 3 registers are large and contain trial protocols from many countries), and the national register of the country of origin of each author. This search was aided by the WHO registry search portal. Within each of these registers, the investigator searched for the citation details of the published report, including the title of the published report, any funding sources, and the first, second, and last authors. Registry entries were confirmed as being related to the published report by matching author, experimental and control interventions, trial name, and country of origin. Trials for which evidence of registration was not found with this procedure were considered unregistered. Trial reports that described only the secondary analysis of the original randomized trial were excluded from the remaining analyses.

Adequacy of Registration

For each registered randomized trial, 2 investigators independently examined the trial registry entry (including its history of amendments) and the published trial report and extracted the following information: registration date, start and end dates of enrollment, primary outcomes, other outcomes, definition of each outcome, time points of assessment for each outcome, and statistical analyses for each outcome. An outcome was defined as a variable that was used to compare randomized groups to assess the efficacy or harm of the intervention. Outcomes were classified as primary in the registry entry or in the published report if they were explicitly described as “primary,” “key,” or “main” outcomes in the relevant document. Otherwise, outcomes were considered nonprimary. If no primary outcome was specified for a trial, all outcomes were considered nonprimary. The investigators who extracted this information had been trained in, and were experienced in, rating the quality of randomized trials. Disagreements were resolved by discussion.

The registration date and the start and end dates of enrollment were used to determine whether registered randomized trials were registered prospectively, concurrently, or retrospectively. Prospective registration was defined as registration of the trial before data collection began and remaining unchanged thereafter. Concurrent registration occurred during the period in which participants were enrolled or data were collected. Retrospective registration occurred after data collection was complete. We also classified trials as having adequate registration if they were prospectively registered with unambiguous primary outcomes.18 Primary outcomes were considered unambiguous if the trial registry entry specified both the definition (ie, measurement and metric as defined by Zarin et al33) and time points for analysis. For example, “lung function” would be considered ambiguous on both criteria, whereas “change in forced vital capacity at 3 months” would be considered unambiguous.

To determine whether the proportion of registered randomized trials is higher among those published in journals with the highest impact factors, we ordered the cohort of 200 randomized trials by the impact factor of the journal in which they were published. We then selected the top 10% of this list and calculated the proportion of randomized trials that were registered and compared this proportion with the proportion among the remainder of the trials.

Inconsistency of Outcomes

The 2 investigators also determined which randomized trials had primary outcomes that were inconsistent between the most recent registry entry (ie, including amendments) and the published report. Discrepancies were classified according to the criteria of Chan et al21 as modified by Mathieu et al18: (1) the registered primary outcome was reported as a nonprimary outcome in the published report, (2) the registered primary outcome was omitted in the published report, (3) a new primary outcome was introduced in the published report (ie, a registered nonprimary outcome became a primary outcome in the published report or an outcome that did not appear at all in the registry entry was introduced as a primary outcome in the published report), (4) the published primary outcome was described as a nonprimary outcome in the registry entry (ie, a subset of category 3), and (5) the timing of assessment of the registered and published primary outcomes differed. Randomized trials also were classified according to whether the nonprimary outcomes were inconsistent between the trial registry entry and the published report. Nonprimary outcomes were considered inconsistent when the nonprimary outcomes listed in the registry entry did not match those listed in the published report. If there were any disagreements at this stage, investigators again reached a consensus through discussion.

Selective Reporting of Outcomes

Selective reporting of outcomes was considered objectively assessable only among those registered randomized trials that had unambiguous primary outcomes. Selective reporting was considered to have occurred when at least 1 of the 5 discrepancies described above was present. Where selective reporting was identified, the 2 investigators recorded together whether it affected the statistical significance of the primary outcome.

Data Synthesis and Analyses

Proportions were used to summarize the characteristics of all randomized trials. Unpaired t tests were used to compare the quality scores and sample sizes between the registered and unregistered randomized trials, and the results were reported as mean differences with 95% confidence intervals. The Mann-Whitney U test was used to compare the quality scores and sample sizes between the registered and unregistered randomized trials if they were not normally distributed. The associations between trial characteristics (continent, funding, number of sites, and impact factor for the publishing journal) and registration were assessed using odds ratios. We listed the registers used and the number and proportion of randomized trials registered in each.

Results

General Characteristics of Selected Articles

On October 21, 2010, PEDro had indexed 927 reports that were coded as randomized trials published in English in 2009. We selected a random sample of 200 of these reports (eAppendix). The characteristics (subdiscipline and PEDro score) of the sample of 200 randomized trials were similar to those of all trials in PEDro published in 2009 (Fig. 1). Sixty-seven (34%) of the 200 randomized trials of physical therapy interventions were registered (Fig. 2). Of the 67 registered randomized trials, 48 (72%) stated in the published trial report that the trial was registered. The remaining trials were identified as registered through contacting authors (n=17, 25%) or hand-searching trial registries (n=2, 3%).

Figure 1.
Figure 1.

Graphs supporting the representativeness of the sample used in this study based upon subdiscipline of physical therapy and Physiotherapy Evidence Database (PEDro) score. The blue bars represent the sample of 200 trials, and the red bars the total 927 trials indexed in PEDro in 2009. (A) Proportion of trials coded in each PEDro subdiscipline. (B) Proportion of trials according to the PEDro score. Sport=sports, Peds=pediatrics, Ortho=orthopedics, Onco=oncology, Neuro=neurology, Musc=musculoskeletal, Ger=gerontology, E&OH=ergonomics and occupational health, C&WH=continence and women's health, CP=cardiopulmonary, N/A=no applicable subdiscipline.

Figure 2.
Figure 2.

Flowchart of manuscript selection and assessment of registration status. PEDro=Physiotherapy Evidence Database

A description of all included randomized trials, as well as separate description of registered and unregistered randomized trials, is provided in Table 2. The mean (SD) quality score out of 10 was 5.9 (1.6) for the registered randomized trials and 5.4 (1.4) for the unregistered randomized trials (mean difference=0.6, 95% confidence interval=0.1–1.0). The median (IQR) sample size was 97 (50–168) for the registered randomized trials and 48 (28–74) for the unregistered randomized trials (median difference=49) (P<.001). Figure 3 shows odds ratios for the association between trial characteristics (continent, funding, and number of sites) and registration.

View this table:
Table 2.

Characteristics of All of the Randomized Trials (N=200), the Unregistered Randomized Trials (n=133), and the Registered Randomized Trials (n=67)

Figure 3.
Figure 3.

Factors associated with trial registration (odds ratios and 95% confidence intervals [95% CI]). Dashed lines indicate the reference group used in the odds ratio analysis

Of the 67 randomized trials that had been registered, most were registered with ClinicalTrials.gov (n=34, 51%), followed by the Australian New Zealand Clinical Trials Registry (n=20, 30%) and the International Standard Randomized Controlled Trial Number Register (n=11, 16%). In addition, 2 other registries (StrokeCenter.org and TrialRegister.nl) each had 1 randomized trial (1%) registered. Only 1 randomized trial had more than 1 registry entry, and both entries contained the same information. Six (9%) of the 67 randomized trials stated in the published report that they also had published the trial protocol in a peer-reviewed journal.

When the randomized trials were ranked according to the impact factor of their publishing journal, 31 had no impact factor and the remaining 169 had a median (IQR) impact factor of 2 (2–4). Among the top 10% of these ranked trials, the median (IQR) impact factor was 12 (7–30). The proportion of randomized trials that were registered among only these top-ranked trials was 75%, which was significantly higher than among the remaining trials (odds ratio=7.4, 95% confidence interval=2.6–21.4).

Of the 67 reports of registered randomized trials, 5 were secondary reports of the original study (Fig. 2). These 5 trials were excluded from analyses of adequacy of registration, inconsistency of outcomes, and selective reporting (Tab. 3).

View this table:
Table 3.

Characteristics of the Registered Randomized Trials in the Study Cohort That Reported the Primary Analysis (n=62)

Adequacy of Registration

Of the 200 randomized trials evaluated, only 12 (6%) had been registered prospectively (Tab. 3). The proportions of trial registry entries that defined the primary outcome, its method of analysis, and the time points at which it would be assessed are presented in Table 3. Thirty-two trial registry entries (16%) specified unambiguous primary outcomes (ie, included both a definition and time points for analysis). Only 5 randomized trials were considered adequately registered (ie, prospectively registered and unambiguous primary outcomes). These 5 adequately registered randomized trials represent only 2.5% of the 200 trials evaluated.

Inconsistency of Outcomes

Consistency of reporting outcomes was analyzed by comparing the registry entry with the published report. At least one major discrepancy of primary outcome was found in 34 registered randomized trials (55%). Table 3 presents the proportion of randomized trials in each category of discrepancy. For the nonprimary outcomes, 53 (85%) of the registered randomized trials had differences between the trial registry entries and the published report.

Selective Reporting of Outcomes

Of the 62 registered randomized trials, 13 (21%) did not explicitly define the primary and nonprimary outcomes in the published report (Tab. 3). Thus, the remaining 49 randomized trials were used in the analysis of selective reporting. Of the 49 randomized trials assessed, 23 (47%) showed evidence of selective reporting. Among 18 of these 23 randomized trials, the effect of the selective reporting on the statistical significance of the primary outcome was assessable. The selective reporting caused the primary outcome to change from nonsignificant to significant in 9 trials (50%) but did not cause a change from significant to nonsignificant in any trial. The significance did not change in the remaining 9 trials (50%).

Discussion

These data suggest that the registration of randomized trials of physical therapy interventions is inadequate to achieve many of its intended benefits. Only a third of randomized trials were registered out of the sample of 200 trials. Such a low rate of registration is unlikely to prevent selective reporting of trials or to facilitate the location of data for systematic reviews. Only one fifth of registered randomized trials were registered prospectively. This rate appears to be too low to prevent unnecessary duplication or to facilitate participation in trials by volunteers. Only one fiftieth of all randomized trials were adequately registered (ie, prospectively, with an unambiguous primary outcome), so the selective reporting of outcomes is not substantially prevented either. This finding was confirmed by the analysis of the small number of randomized trials where selective outcome reporting could be assessed. More than half of these trials had discrepancies between the registered and the reported primary outcomes, and almost all had discrepancies in the other outcomes. Selective outcome reporting often altered the statistical significance of the primary outcome, almost always introducing a significant primary outcome when the original primary outcome would not have been significant.

The prevalence of trial registration observed in our study was 34%. We believe this is an accurate estimate of the prevalence among published randomized trials in physical therapy because the trials were randomly sampled. The sample size of 200 also was substantial (20% of the randomized trials in PEDro published in 2009) and similar to the sample sizes of the previous surveys of registration, which ranged from 158 to 323.17,18 Because we included only English-language trials, we do not know whether the results are generalizable to non–English-language trials. However, non–English-language trials made up only 8% of trials in PEDro published in 2009.

Previous surveys of registration among clinical trials of medical interventions have estimated the prevalence of registration as 59% to 97%.17,18 However, these surveys included only trials from journals with the highest impact factors (median [IQR] impact factor=10 [8–26]). In our survey, the corresponding estimate using the 10% of trials from journals with the highest impact factors (median [IQR]=12 [7–30]) was 75%. In addition, among these journals with the highest impact factors, only 1 out of 14 journals did not require trial registration. This finding suggests that clinical trial registration may be similarly prevalent in physical therapy and the medical disciplines examined in the earlier studies. This result also suggests that the estimates of the prevalence of clinical trial registration from existing surveys of medical trials may be inflated by the sampling of publications from high impact journals only.

The paucity of registration among randomized trials in physical therapy could be explained by a lack of awareness among authors. Although clinical trial registration was recommended as early as 1993, it was more than a decade until the International Committee of Medical Journal Editors officially announced their policy.11 Consequently, it is likely that some trials published in 2009 had started recruitment before the official implementation of this policy by any physical therapy journals. Therefore, analysis of reports published in 2009 might be considered too soon to assess the impact of the registration policy. However, even in 2012, registration is compulsory for only 10 (16%) of the 64 member journals of the International Society of Physiotherapy Journal Editors: Physical Therapy, Physiotherapy, Physiotherapy Theory and Practice, Journal of Physiotherapy, Journal of Physical Therapy, Journal of Manual and Manipulative Therapy, Journal of Orthopaedic and Sports Physical Therapy, Brazilian Journal of Physical Therapy, New Zealand Journal of Physiotherapy, and Norwegian Journal of Physiotherapy. Importantly, the first 4 journals are those that have initiated the transition to accept only trials that are prospectively registered.

Previous studies have examined cohorts of clinical trials for inconsistencies between the primary outcomes in the registry entry and the published report.17,18,21,24,34 Although there are minor differences in the definition of “inconsistency” among these studies and our study, most report the proportion of trials with at least one inconsistency. These proportions range from 30% to 67%.4,17,18,21–24 Our estimate of 55% is consistent with these. Only one of these studies considered inconsistency among nonprimary outcomes and identified a prevalence of 70%.17 We obtained a somewhat higher estimate of 85%, which again may reflect the inclusion of trials only from journals with high impact factors in the previous study.

Our results showed that researchers involved in multicenter and large studies are more likely to register their randomized trial. This also is the case for funded randomized trials, which suggests that grant bodies might be aware of this policy. However, trial registration in physical therapy varies widely across continents, and an extra effort to implement this policy worldwide is needed. Therefore, several groups could contribute to making clinical trial registration more effective. Researchers could undertake to register all of their trials prospectively, comprehensively, and unambiguously. Ethics review committees could make it a requirement of approval that clinical trials are prospectively registered. Clinical trial registers could ensure that all registry entries are comprehensive and unambiguous. The executive of the International Society of Physiotherapy Journal Editors also could encourage the adoption of mandatory clinical trial registration among member journals. Journal editorial boards could request registration for clinical trials, implementing a policy of mandatory prospective registration as soon as the supply of prospectively registered clinical trials allows.35–37 Journal editors and peer reviewers could compare submitted manuscripts with registry entries and insist that any discrepancies be resolved by following the registry entry. Finally, readers should give more credence to trials that are prospectively registered and consistent between the registered and published versions.

In this study, we have attempted to investigate whether methods of statistical analysis for the primary outcome were stated in the registry entry. Although statistical analysis is not included in the minimal registration data set,38 in 8 registry entries authors did prespecify analysis plans. There does not seem to be a consensus about the level of transparency in prespecifying analysis plans, and further discussion may be needed in this area. Selective outcome reporting has previously been shown to favor statistically significant results.18,21,39 Our data support this conclusion.

In this study, we assessed the completeness of registration and the extent of selective reporting of outcomes in a representative sample of published reports of randomized trials in physical therapy. However, the study did not generate any information regarding selective underreporting of trials or whether there is a trend toward favoring publication of trials with positive findings. Hence, readers should bear in mind that some trials about physical therapy intervention that have or have not been registered might never be published. Evidence of bias related to nonpublication has been reported in other fields.1,2 Further research specifically in the field of physical therapy is warranted.

In summary, the data generated by this study suggest that the potential benefits of trial registration are not currently being realized among published randomized trials of physical therapy interventions. Various groups involved in the funding, conduct, review, and publication of trials could promote effective clinical trial registration. In the interim, although registers did not prevent selective reporting of outcomes in many cases, they do allow readers to check for selective reporting. Future surveys of the prevalence of clinical trial registration should avoid restricting the sample of trials by the impact factor of the publishing journal because this factor may affect the estimate obtained.

Footnotes

  • Dr Elkins, Dr Moseley, Dr Sherrington, Dr Herbert, Dr Maher, Dr P.H. Ferreira, and Dr M.L. Ferreira provided concept/idea/research design. Mr Pinto, Dr Elkins, Dr Moseley, Dr Sherrington, Dr Herbert, and Dr Maher provided writing. Mr Pinto, Dr Elkins, and Dr Moseley provided data collection. Mr Pinto, Dr Elkins, Dr Sherrington, Dr Maher, and Dr P.H. Ferreira provided data analysis. Dr Moseley, Dr Maher, Dr P.H. Ferreira, and Dr M.L. Ferreira provided consultation (including review of manuscript before submission).

  • This project was not directly supported by any external grants or funds. Mr Pinto is supported by the Capes Foundation, Ministry of Education of Brazil. Dr Maher is supported by a research fellowship funded by the Australian Research Council. Dr Sherrington and Dr Herbert are supported by research fellowships funded by the National Health and Medical Research Council, Australia. The Capes Foundation, the Australian Research Council, and the National Health and Medical Research Council had no role in the design and conduct of the study; collection, management, analysis, or interpretation of data; or preparation, review, or approval of the manuscript. Dr Moseley is a developer of the Physiotherapy Evidence Database (PEDro) and is currently employed to manage PEDro.

  • Received May 21, 2012.
  • Accepted October 31, 2012.

References

    1. Dickersin K,
    2. Chan S,
    3. Chalmers TC,
    4. et al
    . Publication bias and clinical trials. Control Clin Trials. 1987;8:343–353.
    1. Hopewell S,
    2. Loudon K,
    3. Clarke MJ,
    4. et al
    . Publication bias in clinical trials due to statistical significance or direction of trial results. Cochrane Database Syst Rev. 2009;(1):MR000006.
    1. Chalmers I
    . Misconduct in medical research. BMJ. 1989;298:256.
    1. Hahn S,
    2. Williamson PR,
    3. Hutton JL
    . Investigation of within-study selective reporting in clinical research: follow-up of applications submitted to a local research ethics committee. J Eval Clin Pract. 2002;8:353–359.
    1. Hahn S,
    2. Williamson PR,
    3. Hutton JL,
    4. et al
    . Assessing the potential for bias in meta-analysis due to selective reporting of subgroup analyses within studies. Stat Med. 2000;19:3325–3336.
    1. Chalmers I
    . Underreporting research is scientific misconduct. JAMA. 1990;263:1405–1408.
    1. Dickersin K
    . The existence of publication bias and risk factors for its occurrence. JAMA. 1990;263:1385–1389.
  8. World Health Organization. International clinical trials registry platform. Available at: http://www.who.int/ictrp/trial_reg/en/index.html. Accessed November 16, 2009.
  9. World Health Organization. WHO trial registration data set (version 1.2.1). Available at: http://www.who.int/ictrp/network/trds/en/index.html. Accessed July 9, 2012.
    1. Moher D
    . Clinical-trial registration: a call for its implementation in Canada. CMAJ. 1993;149:1657–1658.
  11. International Committee of Medical Journal Editors. Clinical trial registration: a statement from the International Committee of Medical Journal Editors [editorial]. Available at: http://www.icmje.org/clin_trial.pdf. Accessed November 16, 2009.
    1. Askie L,
    2. Ghersi D,
    3. Simes J
    . Prospective registration of clinical trials. Aust J Physiother. 2006;52:237–239.
    1. D'Alonzo GE Jr
    . Clinical trial registration, a needed addition to the research process. J Am Osteopath Assoc. 2004;104:409–410.
  14. International Journal of Paediatric Dentistry. Author guidelines. Available at: http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1365-263X/homepage/ForAuthors.html
  15. Journal of Dentistry. Author guidelines. Available at: http://www.elsevier.com/wps/find/journaldescription.cws_home/30441/authorinstructions. Accessed November 16, 2009.
    1. Eysenbach G
    . Tackling publication bias and selective reporting in health informatics research: register your eHealth trials in the International eHealth Studies Registry. J Med Internet Res. 2004;6:e35.
    1. Ewart R,
    2. Lausen H,
    3. Millian N
    . Undisclosed changes in outcomes in randomized controlled trials: an observational study. Ann Fam Med. 2009;7:542–546.
    1. Mathieu S,
    2. Boutron I,
    3. Moher D,
    4. et al
    . Comparison of registered and published primary outcomes in randomized controlled trials [erratum in: JAMA. 2009;302:1532]. JAMA. 2009;302:977–984.
    1. Macedo LG,
    2. Elkins MR,
    3. Maher CG,
    4. et al
    . There was evidence of convergent and construct validity of Physiotherapy Evidence Database quality scale for physiotherapy trials. J Clin Epidemiol. 2010;63:920–925.
    1. Dwan K,
    2. Altman DG,
    3. Cresswell L,
    4. et al
    . Comparison of protocols and registry entries to published reports for randomised controlled trials. Cochrane Database Syst Rev. 2011;(1):MR000031.
    1. Chan AW,
    2. Hrobjartsson A,
    3. Haahr MT,
    4. et al
    . Empirical evidence for selective reporting of outcomes in randomized trials: comparison of protocols to published articles. JAMA. 2004;291:2457–2465.
    1. Vedula SS,
    2. Bero L,
    3. Scherer RW,
    4. Dickersin K
    . Outcome reporting in industry-sponsored trials of gabapentin for off-label use. N Engl J Med. 2009;361:1963–1971.
    1. Al-Marzouki S,
    2. Roberts I,
    3. Evans S,
    4. Marshall T
    . Selective reporting in clinical trials: analysis of trial protocols accepted by The Lancet. Lancet. 2008;372:201.
    1. Chan AW,
    2. Krleza-Jeric K,
    3. Schmid I,
    4. Altman DG
    . Outcome reporting bias in randomized trials funded by the Canadian Institutes of Health Research. CMAJ. 2004;171:735–740.
    1. Moseley AM,
    2. Elkins MR,
    3. Herbert RD,
    4. et al
    . Cochrane reviews used more rigorous methods than non-Cochrane reviews: survey of systematic reviews in physiotherapy. J Clin Epidemiol. 2009;62:1021–1030.
    1. Moseley AM,
    2. Sherrington C,
    3. Elkins MR,
    4. et al
    . Indexing of randomised controlled trials of physiotherapy interventions: a comparison of AMED, CENTRAL, CINAHL, EMBASE, Hooked on Evidence, PEDro, PsycINFO and PubMed. Physiotherapy. 2009;95:151–156.
    1. Michaleff ZA,
    2. Costa LO,
    3. Moseley AM,
    4. et al
    . CENTRAL, PEDro, PubMed, and EMBASE are the most comprehensive databases indexing randomized controlled trials of physical therapy interventions. Phys Ther. 2011;91:190–197.
    1. de Morton NA
    . The PEDro scale is a valid measure of the methodological quality of clinical trials: a demographic study. Aust J Physiother. 2009;55:129–133.
    1. Maher CG,
    2. Sherrington C,
    3. Herbert RD,
    4. et al
    . Reliability of the PEDro scale for rating quality of randomized controlled trials. Phys Ther. 2003;83:713–721.
    1. Juni P,
    2. Witschi A,
    3. Bloch R,
    4. Egger M
    . The hazards of scoring the quality of clinical trials for meta-analysis. JAMA. 1999;282:1054–1060.
  31. Thomson Reuters. Journal citation reports, science edition 2010. Available at: http://wokinfo.com/products_tools/analytical/jcr/. Accessed September 21, 2011.
  32. International Society of Physiotherapy Journal Editors. Available at: http://www.wcpt.org/ispje/membership. Accessed July 9, 2012.
    1. Zarin DA,
    2. Tse T,
    3. Williams RJ,
    4. et al
    . The ClinicalTrials.gov results database: update and key issues. N Engl J Med. 2011;364:852–860.
    1. Ross JS,
    2. Mulvey GK,
    3. Hines EM,
    4. et al
    . Trial publication after registration in ClinicalTrials.Gov: a cross-sectional analysis. PLoS Med. 2009;6:e1000144.
    1. Craik RL,
    2. Maher C
    . Still “not satisfied”…yet [editorial]. Phys Ther. 2008;88:423–425.
    1. Elkins M
    . New initiatives in Journal of Physiotherapy. J Physiother. 2011;57:67–68.
    1. Harms M
    . Clinical trial registration. Physiotherapy. 2011;97:181.
    1. De Angelis CD,
    2. Drazen JM,
    3. Frizelle FA,
    4. et al
    . Is this clinical trial fully registered: a statement from the International Committee of Medical Journal Editors. N Engl J Med. 2005;352:2436–2438.
    1. Dwan K,
    2. Altman DG,
    3. Arnaiz JA,
    4. et al
    . Systematic review of the empirical evidence of study publication bias and outcome reporting bias. PLoS One. 2008;3:e3081.