We searched for all low back pain randomized controlled trials, published between 2010 and 2015, and indexed on PEDro database on February 1st 2016. Then we randomly selected a sample of approximately 40% of all eligible trials, using the random number function in Excel. The eligibility criteria were: full-published articles; written in English, Spanish or Portuguese; and published between 2010 and 2015. The languages were restricted to English, Spanish and Portuguese, as they are languages spoken by the authors of this manuscript and are within the most published languages on PEDro.14,23 Trials that were still in the recruitment stages, protocols, duplicates and those involving any topic other than low back pain were excluded. We used the advanced search strategy on PEDro, as follows: “clinical trial” for method; “lumbar spine, sacroiliac joint or pelvis” for body part; “pain” for problem; and 2010–2015 for year of publication.

One independent author extracted data in an Excel spreadsheet previously formatted to describe the trials’ characteristics. Data extracted from the trials were: (1) continent where trial was conducted; (2) language of publication, categorized into English or non-English; (3) methodological quality, which was measured by total PEDro score and downloaded from PEDro database (0–10 points with higher values indicating better methodological quality); (4) number of years since publication, calculated by subtracting the year of publication from 2015; (5) number of citations normalized by the number of years since publication (extracted from the Web of Science Clarivate Analytics); (6) total sample size (i.e., number of participants randomized); (7) primary outcomes, which were collected at most two outcomes. For articles that did not specify the primary outcome, but had two outcomes mentioned only, we considered both as primary. For articles that did not specify the primary outcome but had more than two outcomes, we considered “pain” and “disability” as primary.24 In the cases which the article did not have any of these two outcomes, we collected the first two mentioned by the authors. All primary outcomes were categorized into: pain, disability, function, quality of life or other; (8) interventions, which included any type of physical therapy intervention. Each intervention was categorized according to the intervention codes used by PEDro25 (i.e., Acupuncture; Behavior modification; Education; Electrotherapy, heat, cold; Fitness training; Health promotion; Hydrotherapy, balneotherapy; Neurodevelopmental therapy, neurofacilitation; Ortheses, taping, splinting; Respiratory therapy; Skill training; Strength training; Stretching, mobilization, manipulation, massage; No appropriate value in this field); (9) comparators, which were categorized into no-treatment control,26 placebo,27 addition of other treatment,28 waiting list,29 minimal intervention,28 usual care30 and/or other intervention.31 We also classified the trial in one of these groups if the authors of the trial classified the comparator group as such.

Data extracted from journals were: (1) 2015 journal Impact Factor (from the Web of Science, InCites Journal Citation Reports – Thomson Reuters’ website Clarivate Analytics), and categorized in three strata (Journals without Impact Factor; Journals with Impact Factor less than 2.0; Journals with Impact Factor more than 2.0); (2) if journal was open access (yes or no), which was obtained from the Directory of Open Access Journals, PubMed Central or the journal's web-site; and (3) if journal endorses the CONSORT recommendations32 (yes or no), which was extracted from the “Instructions to authors” for each of the journals or from the CONSORT's website.13

The main outcome measures were: journal Impact Factor; if the paper was published as open access; CONSORT recommendations endorsement by the journal; and methodological quality measured by total PEDro score.

All trials characteristics were analyzed descriptively. Normality tests were conducted and all data were normally distributed, except for the total sample size, which was reported as median and interquartile range. The correlation between trials methodological quality and journals Impact Factor was calculated using Pearson correlation. We used SPSS software (Version 20.0) for all analyses.

Low back pain randomized controlled trials have been published in a variety of healthcare journals, with Impact Factors ranging from 0.4 to 19.7, and even in some journals that were not even indexed in the Web of Science. Most of these trials had moderate methodological quality (with a PEDro mean score of 5.8 points out of 10.0) and were published in open access journals that endorsed CONSORT recommendations. Similar to a previous study in musculoskeletal physical therapy,14 our results indicate a positive, but weak correlation between methodological quality and Impact Factor. This means that high quality low back pain trials are not necessarily published in journals with high Impact Factor.

The sample size showed a median of 73.5 participants (IQR 109, minimum 10–maximum 4.325), which shows that some studies analyzed a small group of participants, which probably did not reach the statistical power of detecting a clinical important difference among the intervention and control groups.33

The primary outcomes most evaluated by the trials were pain and disability, which agrees with the core outcome sets found in panel consensus for low back pain, together with quality of life.34 The instruments mostly recommended for physical functioning are the Oswestry Disability Index and 24-item Roland-Morris Disability Questionnaire, and the Numerical Pain Rating Scale.34 This shows that most of back pain researchers are aligned with these recommendations.

The most frequent interventions analyzed in our sample were exercises (strength training and stretching), manual therapy and education, which are the main recommended interventions for low back pain according to the National Institute for Health and Care Excellence (NICE) guidelines,35 and also to a recent low back pain series published by a panel of experts in the field.2–4 Maybe authors should investigate further the effectiveness of interventions that were poorly or not investigated in these trials for low back pain and which do not present much evidence according to previously published guidelines,2–4,35 such as hydrotherapy and neurodevelopment therapy, for example.

A particular strength of our study is that we presented 30 possible options of high Impact Factor journals that published low back pain trials, from which six journals are highly recommended for reading and referred to when taking a clinical decision. Also, we are confident that we have a representative sample of low back pain trials published on PEDro in 2010–2015, once we selected 40% of the total ‘population’ of eligible trials. However, our main limitation is that some journals might have been missed and we could not generalize our results for all low back pain trials, including trials involving surgery or drug trials.

The analyzed trials comparing interventions for low back pain presented moderate methodological quality. Readers should be very cautious on taking clinical decisions based on poor methodological quality trials, which may increase the existing gap between evidence and practice.2–4 The research efforts and global initiatives to treat this public health problem need also to focus on improving reporting and methodological quality for future studies. Although 55.5% of journals already endorse CONSORT recommendations, we suggest that the EQUATOR recommendations (Enhancing the QUAlity and Transparency Of health Research; https://www.equator-network.org/) become mandatory requirements for publication, as well as following the items evaluated in methodological guidelines (e.g. Cochrane Risk of Bias tool or the PEDro scale36). Journal editors and peer reviewers should strictly follow those requirements in order to improve reporting and methodological quality of randomized controlled trials. Therefore, actions of better reporting and conducting the studies following adequate methodological guidelines will better address authors and journals editors/reviewers to write and publish more trustworthy studies.