From:  A survey of the composition, safety, efficacy and regulation of non-autologous biologic treatments in human musculoskeletal pathologies

 Description of clinical study or RCT attributes including a bias estimate and statistical methods

ReferencesStudy typeClinicaltrials.gov identifierEnrollmentRisk of biasStatistical methods for analysis
Gaudilliere et al. [6]RCTNCT0398141938Randomized with control arm; double blinded; lowDue to the large number of proteomic targets available, the authors used an advanced regression model to identify analytes of interest for comparison between treatment cohorts. Paired comparisons were analyzed with a non-parametric method.
Beall et al. [8]RCTNCT03709901218Randomized with two control arms; patient blinding; moderateStandard statistical assessment of results was performed.
Hunter et al. [9]—a post hoc analysis of the study reported in [8]RCTNCT03709901218Randomized with two control arms; patient blinding; moderatePost hoc analysis of the primary dataset reported in [8] was performed with non-parametric methods appropriate for a three-group comparison; significance of the three-group non-parametric analysis was confirmed by a subsequent ranked-based non-parametric analysis.
Psathas et al. [13]RetrospectiveNA32Standardized review conducted; not all patients provided consent; moderateA standard statistical assessment for significance between categorical variables was performed.
Tettelbach et al. [14]RCTNCT01693133110Randomized with control arm; outcomes validated by blinded adjudicator panel; lowDifferences in continuous variables were assessed by standard methods, including a non-parametric analysis of multiple groups. Categorical variables were assessed by standard methods, including regression modeling with fixed effects.
Ahuja et al. [15]RetrospectiveNA30Chart review; highNo comparative statistical analysis was performed.
Gomoll et al. [17]RCTNCT02318511200Randomized with two control arms; single blinded; moderateOutcomes were assessed for change from baseline with standard statistical assessment of significance between treatment groups.
Zelen et al. [18]RCTNCT0165982745Randomized with control arm and two treatment arms; patient blinded; moderateAppropriate non-parametric methods were used for comparing two or more groups of continuous data. Parametric methods were used for comparing binary data.
Hanselman et al. [19]RCTNA24Randomized with control arm; double blinded; lowStandard methods were used for assessing the significant differences between control and treatment arm outcomes, including a separate assessment of the number of injections on treatment outcomes.
Alden et al. [20]RetrospectiveNA82 (100 knees)Chart review; highSubscores of KOOS were averaged for each timepoint, and an arbitrary cutoff of a change of “10 pts” in any of the subscores compared to baseline was considered to represent a “clinically meaningful improvement”.
Natali et al. [21]ProspectiveNA25Non-randomized; highNon-parametric analysis was performed for the outcome metrics comparing pre- and post-treatment results, with a specific assessment of age.
Meadows et al. [22]ProspectiveNA10Non-randomized; highAppropriate non-parametric analysis was performed to assess the significance of all metrics reported post-treatment compared to baseline.
Noriega et al. [24]RCTNA24 (23 at 3.5 year follow-up)Randomized with control arm; blinding through all phases; lowAppropriate non-parametric and parametric assessments were performed to assess the significant difference in the results.
Amirdelfan et al. [25]RCTNCT01290367100Randomized with two control and two treatment arms; blinding of participants and radiological evaluation; moderateA complex analysis of outcomes was performed, including the use of appropriate statistical methods for dealing with missing data. The impact of multiple parameters was assessed by repeated measures mixed modelling.
Gornet et al. [26]RCTNCT0334770860Randomized with two control and two treatment arms; double blinded; lowA complex analysis of VAS based on repeated measures mixed-effects linear modelling was performed, adjusting for missing data. Validation of the analysis was performed with a separate statistical assessment with its own null hypothesis. Group differences were assessed by standard methods, including an analysis of co-variance.
Abdullah et al. [32]RCTNA40Randomized with control and treatment arms; double blinded; lowAnalysis with an appropriate parametric or non-parametric standard method was employed after assessing the distribution of the data as either normal or non-normal.
Mazzotta et al. [33]RetrospectiveNA96Consecutive case review with blinded to treatment matched pair selection; moderateA stratified approach was used to confirm normality of results, as well as the consistency in variances for datasets. Differences in outcomes over time were assessed in a linear model, along with an ANOVA assessment of between-groups differences when the data was normally distributed and variances were constant. Where datasets didn’t meet these requirements, non-parametric methods were employed. Where appropriate, non-parametric correlative measures were assessed. Standard statistical methods were used for parametric analysis.
Zhu et al. [35]RCTChiCTR2100048624 (Chinese Clinical Trial Registration)80Randomized with treatment and control arms; blinded to cohort; lowMultiple assessments of the normality of the distribution of various datasets were used to perform either parametric or non-parametric analysis. Multiple outcome measures were evaluated in a mixed linear model, with assessments of interactions. Regression models and ANOVA analysis were performed to characterize the influence of secondary outcomes with interactions. Of note, blinding was assessed with the James blinding index.
Lightner et al. [40]RCTNCT04493242102Randomized with two treatment arms and a control arm; double-anonymized; lowPre-planned standard methods of analysis of the primary outcome of 60-day mortality rate were used. Pre-defined subgroup analyses of mortality data also were performed with standard methods.
Gibson et al. [42]RCTNCT0300510671Randomized with treatment and control (autograft) arms; moderateNon-parametric methods of analysis were used to assess significant differences in outcome measures of healing between the test article and autograft. Standard methods of analysis were used for parametric assessment of participant-reported outcomes, or for non-participant assessments of healing.

KOOS: Knee Injury and Osteoarthritis Outcome Score; RCT: randomized controlled trial; VAS: visual analog scale