Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
Editorial
. 2022 Mar;30(3):565-570.
doi: 10.1002/oby.23373.

A practical decision tree to support editorial adjudication of submitted parallel cluster randomized controlled trials

Yasaman Jamshidi-Naeini  1 , Andrew W Brown  2 , Tapan Mehta  3 , Deborah H Glueck  4 , Lilian Golzarri-Arroyo  1 , Keith E Muller  5 , Carmen D Tekwe  1 , David B Allison  1
Affiliations
Editorial

A practical decision tree to support editorial adjudication of submitted parallel cluster randomized controlled trials

Yasaman Jamshidi-Naeini et al. Obesity (Silver Spring). 2022 Mar.
No abstract available

PubMed Disclaimer

Conflict of interest statement

CONFLICT OF INTEREST

In the last 36 months prior to the initial submission, DBA has received personal payments or promises for same from the following: Alkermes, Inc.; American Association for the Advancement of Science; American Society for Nutrition; American Statistical Association; Amin Talati Wasserman for KSF Acquisition Corp (Glanbia); Big Sky Health, Inc.; Biofortis Innovation Services (Merieux NutriSciences); California Walnut Commission; Clark Hill PLC; Columbia University; eLife Sciences; Frontiers Publishing; Henry Stewart Talks; Indiana University; Johns Hopkins University; Kaleido Biosciences; Law Offices of Ronald Marron; Medical College of Wisconsin; Medpace/Gelesis; NIH; National Academies of Science; Nestec/Nestle; Novo Nordisk Fonden; Sage Publishing; Sports Research Corp.; The Elements Agency, LLC; The Obesity Society; Taylor and Francis; Tomasik, Kotin & Kasserman LLC; University of Alabama at Birmingham; the University of Miami; and WW (formerly Weight Watchers International, LLC). Donations to a foundation have been made on his behalf by the Northarvest Bean Growers Association. DBA was previously an unpaid member of the International Life Sciences Institute North America Board of Trustees. In the last 36 months prior to the initial submission, AWB has received travel expenses from University of Louisville; speaking fees from Purdue University and University of Arkansas; consulting fees from LA NORC and Pennington Biomedical Research Center; and award honorarium from the American Society for Nutrition Foundation. His wife is employed by Reckitt Benckiser. The institution of DBA and AWB, Indiana University, and the Indiana University Foundation have received funds, contracts, or donations to support their research or educational activities from the following: NIH; USDA; Soleno Therapeutics; National Cattlemen’s Beef Association; Eli Lilly and Co.; Reckitt Benckiser Group PLC; Alliance for Potato Research and Education; American Federation for Aging Research; Dairy Management Inc.; American Egg Board; California Walnut Commission; Almond Board; Peanut Institute; Mondelez; Morris Animal Foundation; Oxford University Press; Hass Avocado Board; Whistle Labs, Inc.; Dynamic Air Quality Solutions (AQS); Herbalife International; Mars, Inc.; Arnold Ventures; the Gordon and Betty Moore Foundation; the Alfred P. Sloan Foundation; Indiana CTSI; Center for Open Science, and numerous other for-profit and nonprofit organizations to support the work of the School of Public Health and the university more broadly. In the past 3 years prior to the initial submission, TM has received personal payments from University of Wisconsin Milwaukee, University of Alabama, American Physical Therapy Association, The Obesity Society, and PLOS One. The other authors declared no conflict of interest.

Figures

FIGURE 1
FIGURE 1. Editorial decision tree for human studies that are or purport to be cluster randomized trials (cRCTs).
The focus of the decision tree is on trials that are cluster randomized, are parallel by design, and have two or more clusters per condition. Recommendations herein specifically apply to one level of clustering. *Asterisks indicate the CONSORT extension for cluster trials checklist items where relevant information should be obtained. §The first step through the question “Are there at least two clusters randomized to each condition?” could be operationalized as part of the editorial screening process, while the subsequent decision flow check will require an expert peer review. (1) The following should be identified from such a statement: independent sampling unit, unit of randomization, and randomization process. See Glossary for definitions. Note that any nonrandom allocation of independent sampling units (e.g., mix of random and nonrandom assignment) to study conditions disqualifies the study from being an RCT or a cRCT (14). (2) There may be clustering involved that should be noted and considered in the analysis (15). For an example of an individually randomized trial that has clustering involved, see The ACTonHEART study (16). (3) Sets of grouped study participants that are randomly assigned are groupings that exist prior to randomization, such as families, couples, classrooms, schools, clinics, patients cared for by the same physician, work sites, counties, or communities (9). It is also possible that randomized groups are created by directing individuals to the same groups under conditions that induce dependency among participants (e.g., time-tabling constraints, participant choice) (17). (4) a) Regarding the minimal number of clusters to permit estimation and testing of the causal effect of treatment assignment, the absolute minimum number is theoretically three. If all clusters are equal in size, then two clusters in one condition and one cluster in another condition could be used, though in truth such a design would have such low power and so little robustness, that it would arguably be invalid for practical purposes. In contrast, if the clusters are unequal in size, then a test cannot be guaranteed to exist because iterative solution of the likelihood equations to find estimates will frequently fail. This has recently been shown by simulation results for two clusters in each of two groups with 100 members (18); b) Note that this should not be reclassified as an individually randomized trial. It might be better to reclassify the study as a quasi-experiment (19). In this case, authors should remove any language describing this as a cRCT or RCT and rephrase any causal inference accordingly. (5) For recommendations for choosing an analysis method for unbalanced cluster sample designs, see (20). (6) There are three considerations here: a) One example of methods to account for clustering (i.e., intraclass correlation coefficient) in statistical analyses is mixed-effects regression, in which clusters are modeled as random effects. Note that using generalized estimating equations that model the correlation structure directly for analysis of cRCTs may have inflated type I error rate (21). Nesting is accounted for by adjusting degrees of freedom for testing the intervention effect (22); b) Note that estimating random effects for individual study participants accounts for repeated measurements nested within study participants and should not be confused with accounting for the clustering due to cluster randomization; c) If more than one level of clustering and nesting is expected in the design (e.g., students nested in classrooms nested in schools), consult a statistician about necessity of accounting for all levels of clustering and nesting other than those which are attributable to study participants within randomized clusters. (7) Arguments that the study is a pilot study, that there are too few clusters to account for clustering and nesting, or that the intraclass correlation coefficient is low do not provide sufficient justification for ignoring clustering in analyses (22). (8) Note that analysis of cRCTs involves many complexities. Mistakes in calculating and using degrees of freedom are common. The best practice for the verification of analyses is asking the authors to supply their analytic code (e.g., R or SAS code) and having the code and manuscript reviewed by a professional statistician with expertise in cRCTs. (9) Such a statement should address the number of experimental conditions, the number of clusters per condition, the cluster size, whether equal cluster sizes were assumed, the choice and specification of appropriate intraclass correlation coefficient, and the methodology used to account for repeated measures (if applicable). Note that other important design and analytical concerns still apply (e.g., order of cluster and participant recruitment and randomization; whether missingness is at random or not; the use of intention-to-treat vs. completers-only analysis approach), but are beyond the scope of this tool. Important and detailed guidelines are available elsewhere (11,12)
See this image and copyright information in PMC

Similar articles

See all similar articles

Cited by

See all "Cited by" articles

References

    1. Murray DM, Varnell SP, Blitstein JL. Design and analysis of group-randomized trials: a review of recent methodological developments. Am J Public Health. 2004;94:423–432. - PMC - PubMed
    1. Sedgwick P. Clinical trials: units of randomisation. BMJ. 2014;348:g3297.doi:10.1136/bmj.g3297 - DOI - PubMed
    1. Golzarri-Arroyo L, Oakes JM, Brown AW, Allison DB. Incorrect analyses of cluster-randomized trials that do not take clustering and nesting into account likely lead to p-values that are too small. Child Obes. 2020;16:65–66. - PMC - PubMed
    1. Golzarri-Arroyo L, Vorland CJ, Thabane L, et al. Incorrect design and analysis render conclusion unsubstantiated: comment on “A digital movement in the world of inactive children: favourable outcomes of playing active video games in a pilot randomized trial”. Eur J Pediatr. 2020;179:1487–1488. - PMC - PubMed
    1. Tekwe CD, Allison DB. Randomization by cluster, but analysis by individual without accommodating clustering in the analysis is incorrect: comment. Ann Behav Med. 2020;54:139. doi:10.1093/abm/kaz065 - DOI - PubMed

Publication types

LinkOut - more resources