Secondary Logo

Journal Logo


Accounting for Center Effects in Multicenter Trials

Tangri, Navdeep; Kitsios, Georgios D.; Su, Shi Hann; Kent, David M.

Author Information
doi: 10.1097/EDE.0b013e3181f56fc0
  • Free

To the Editors:

Individual patients in multicenter trials may not represent truly independent observations. Center characteristics and practice patterns, particularly in cases with substantial between-center variability in outcome rates, can lead to potentially misleading conclusions, if ignored. More specifically, failure to consider the center can lead to incorrect standard errors and P values (due to clustering), biased estimates (from uncontrolled confounding), and unrecognized heterogeneity across centers (from effect modification).1–3

Although statistical methods for dealing with center-level clustering in the analysis of randomized controlled trials (RCTs) have been well-described and advocated,1–5 there is some evidence that the application of these methods is limited.6,7 We believe that accounting for center effects is of crucial importance in RCTs of medicinal products, particularly in large multicenter studies.2,3 Because the extent of center-effect adjustment has not previously been described for medicinal products, we performed an empirical evaluation of adjustment for center-level clustering in reports of multicenter RCTs in 4 major medical journals.

A systematic search for RCTs published during the year 2007 in 4 prominent medical journals (British Medical Journal, Journal of the American Medical Association, Lancet, and New England Journal of Medicine) was conducted in PubMed. We evaluated the retrieved articles for the following inclusion criteria: adult human study population enrolled, use of randomized design, multicenter enrollment, and testing for efficacy/effectiveness of medicinal products.

The main characteristics of the 101 included RCTs are shown in the Table. The majority of the trials used a superiority study design; cardiovascular and oncology disorders were the most commonly studied conditions (32% and 25%, respectively); and binary and time-to-event outcomes were examined in almost equal proportions.

Characteristics of RCTs Included in the Analysis (n = 101)

The number of centers included in the RCTs ranged from 2–707 (median = 64, interquartile range = 22–117), representing 70–22,949 patients. Of the 101 studies, 36 (36%) performed random allocation stratified by center. Statistical analysis adjusting for the clustering of patients by center was present in 18% of the reports. Of these, only 1 used a random term for the center-effect, whereas the remaining used a fixed-effects model. Thus, a total of 82% did not adjust for center effects.

Previous investigators have studied RCTs of surgical interventions; they reported similar results for allocation stratification on center (38%) and adjusted statistical analysis (6%).6–8 Our literature sample is more contemporary and focused on 4 major medical journals publishing RCTs of medical interventions. Studies published in these journals are expected to be of high quality and have the potential to disproportionately influence clinical practice. Nonetheless, the similarities between our results and those of previous investigators highlight the lack of accounting for center effects in RCTs as a widespread problem.

Our analysis has some limitations. Our results may not be generalizable to the entire medical literature. (The proportion of RCTs accounting for center effects is likely to be even lower among the less prestigious journals.) Second, our conclusions regarding center effect accounting were based on the published statistical methods. It is possible that appropriate accounting was performed, but not reported due to space constraints.

In summary, using a contemporary sample of RCTs from 4 major medical journals, we find that center effects are not accounted for in the recruitment stage or in the statistical analysis of the majority of RCTs evaluating medical interventions. The recent extension of the CONSORT statement advocates center-effect reporting and adjustment for trials, involving nonpharmacologic interventions. Our analysis highlights the need for similar recommendations in trials of medical therapies.

Navdeep Tangri

Georgios D. Kitsios

Shi Hann Su

David M. Kent

Tufts Clinical and Translational Science Institute

Institute for Clinical Research and Health Policy Studies

Tufts Medical Center

Boston, MA

[email protected]


1. Andersen PK, Klein JP, Zhang MJ. Testing for centre effects in multi-centre survival studies: a Monte Carlo comparison of fixed and random effects tests. Stat Med. 1999;18:1489–1500.
2. Localio AR, Berlin JA, Ten Have TR, Kimmel SE. Adjustments for center in multicenter studies: an overview. Ann Intern Med. 2001;135:112–123.
3. Agresti A, Hartzel J. Strategies for comparing treatments on a binary response with multi-centre data. Stat Med. 2000;19:1115–1139.
4. Donald A, Donner A. Adjustments to the Mantel-Haenszel chi-square statistic and odds ratio variance estimator when the data are clustered. Stat Med. 1987;6:491–499.
5. Boutron I, Moher D, Altman DG, Schulz KF, Ravaud P. Extending the CONSORT statement to randomized trials of nonpharmacologic treatment: explanation and elaboration. Ann Intern Med. 2008;148:295–309.
6. Biau DJ, Halm JA, Ahmadieh H, et al. Provider and center effect in multicenter randomized controlled trials of surgical specialties: an analysis on patient-level data. Ann Surg. 2008;247:892–898.
7. Biau DJ, Porcher R, Boutron I. The account for provider and center effects in multicenter interventional and surgical randomized controlled trials is in need of improvement: a review. J Clin Epidemiol. 2008;61:435–439.
8. Gabler NB, Duan N, Liao D, Elmore JG, Ganiats TG, Kravitz RL. Dealing with heterogeneity of treatment effects: is the literature up to the challenge? Trials. 2009;10:43.
© 2010 Lippincott Williams & Wilkins, Inc.