Concise Review
Randomization Procedures in Orthopaedic Trials
Pietro Randelli, M.D., Paolo Arrigoni, M.D., James H. Lubowitz, M.D.,
Paolo Cabitza, M.D., and Matteo Denti, M.D.
Abstract: Historically, the surgical literature has lacked in reports of randomized clinical trials. This
deficit is now being addressed, but the best methods are not always followed. One opportunity for
improvement is in the area of randomization. Randomization is of central importance in clinical trials
because it reduces bias and represents a basis for ensuring the validity of data analysis using statistical
testing. Randomization requires a table of random numbers. Simple randomization is adequate for
large trials. Block randomization is a method of balancing equal numbers of patients in each
treatment group. Stratification allows balanced distribution of one or more confounding prognostic
variables among treatment groups to ensure that groups have similar prognoses (minimizing selection
bias). Block randomization and stratification improve validity in trials with fewer patients. Com-
mercially available computer software facilitates randomization. Key Words: Allocate—Block—
Randomization—Selection bias—Simple—Stratify.
A
prospective, randomized controlled trial (RCT) is
an essential tool for comparing the efficacy of
surgical treatments and provides the highest level of
evidence when attempting to resolve a primary re-
search question. With proper randomization, treatment
is not known to the researcher at the time a patient is
registered for a trial; thus, selection bias (bias result-
ing from patients in different treatment groups having
differing prognoses) is minimized, because random-
ization results in a balance of unknown (as well as
known) prognostic variables when patients are as-
signed to various treatment arms.
Review of the surgical
1-3
and orthopaedic surgical
4
literature reveals that few published studies (3% to
6%) represent RCTs, and a minority of trials employ
proper methods of randomization. A review of pub-
lished “RCTs” of upper extremity disorders reveals
that only 43% were truly randomized; of these, 23%
described an inappropriate randomization method, and
in 34%, the randomization method was not described.
5
The purpose of this review is to concisely consider
and suggest an ideal method of randomization for
orthopaedic surgical trials.
RANDOM NUMBERS
Randomization is ideally based on a table of random
numbers, which is readily generated by computer. Such
a table consists of a group of digits that are completely
independent of one another, such that knowledge of
proceeding digits of a series does not allow prediction of
the next digit, because each digit has an equal chance
of appearing in each position. A simple way to obtain
such a list of random numbers is to use the “Randbe-
tween” worksheet function provided in Microsoft Ex-
cel (version 2003; Microsoft, Redmond, WA).
From the Dipartimento di Scienze Medico-Chirurgiche (P.R.,
P.A., P.C.), Università degli Studi di Milano, IRCCS Policlinico
San Donato, Milan, Italy; Taos Orthopaedic Institute (J.H.L.),
Taos, New Mexico, U.S.A.; and the Centro di Traumatologia dello
Sport (M.D.), Istituto Ortopedico Galeazzi, Milan, Italy.
The authors report no conflict of interest.
Address correspondence and reprint requests to Pietro Randelli,
M.D., IRCCS Policlinico San Donato, Via Morandi 30 20097, San
Donato Milanese, Milan, Italy. E-mail: pietro.randelli@tin.it
© 2008 by the Arthroscopy Association of North America
0749-8063/08/2407-0822$34.00/0
doi:10.1016/j.arthro.2008.01.011
834 Arthroscopy: The Journal of Arthroscopic and Related Surgery, Vol 24, No 7 (July), 2008: pp 834-838