| HOME | HELP | CONTACT US | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Sign In to gain access to subscriptions and/or personal tools.
|
|
|
|||||||
Sign In to gain access to subscriptions and/or personal tools. |
|||||||||
Letter to the Editor |
1 L’Aquila, Italy2 Rome, Italy
Dear Editor:
We are writing with reference to the article by Matsumoto et al, "A Comparison of Bone–Patellar Tendon–Bone and Bone–Hamstring Tendon–Bone Autografts for Anterior Cruciate Ligament Reconstruction" (February 2006, pages 213–219).
Randomized controlled trials (RCTs) are at the core of evidence-based research in that by random allocation, each patient has an equal chance of being given each treatment. Thus, RCTs are the most effective way of preventing bias in research.1 However, random allocation is not equal to haphazard allocation, and randomization procedures should follow formal methods described in the literature, for example, simple randomization, block randomization, stratified randomization, and using random-number tables or random-number generators.1 In addition, the randomization procedure should be concealed to researchers performing the study,3,4,9 and an independent observer would be required to assess physician-based objective measures.4 This is all the more important in that the selection of quality studies to be included in a systematic review or in a meta-analysis requires assessing, among other criteria, method of randomization, concealment, and independent observer.6 Furthermore, given the known difficulties in performing randomized trials in surgery,7 RCTs are often conducted on relatively small samples. Systematic reviews and meta-analyses attempt to overcome the limitations of studies based on small samples by pooling data from published RCTs, summarizing current evidences, if any, on a specific topic. The quality of systematic reviews and of meta-analyses is affected by the quality of RCTs selected for inclusion,5 and it is therefore vital that published RCTs explicitly report detailed information pertaining to statistical methodology in the "Materials and Methods" section of the article. Often, authors of systematic reviews and meta-analyses have to contact the authors of RCTs to determine this information when it is not published in the original article.
This considered, Matsumoto et al reported that patients enrolled in their study were randomized to undergoing either bone–patellar tendon–bone or bone–hamstring tendon–bone autografts according to their birth dates (page 214 in their article). We would be grateful if the authors could further specify their method of randomization (eg, was it based on odd or even number ending of birth date). However, using birth dates as a randomization procedure may be methodologically incorrect for 2 reasons. First, if the allocation criterion is not concealed to researchers, and we were unable to find out any mention of concealment in the article, bias may be introduced, especially if objective measurements are not conducted by an independent observer. Unfortunately, the authors do not declare whether objective measurements were assessed by an independent observer. Second, allocation by birth date is not a formally recognized method of randomization1,2 and, as such, may introduce bias, thus flawing the study. Particularly, Matsumoto et al recognized that 1 issue in their study may have been a difference in gender distribution and length of follow-up, although no significant difference between groups was observed (Table 1 and page 218 in their article). In addition to the consideration that a type II or beta error should be calculated when tests are not significant, as the test power may be inadequate in small-sample studies, the imbalance in baseline variables between 2 groups may bias the outcome. It has been recognized that lack of balance in baseline variables known to affect outcome can bias statistical tests, a property defined as chance bias.8 As a result, observed differences in outcome between groups in a particular trial may derive from the characteristics of the patients, not treatment.8 Simple randomization, which is a different procedure from allocating patients according to birth date, removes bias from the allocation procedure, yet it does not ensure that the individuals in each group have, for example, similar gender or age distribution. Chance imbalance may consequently occur, especially in small-sample studies.8 If gender imbalance was deemed to be an issue by the authors in the phase of designing the study, the use of stratified randomization might have achieved a better balance of characteristics while maintaining random allocation. On the other hand, block randomization might have been used if the authors wished to allocate an equal number of patients in each group.
Although we praise the authors for conducting their RCT, we regret detecting some methodological inaccuracies that may limit the quality of their RCT. We believe that researchers conducting RCTs should consider their work in the broader perspective of evidence-based research and comply to recognized standards of methodology. Thus, the efforts of other researchers performing systematic reviews or meta-analyses at a later stage may be facilitated when published RCTs are screened for their quality.
REFERENCES
Kobe, Japan
We are responding to the letter to the editor (from Dr Stefano Lupparelli, Dr Vittorio Calvisi, and Dr Emilio Romanini) regarding the article by Matsumoto et al, "A Comparison of Bone–Patellar Tendon–Bone and Bone–Hamstring Tendon–Bone Autografts for Anterior Cruciate Ligament Reconstruction" (February 2006, pages 213–219).
In our study, objective measurements were assessed by an independent observer to prevent a bias in the evaluation.
We "randomized" the patients by the month of birth. This, however, does not seem to be a true randomization. As they pointed out, randomizing patients by the medical record number, day of admission, and day, month, and year of birth does not ensure random occurrences because there is no concealment of randomization. Without concealment of randomization, the investigator can systematically influence the patient population that will be assigned to the different treatment groups by changing the indication for surgery, encouraging patients to participate in the study, interpreting the inclusion/exclusion criteria, and so forth. Random allocation is not equal to haphazard allocation, and randomization procedures should follow formal methods, for example, simple randomization, block randomization, stratified randomization, and using random-number tables or random-number generators. Thus, patient assignment will be based not only on a random process but also on the investigator’s choice. Although we mentioned that no significant difference in gender distribution and length of follow-up between groups was observed, stratified randomization should have been used for random allocation considering gender, age, and so forth in our study. Therefore, our study was not a high-quality randomized controlled trial but was an evidence-based medicine level 2 study. However, surgeries for the included patients in this study were performed between 1994 and 1996, when importance of randomization method was not well recognized (8 of the 9 references cited in the letter were published after 1999). We would also appreciate if they could take this situation into account.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | CONTACT US | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
