The article below summarizes a roundtable discussion of a study published in this issue of the Journal in light of its methodology, relevance to practice, and implications for future research. Article discussed:
Candiani M, Izzo S, Bulfoni A, Riparini J, Ronzoni S, Marconi A. Laparoscopic versus vaginal hysterectomy for benign pathology. Am J Obstet Gynecol 2009;200:368.e1-368.e7.
Was the hypothesis clearly stated?
What type of study was this?
How was the sample size determined?
How were the outcomes measured?
Were the statistical methods adequate?
What information is in the tables?
What proportion of patients underwent adnexectomy?
Was the main conclusion supported?
How might this study have been improved?
Hysterectomy is the most common nonobstetric surgical procedure performed in women. Each of the 3 main approaches to removing the uterus (open abdominal, vaginal, and laparoscopic) has its own set of risks and benefits. In discussing hysterectomy routes with a patient, the surgeon considers her specific indications, characteristics such as uterine size and history of surgical procedures and her personal preferences. The goal, of course, is to work toward a decision that optimizes a procedure's advantages and minimizes its disadvantages.
See related article, page 368
How should a gynecologist counsel a patient on hysterectomy method if her clinical scenario does not clearly dictate the superiority of 1 approach over another? Generally, because of decreased operative time, shorter hospital stay, and reduced recovery time, vaginal hysterectomy has been considered the better approach when compared with laparotomy. However, when there is no evidence of uterine or adnexal malignancy and the uterus is mobile and of a limited volume, it has yet to be determined whether the vaginal approach is a wiser option than laparoscopic methods. In their randomized clinical trial, Candiani et al suggest that, in these circumstances, laparoscopic methods might be a better choice.
Assigned at Random
The randomized controlled trial (RCT) is the preferred study design for comparing 2 interventions. By randomly assigning participants to an intervention, researchers theoretically control for known and unknown confounders by producing groups with similar baseline characteristics. However, in studies with smaller sample sizes, simple randomization may not sufficiently balance the distribution of measured and unmeasured confounders between 2 treatment groups.
Using an RCT allowed Candiani et al to control for confounders, to define the participant population strictly, and to standardize the surgical procedure. The data indicated that vaginal hysterectomy was associated with an increased hospital stay, compared with laparoscopic hysterectomy. Addressing the same research question with a cohort study design would likely have introduced more bias than would be expected to occur with an RCT design. For example, with a cohort design, if the investigators differentially assigned patients to procedures based on their individual characteristics (assigning healthier patients to laparoscopic hysterectomy, for example), it could affect the length of stay.
Study participants included women with an indication for vaginal hysterectomy and benign disease. Recruiting for surgical trials can be challenging because patients often have a preference for a certain procedure that is based on information they have gathered; therefore, they might be hesitant when assigned to a surgical procedure at random. Yet, Candiani et al had only a 7% refusal rate, which led Journal Club members to want more information about their recruitment technique.
The inclusion and exclusion criteria of an RCT must be clearly defined. This investigation defined a very specific subset of women who were undergoing hysterectomies: women for whom neither vaginal nor laparoscopic hysterectomy was an obviously better choice. Equipoise, as this uncertainty is termed, unfortunately comes at the expense of generalizability. Because such a narrow subset of women who needed hysterectomy was included in this study, the results can be generalized only to women with the same set of characteristics: small uterus, no malignancy, no previous surgery for pelvic inflammatory disease or endometriosis, minimal pelvic organ prolapse, and no ovarian cysts > 4 cm.
Turn a Blinded Eye
Randomization controls for confounders at the outset of a study; blinding controls for confounding during the follow-up phase by preventing the intervention groups from being managed differently and by preventing knowledge of the intervention from influencing assessment of the outcome. The investigators did not use blinding at any stage of this study.
Journal Club members suggested that the lack of blinding was a major source of bias in this study. The surgical intervention could influence postoperative treatment of participants and the evaluation of outcome (appropriateness of discharge). Staff on the postoperative ward might be inclined to alter the care they provided depending on whether patients had a vaginal or laparoscopic hysterectomy. For example, they might offer more pain medication after 1 procedure or counsel patients differently on when to expect to be discharged home. Patients could have preconceived ideas pertaining to the expected length of stay and the pain associated with each procedure. Surgeons who determined appropriateness for patient discharge may have taken the actual procedure into account when deciding to discharge, despite well-defined discharge guidelines. These factors would all influence the main outcome, length of stay.
Although it is impossible to have a completely blinded RCT of surgical interventions (the surgeons must know what procedure was performed), individuals who determined length of stay and providers of postoperative care could have been blinded to procedure, which is a tactic that could have decreased bias. Given the magnitude of this limitation, Journal Club members agreed that the investigators should have discussed in more detail how blinding might have been accomplished and how lack of blinding could have shaped their results.
Sizing Up the Study
Calculating the appropriate sample size is essential for good research. A sample size that is too small prevents investigators from making accurate and reliable statistical judgments. If the sample size is too large, limited financial resources and participants' time are squandered. In the study by Candiani et al, the sample size of 60 patients was based on the proportion of patients who were discharged on day 2. Although a test of proportions was used for sample-size estimation, the probability values that were reported throughout the study were based on a 2-sided t-test of mean hospital stay in days. Thus, there is some inconsistency in the methods.
The procedures that were used to analyze data must be scrutinized carefully. As noted, hospital stay in days was compared by the t-test. Other outcomes that were compared by a t-test included operative time, blood loss, pain score, and number of days that analgesia was requested by patients. T-tests assume that the continuous data are distributed normally or that the sample size in each group is large enough to assume that the sample means are distributed normally by the central limit theorem. With 30 patients per group, the latter assumption may not hold; nonparametric statistical methods might have been more appropriate. The authors should have commented on whether distributional checks were performed to assess normality.
Through the Laparoscope
The authors of this study concluded that, based on the longer hospital stay that is associated with vaginal hysterectomy, the procedure was not superior to laparoscopic hysterectomy in this specific patient population. They also suggested that the laparoscopic approach may be preferable if a salpingo-oophorectomy is planned.
Journal Club members thought about how these results might affect clinical practice. The findings are applicable to a very limited subset of women who need a hysterectomy. Clinical decisions on surgical approach are individualized, and the surgeon's experience with a particular technique is also an important factor when choosing an approach, although this study tried to control for the participating surgeons' capabilities.
Still, Candiani et al embarked on an important task: questioning whether 1 type of hysterectomy was preferable to another—vaginal vs laparoscopic—in a group of patients where either was feasible. Journal Club members thought the authors should be applauded for investigating this question with an RCT. Although the findings are unlikely to make gynecologists abandon vaginal hysterectomy in favor of laparoscopic hysterectomy, it might cause them to consider the reason that they prefer a particular route and to examine their own preconceived ideas.
Kristen A. Matteson, MD, MPH; George A. Macones, MD, MSCE, Associate Editor
© 2009 Published by Elsevier Inc.