Metronidazole for Posthemorrhoidectomy Pain: Does No Statistical Significance Mean Insignificant Effect on Pain Relief?

    loading  Checking for direct PDF access through Ovid

Excerpt

With great interest, we read the meta-analysis recently published in Diseases of the Colon & Rectum.1 We congratulate and applaud Wanis et al1 for their interesting work, but several important issues should be noted.
First, the meta-analysis merely included patients who were administrated oral metronidazole after hemorrhoidectomy, with the conclusion that the prescription of oral metronidazole seemed to have no pain reduction effect on most postoperative days. Regardless of no statistical significance, oral metronidazole was associated with a possible tendency in relieving postoperative pain after visual inspection of the pooled effect in Figure 2. Understandably, the lack of statistical significance largely resulted from including studies with 337 patients. When limited studies with small sample size are included in a meta-analysis, the pooled effect should be interpreted with caution after considering subsequent random error and false negatives.2 Accordingly, with sample size accumulation, what we can foresee is that updated meta-analyses including more studies with larger sample sizes will reach a favorable conclusion with respect to the use of oral metronidazole and achieve statistical significance. Under current circumstances, should we ignore the potentially significant pain-relieving effect of metronidazole because of the lack of statistical significance arising from a small sample-size effect? The authors should at least make some mention of this in the Discussion section. In addition, the meta-analysis only incorporated patients with a prescription of oral metronidazole, but the title involved systemic metronidazole. Is oral equal to systemic? Maybe this issue also puzzled other readers.
Lastly, the standardized mean difference was unreasonably used to pool continuous variables. Generally, the standardized mean difference is a better alternative when included studies all assess the same outcome but measure it in different ways.3 Given that all of the included studies assessed pain using visual analog scale, the weighted mean difference should be more preferentially chosen to pool continuous variables in the current meta-analysis.
    loading  Loading Related Articles