Advertisement

Reply

      To the Editors:

      Our study found a significant p value when objective cure rates of Burch colposuspension and paravaginal repair were compared (100% vs 61%, p = 0.004). Thus we appropriately rejected the null hypothesis (which states that no difference is expected between the two operations) for the alternative hypothesis (namely, there is a real difference in the two cure rates).
      • Peipert JF
      • Metheny WP
      • Schulz K
      Sample size and statistical power in reproductive research.
      I used Fisher's analysis because an expected value of the contingency table was <5. However, if I had used the χ2 test with Yates' correction, the p value would still have reached significance (χ2 6.38, p = 0.01). The relative risk is 1.64 (95% confidence interval 1.13 to 2.37).
      A study that finds a significant p value is defined a positive study. Power analyses must be performed in studies that find a nonsignificant p value, that is, when the null hypothesis cannot be rejected (negative studies).
      • Peipert JF
      • Metheny WP
      • Schulz K
      Sample size and statistical power in reproductive research.
      This is not our case. In fact, a simple calculation gives a power of 86% for the above comparison between our two objective cure rates. Furthermore, only 15 patients in each arm (a total of 30 patients) would have been enough to reach a power of 80%. As every reader can see, our data do not lack power at all. However, this is not simply a problem of statistical nature. Even if, to date, our study is the only available objective assessment of paravaginal repair, the experience of a single center must obviously be confirmed. I hope that other authors will conduct further objective trials to support or counteract our observations.
      Miklos and Saye assumed a general success rate of 85% for the Burch colposuspension. However, I take the liberty of remembering that none of our patients had undergone previous surgery and that those with urethral sphincter weakness or detrusor instability were excluded. This means that our population represents selected patients with primary uncomplicated pure genuine stress incontinence. Moreover, subjects were followed up for a relatively brief period. I believe that in such circumstances a success rate around 95% should be presupposed.
      • Penttinen J
      • Kaar K
      • Kauppila A
      Colposuspension and transvaginal bladder neck suspension in the treatment of stress incontinence.
      • Kilholma P
      • Mäkinen J
      • Chancellor MB
      • Pitkänen Y
      • Hirvonen T
      Modified Burch colposuspension for stress urinary incontinence in females.
      We began our trial with the conviction that the paravaginal repair would have been an excellent antiincontinence operation. I now discourage its use in treating stress incontinence. Nonetheless, I do not believe that it should be excluded from our surgical armamentarium. At our institution we are continuing to perform the paravaginal repair but exclusively in continent patients (or those with only minimal degrees of incontinence) with a cystocele from a lateral defect who are undergoing abdominal surgery for other reasons.

      References

        • Peipert JF
        • Metheny WP
        • Schulz K
        Sample size and statistical power in reproductive research.
        Obstet Gynecol. 1995; 86: 302-305
        • Penttinen J
        • Kaar K
        • Kauppila A
        Colposuspension and transvaginal bladder neck suspension in the treatment of stress incontinence.
        Gynecol Obstet Invest. 1989; 28: 101-105
        • Kilholma P
        • Mäkinen J
        • Chancellor MB
        • Pitkänen Y
        • Hirvonen T
        Modified Burch colposuspension for stress urinary incontinence in females.
        Surg Gynecol Obstet. 1993; 176: 111-115