986 Besinger et al.

13. Gerson A, Abbasi S, johnson A, et al. Safety and efficacy of long-term tocolysis with indomethacin. Am j Perinatol 1990;7:71-4.

14. Grella P, Zanor P. Premature labor and indomethacin. Prostaglandins 1978; 16: 1007-17.

15. Rubaltelli FF, Chiozza ML, Zanardo V, Cantarutti F. Effect on neonate of maternal treatment with indomethacin. j Pediatr 1979;94:161.

16. Goldenberg RL, Davis RO, Baker RC. Indomethacin­induced oligohydramnios. AM j OBSTET GVNECOL 1989; 160:1196-7.

17. Kirshon B, Moise Kj, Wasserstrum N, et al. Influence of short-term indomethacin therapy on fetal urine output. Obstet Gynecol 1988;72:51-3.

18 .Cabrol D, Landesman R, Muller j, et al. Treatment of polyhydramnios with prostaglandin synthetase inhibi­tor (indomethacin). AM j OBSTET GVNECOL 1987;157: 422-6.

Editors' note: This manuscript was revised after these discussions were presented.

Discussion DR. THOMAS H. KIRSCHBAUM, Los Angeles, Califor­

nia. This article extends a related study published in 1980. [ In the earlier work, which was a randomized blind study, the efficacy of indomethacin in inhibiting preterm labor was compared in two groups of patien~s: those treated with indomethacin and those who did not receive indomethacin. The study was not placebo con­trolled in a rigorous sense, because 30% of the women in the control group ultimately received other tocolytic drugs. Entry criteria were virtually the same as in this study. Treatment failure was defined as a progressive change in the cervix that occurred >2 hours after initial treatment or the attainment of >4 cm of cervical di­latation. Fifteen patients were studied in each group, and there was a statistically significantly larger inci­dence of treatment failure in the control group as op­posed to drug-treated patients at 24 and 48 hours after the initiation of therapy. No effect on outcome as measured by life-threatening newborn complications or perinatal death was seen, although the numbers of cases were relatively small. Mean values indicate that the drug-treated group averaged 3.2 weeks of prolongation of pregnancy after treatment, whereas the control group averaged 3.0 weeks of prolon­gation.

In this article the authors make two points. The first. is that indomethacin is less expensive than ritodrine, has fewer maternal side effects, and is as effective as ritodrine, as the outcome measurements indicate. The second point is that pulmonary hypertension was seen in two of the pregnancies and in three of the infants of 22 women who received indomethacin, whereas none of the 18 women treated with ritodrine experi­enced those neonatal complications. There are, I be­lieve, some questions that concern the validity of those conclusions, and I should like to pose two additional questions.

The differences in the reproductive histories of the two study groups are regrettable because they consti­tute uncontrolled variables. Eighteen of the ritodrine-

April 1991 Am J Obstet Gynecol

treated women had 20 prior pregnancies, which in­cluded seven abortions; 22 of the women treated with indomethacin had 52 prior pregnancies, which in­cluded 26 abortions. The interval to delivery and hence of drug therapy for both groups shows a great deal of scatter from the respective means. I calculate a mean and SD of 27.9 ± 22.9 days for ritodrine and 25.5 ± 20 days for indomethacin. Whether the unpaired t test might not be misleading for differences between means when frequency distributions are not in a gaussian curve is a serious question. Also, these results were obtained with the use of magnesium sulfate in about 30% of both groups after the initial drug failed to con­trollabor. This means that a comparison of the effects of indomethacin and ritodrine alone cannot be made from these data. In fact, the data suggest bimodal or multi modal frequency distributions. This problem is further exhibited in the authors' Table IV, in which the 8.9 average number of days on the ventilator for the infants born to women treated with ritodrine has a standard deviation of 13.9 days. This indicates a num­ber of extreme outliers form the mean value. Again, the choice of the statistical analytic technique used here is questionable.

These factors do not help to prove a negative hypothesis-here, that there is no difference between these two forms of therapy. This premise differs from that of the article written in 1980, in which a positive hypothesis was pursued. A negative hypothesis can never be proved; it simply displays the experimental design, the number of cases studied, and the data anal­ysis and makes the argument that if there were a dif­ference, it would have been discerned. In this study all three of those elements are arguable. As the authors acknowledge, the number of cases studied is too small to minimize the chance of false-positive errors or 13-errors. The presence of a difference between these two forms of therapy might have been seen if more cases had been studied or, for example, if a design that included no magnesium sulfate therapy had been used.

Would such a study have been ethically acceptable? I think so, if the work of others in this area is examined. The authors' 1980 article is one of two studies that purport to show the effects of indomethacin in inhib­iting labor for at least 24 hours. [.2 Two other studies show that the effect of indomethacin tended to delay labor for 48 hours2. 3 without effect on fetal and neo­natal deaths (five studies[") or respiratory distress syn­drome (four studies[' 2. 4. 5). I3-Sympathomimetic toco­lysis has been shown to be effective in delaying the delivery of infants in treated patients (as compared with control subjects) for 24 hours (five studies6

.1O

) or for 48 hours (two studies9

. [0) without affecting perinatal mortality (14 studies6

.1O

) or respiratory distress syn­drome (12 studies7

.1O

). It is a recurring observation that data in this area fail to show an effect on morbid­ity or survival, and in the absence of that evidence of benefit, not treating patients would seem to be accept­able. The maintenance of pregnancy with oral 13-sympathomimetic agents has not been shown to be of

Volume 164 Number 4

value, and no data exist in regard to the effectiveness of oral maintenance doses of prostaglandin synthetase inhibitors.

A final question relates to the average time of nearly 1 month that elapsed between treatment and delivery in these patients. The question, of course, is whether these patients were ever truly in preterm labor. Entry criteria consist of uterine contractions that occur roughly every 5 to 7Y2 minutes, cervical change, or a cervix with dilatation >2 cm or effacement >75%. It is hard to know what constitutes 75% of normal ef­facement, particularly in multiparous women, and the term change allows subjective evaluation. All of us have seen patients who were clearly not in labor but who met the entry criteria. Until direct biochemical markers of labor appear, this question will continue to make good experimental design, large numbers of cases, and careful statistical evaluation essential.

The data allude to the effect of long-standing in­domethacin treatment on ductus arteriosus stenosis and resultant pulmonary hypertension. Ductal patency is dependent on the production oflarge amounts of pros­taglandins by the ductus endothelium and muscularis. Inhibition of uterine contractions with indomethacin in fetal animals and in human neonates has been dem­onstrated to constrict the ductus and to produce fetal pulmonary hypertension with possible neonatal con­sequences. Two of 22 treated with indomethacin con­stitute evidence that suggests that the risks of indo­methacin treatment may be greater than the benefits. The use of indomethacin without clear evidence of ben­efit beyond the prolongation of pregnancy by 24 to 48 hours may be seriously questioned.

The authors have added new data that address this important area. Regrettably, they have not provided a great deal of clarification as a result of the intrinsic difficulty of research in which labor cannot rigorously and objectively be defined.

REFERENCES I. Niebyl JR, Blake DA, White RD, et al. The inhibition of

premature labor with indomethacin. AM J OBSTET Gy­NECOL 1980;136:1014-9.

2. Zuckerman H, Shalev E, Gilad G, Katzuni E. Further study of the inhibition of premature labor by indometh­acin. Part II. Double-blind study. J Perinat Med 1984; 12:25-9.

3. Spearing G. Alcohol, indomethacin and solbutomel: a comparative trial of their use in preterm labor. Obstet Gynecol 1979;53: 171.

4. Katz Z, Lancet M, Yemini M, Mogilner BW, Feigl A, Ben­hur H. Treatment of preterm labor with combined rito­drine and indomethacin. Int J Gynaecol Obstet 1983; 21:337-42.

5. Gamissan 0, Balasch J. Prostaglandin synthetase inhibi­tors in the treatment of pre term labor. In: Fuchs F, Stub­blefield PC, eds. Preterm birth: causes and management. New York: Macmillan, 1984:223-48.

6. Wesselius-De Casparis A, Thiery M, Yo Le Sian A, et al. Results of double blind, multicenter study with ritodrine in preterm labor. BMJ 1971:3,144-7.

7. Ingemarsson I. Effect of terbutaline on premature labor: a double blind placebo-controlled study. AMJ OBSTET Gy­NECOL 1976;125:520-4.

Comparative trial of indomethacin and ritodrine 987

8. Christensen KK, Ingemarsson I, Leideman T, Salum H, Sveuningsen M. Effect of ritodrine on labor after pre­mature rupture of the membranes. Obstet Gynecol 1980;55: 187-90.

9. Larsen JF, Eldon K, Lange AP, et al. Ritodrine in the treatment of preterm labor: second Danish multicenter study. Obstet Gynecol 1986;67:607-13.

10. Leveno KJ, Guzick DS, Hankins GD, Klein VR, Young DC, Williams ML. Single-centre randomized trial of rito­drine hydrochloride for preterm labor. Lancet 1986; I: 1293-6.

DR. DAVID COTTON, Houston, Texas. I have some comments and then a question. As you know, Dr. Moise in our group has been very interested in the effects of indomethacin on the fetus. Because of his work that associates closure of the fetal ductus and oligohydram­nios with indomethacin treatment of pregnant women, we have a rather cumbersome protocol for follow-up of these patients. In that protocol we evaluate ductal patency once a week and assess amniotic fluid volume twice a week.

We have seen the occurrence of fetal ductal closure and tricuspid insufficiency within about 48 hours of the initial administration of indomethacin, and we have also seen tricuspid insufficiency that did not resolve in 48 hours. In one case tricuspid insufficiency actually lasted 3 days but did finally resolve. Thus the adverse effects of indomethacin can sometimes be seen fairly quickly.

We have never been able to define the timing of amniotic fluid reduction. In some women a reduction in amniotic fluid volume develops after only 3 or 4 days of indomethacin therapy. Other patients will be treated with indomethacin for as long as 3 weeks before a de­crease in amniotic fluid results. We still feel. that it is very important to monitor amniotic fluid volume in patients, because we have had some evidence that neo­natal renal failure may occur in women who have had prolonged oligohydramnios.

However, the side effects described in this study pertain to those in the J3-sympathomimetic group that are usually seen very early on and that in many instances start to decrease over a 3-day to 7-day period if the patient becomes asymptomatic. Terbuta­line may continue to have its effect of relaxation of the uterus.

Amniotic fluid must be measured on a regular basis via ultrasonography. The side effects of a medication may initially be more complicated to follow, but after an initial evaluation, sophisticated monitoring is not required, even if the ductus is not evaluated. Do you think these agents really are comparable in terms of how easy and how expensive they are for the patients to take them?

DR. ROBERT RESNIK, San Diego, California. Several years ago, Dr. Irwin Merkatz wrote a multiinstitutional collaborative study on the effects of ritodrine, which led to the approval of ritodrine by the Food and Drug Administration. As I recall (and he may wish to com­ment on the clarity of what I'm going to say), the mean

988 Besinger et al.

number of days gained before delivery in the control group was 21. In Dr. Niebyl's study the mean was 25 days in the indomethacin group and 28 days in the ritodrine group.

Would you comment on this, in light of Dr. Kirsch­baum's previous comments, and define the onset of true labor to aid in determining the efficacy of these drugs?

DR. WARREN M. CROSBY, Oklahoma City, Oklahoma. The neonatologist who treats primary pulmonary hy­pertension in which· the ductus remains open is in a real bind, because once he or she closes the ductus, the pulmonary hypertension becomes worse. This study does not comment on the three babies that were treated in that manner. I would like to know if those babies became clinically better or worse.

DR. THOMASJ. GARITE, Orange, California. We have just completed an analysis of some data and have found an interesting association between indomethacin and necrotizing enterocolitis. Did you see any cases of nec­rotizing enterocolitis in your group?

DR. NiEBYL (Closing). First of all, I'll try to answer Dr. Kirschbaum: His questions about the statistics are well taken, but I think that we used the appropriate tests for the number of patients. We had only 40 pa­tients, and I think that applying a more sophisticated method might have gotten us into trouble.

The point about untreated controls is an important one that several people have reiterated. We had orig­inally done a placebo-controlled trial with indometha­cin, which showed that indomethacin treatment was effective, at least during that 48 hours of therapy. The important questions really concern the effectiveness of indomethacin in prolonging pregnancy or in making a significant difference in the health of the baby. I'm not sure that it doel'.

The reason we designed this trial, in which patients receive treatment with the drug for a month, was really to answer those questions. This study was simply a com­parative study of the effects of ritodrine and indo­methacin treatment. The study was not done to ex­amine the whole issue of whether untreated control subjects do as well as drug-treated subjects.

I think that the answers may be somewhere in the

April 1991 Am J Obstet Gynecol

middle; a certain subgroup of patients will benefit from treatment, but there are also certain patients who are undoubtedly not in preterm labor and who, if left un­treated, would not be delivered of their infants. It's very difficult to identify those patients. We certainly know that patients who come in repeatedly for exam­inations and who do not experience cervical change are still at risk of preterm delivery. Thus there are some groups of patients who may benefit.

There's certainly a placebo effect if all of the placebo­controlled trials or Untreated control trials are exam­ined. There is a success rate of 20%,30%, or sometimes even as high as 50% to 60% in avoiding preterm de­livery without treatment. Our problem is the difficulty in diagnosing which patients are really in preterm labor.

Dr. Cotton mentioned the work on Doppler ultra­sonography by the group at Baylor. I think that if I were designing this study to start today; I would in­corporate some sort of Doppler follow-up with these babies, because as the Baylor study indicates the con­striction can happen very quickly. We did monitor am­niotic fluid volume. We did not look at the ductus. Dr. Cotton's question really related more to which med­ication is the more difficult to use. I think his point that ritodrine is more difficult to use initially is well taken. Tachycardia ot other problems may result immediately with ritodrine treatment, but tolerance to the medica­tion may develop ina patient with long-term therapy. I feel that trouble with indomethacin may occur in long­term therapy, as it did in the few cases that we reported.

Dr. Crosby asked me whether infants who had pri­mary pulmonary hypertension in which the ductus re­mained open improved after closure of the ductus. All those babies responded very well, did well in the nurs­ery, and went home without problems. They were sick initially, but they did respond to therapy and did not have any significant long-term problems.

Dr. Carite asked about necrotizing enterocolitis. We didn't see any, but the association has been reported in the literature. However, this type of enterocolitis does not occur often, and there has been no significant as­sociation of the disease with indomethacin treatment.