| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Letters |
Stanford University Palo Alto, CA 94310
Drs. Spring and Barkan claim that the findings of their prospective randomized study (conducted with Dr. Pruyn) [1] argue against the notion that use of oil-soluble contrast agents in hysterosalpingography (HSG) offers a statistically significant fertility benefit compared with use of water-soluble contrast agents. Their results are incongruent with those of numerous earlier trials, including a prospective randomized study by Rasmussen et al. [2] that showed a statistically significant fertility benefit for oil-soluble contrast agents. Whether discrepancies in methodology led to the contrasting conclusion in the trial by Spring et al. is unclear. Unfortunately, because the report by Spring et al. does not offer detailed background characteristics of each arm of the study, we cannot know if confounding variables such as prior surgery or infertility treatment contributed to the lack of observed benefit of using oil-soluble contrast agents.
In their commentary [3], Spring and Barkan state that their study was designed to have the statistical power sufficient to detect a 10% difference in live births and did not show a statistically significant benefit for the sequential use of water- and oil-soluble contrast agents. However, these statements do not reconcile with the original published report of their trial, in which the authors stated that use of the same combination of water- and oil-soluble contrast agents had been prematurely halted soon after being administered to 100 patients because of slow recruitment and that the statistical power to detect a 10% difference was neither sought nor achieved [1]. The authors also comment that the increased rates of term pregnancy after HSG may have been due to confounding variables such as frequency of intercourse. Although such factors may be relevant when comparing HSG with a placebo, theses issues are not germane when comparing different agents used for HSG because both groups of patients would be influenced by such variables.
On the other hand, the earlier prospective randomized study by Rasmussen et al. [2] showed a fertility benefit despite randomizing patients 3 months after HSG when all patients who became pregnant were eliminated from the study. The pregnancy rate during this 3-month waiting period was almost identical to the spontaneous pregnancy rate of normal fertile couples, suggesting that the diagnosis of infertility may often be overstated at the time of HSG. The study by Spring et al. [1] does not account for this factor. By randomizing patients after the waiting period, Rasmussen et al. excluded the group in which the diagnosis of infertility may have been overstated and ran the trial in the more challenging group of remaining patients who met infertility criteria that were more rigorous and severe than the criteria set in the trial by Spring et al. Despite running the study in a population who had more severe infertility, Rasmussen et al. showed a statistically significant fertility benefit for oil-soluble contrast agents versus water-soluble contrast agents. In addition to a higher pregnancy rate, the group who underwent HSG with oil-soluble contrast agents also had a statistically significant increase in the rate of normal outcomes of these pregnancies, which further supports the immunomodulation hypothesis [2, 4]. In the excluded group of patients who became pregnant within 3 months after their HSG, more patients who received oil-soluble contrast agents (13.3%) became pregnant than those who received water-soluble contrast agents (7%), suggesting that the exclusion at least did not bias the study in favor of oil-soluble contrast agents [4].
Three other earlier randomized trials did not report powers calculations and were too small to detect even a doubling in pregnancy rates [5–8], but a meta-analysis of these randomized clinical trials as well as six other nonrandomized controlled trials also showed the benefit of using oil-soluble contrast agents in HSG [4]. Although the debate is far from over, the balance of clinical evidence so far viewed in sum supports the notion that oil-soluble contrast agents may offer greater fertility benefits than water-soluble contrast agents, an effect we hypothesize may be partly attributable to the ability of oil-soluble contrast agents to modulate T helper (Th) cell balance in the female reproductive tract [4]. Our postulation is buttressed by emerging basic scientific evidence that reveals the adaptive significance of host-driven immune shift to Th2 bias during the luteal phase and gestation to promote fertility [4]. Although intriguing in its plausibility based on the evidence to date, the hypothesis requires further empiric validation through rigorous clinical and basic scientific investigations.
References
1 Kaiser Permanente Oakland Medical Center Oakland, CA 94611
2 Kaiser Permanente Santa Rosa Medical Offices Santa Rosa, CA
95403
Our randomized trial evaluated the relative fertility benefit of contrast media used in hysterosalpingography (HSG) [1]. The original design called for a three-arm trial. Dr. Yun suggests that discontinuing the sequential mixed media arm of our original three-arm study weakened our study's statistical power, and that its failure to find differences between the study groups in post-HSG reproductive success could be attributed to the resulting inadequate statistical power.
On the contrary, after discontinuation of recruitment for the study's third arm at 133 subjects, our study maintained adequate statistical power both for a two-way comparison of reproductive success associated with use of each of the two contrast media alone, and for a three-way comparison among all three arms. As we noted in the original paper [1], the statistical power analysis targeted success rates of 10% in the Sinografin (diatrizoate meglumine, Gensia Sicor Pharmaceuticals) group and 20% in the Ethiodol (ethiodized poppy seed oil, Savage Laboratories) group (a much smaller difference than between the 12% and 33% rates reported by Rasmussen et al. [2]). Our pretrial statistical power analysis found that a sample size of 237 per group was needed to have a statistical power of 80% to find that difference statistically significant at p < .05 [1]. After discontinuation of recruitment for the third arm, our study recruited a final sample of 260 for the Sinografin group and 273 for the Ethiodol group. The resulting statistical power for this two-way comparison is 87% [3]. Alternatively, let us examine the statistical power for detecting a two-fold difference of 24% following use of Ethiodol from the 12% reproductive success rate following use of Sinografin found by Rasmussen et al. (note that their study found a 33% reproductive success rate following Ethiodol). Our study had a 94% power to detect this difference at p < .05.
We also examined our statistical power to detect differences among success rates of 10%, 15%, and 20% in the three arms of our study using the sample sizes actually recruited for those arms. Our study had a power of 80% to detect this difference at p < .05 [3]. Our study thus had more than adequate statistical power to detect two-fold intergroup differences of other groups from the 12% reproductive success rate in the Sinografin arm. This is much smaller than the differences reported by Rasmussen et al. [2].
Yun states that Rasmussen et al. [2] waited for 3 months after the HSG before conducting the randomization. This confuses Rasmussen's study design. To evaluate the spontaneous pregnancy rate, the study by Rasmussen et al. had a 3-month waiting period after intake before conduct of the HSG (and, we presume, before randomization to determine which contrast media would be used in that examination). Forty patients successfully conceived during this waiting period. Rasmussen et al. then conducted the HSG using the contrast medium indicated by the randomization process. The randomization thus did not occur 3 months post-HSG. It is unclear what would be accomplished by randomizing 3 months after the intervention. The 40 subjects who conceived during this 3-month lag between recruitment and examination and 10 additional patients who did not want to undergo the examination were then excluded from the study by Rasmussen et al. The 3-month post-HSG conception rates reported by Yun in the above note occurred in the patients remaining in the sample after these 50 patients had been excluded. While this exclusion could be seen as introducing a sampling bias, it seems safer to conclude that this strategy ensured that those patients who underwent HSG might be less fertile. We used no such lag time in our project. Rather, we accepted the referring clinicians' diagnoses of infertility. However, we did investigate the duration of unsuccessful attempts to conceive. The average duration of the period during which our study participants attempted unsuccessfully to conceive was 3.12 years (95% confidence interval, 2.79–3.46 years). Of these women, 55.5% had been attempting to conceive for a year or more. We compared the duration of attempts to conceive between our study groups. The durations were statistically virtually identical (Sinografin, 3.15 years; Ethiodol, 3.13. yrs; mixed 3.09 years; F = .0086, p = .99, NS). We are confident both that participants in our study were indeed experiencing difficulty conceiving, and that there was no biased selection with respect to that duration among the study groups.
While Yun's hypothesis is interesting, he presents no actual new contrast-media-related data from fertile or infertile women in support of his hypothesis. He relies instead on reinterpretation of the results of selected prior studies. We would welcome the collection of any new data to examine immunomodulation resulting from the introduction of any contrast media that could impact reproductive success.
In conclusion, we too would be happy if one variation in radiographic technique could carry a therapeutic benefit for women seeking clinical attention for infertility. Unfortunately, the findings from our study, which had adequate and strong statistical power, show no such benefit.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
