Evening Primrose Oil and Facilitation of Labor
July 2000; Volume 2; 53-55
Source: Dove D, Johnson P. Oral evening primrose oil: Its effect on length of pregnancy and selected intrapartum outcomes in low-risk nulliparous women. J Nurse Midwifery 1999;44:320-324.
Design: Retrospective cohort with data derived from record review.
Setting: Free-standing birth center.
Subjects: 108 nulliparous women who delivered at term with accurate dates and no pregnancy complications.
Treatment: Evening primrose oil in 54 women; no evening primrose oil in 54 women. Use of other treatments by women in both groups was not recorded.
Dose/Route/Duration: 500 mg three times daily for one week beginning gestation week 37 and once a day thereafter until labor ensued.
Outcome Measures: Apgar score, length of pregnancy, length of labor, birth weight, complications of labor.
Results: There were no differences in the planned outcome measures. Post hoc subgroup analyses suggested that evening primrose oil might have been associated with protracted active phase of labor. The authors additionally concluded that evening primrose oil was associated with a "trend" to an increase in arrest of labor and vacuum extraction (although trend testing was not done; see comments below).
Funding: Not indicated.
Comment by Anthony R. Scialli, MD
Evening primrose oil has been recommended in midwifery writings for the prevention of postdate pregnancies and for the shortening of long prodromal labors. The route of administration varies among midwifery practitioners; some recommend oral administration and others recommend topical application to the vagina or cervix. The mechanism of presumptive action is not known, but is presumed to involve cervical ripening.
One early factor in the initiation and maintenance of labor is a change in the connective tissue of the cervix, producing a soft, pliable organ that is prepared to efface and dilate. This effect probably is mediated by endogenous prostanoids and can be induced or accelerated by the local application of pharmaceutical prostaglandins.
This retrospective chart review asks whether using evening primrose oil is more effective than not using evening primrose oil in advancing the timing and shortening the duration of labor. Medical record documentation was the only source of information on whether the subject had been given evening primrose oil. The research method did not permit the investigators to discover potentially important information:
1. Were there clinically important differences between women who were given evening primrose oil and women who did not receive this therapy?
2. What other labor-promoting techniques were used? The authors indicate that there was no available information on the use of castor oil or intercourse by these women. Nipple stimulation, a cervical ripening technique recommended by some practitioners, also was not addressed.
3. Was the evening primrose oil used by those for whom it was prescribed, and not used by those for whom it was not prescribed? The prescription of evening primrose oil in the medical record may not reflect its use according to the recommended regimen. Conversely, the lack of a prescription of evening primrose oil in the record may not reflect lack of use of this intervention.
Only a prospective randomized clinical trial can answer these questions. Such a trial should include women with similar pretreatment cervical status (or at least cervical status information could be collected prior to treatment for inclusion in the data analysis). A randomized clinical trial permits instruction concerning compliance and avoidance of additional, unstudied interventions, although this instruction is no guarantee that subjects will not deviate from the protocol.
A retrospective record review might have been useful in identifying a possible effect of evening primrose oil if a significant difference in outcome had been shown between women receiving this treatment and women not receiving it. Had a large association existed between evening primrose oil treatment and short labor duration, for example, the authors would have been correct in proposing that the intervention may have been effective, even though unmeasured confounders actually may have caused the apparent treatment effect. Such a result would have increased interest in, and perhaps funding opportunities for, a placebo-controlled clinical trial. Given the current results, however, is it reasonable for the authors to conclude that there is "no apparent benefit from taking oral evening primrose oil for the purposes of reducing the incidence of adverse labor outcomes or decreasing the overall length of labor"?
There is no justification for this conclusion. The lack of an appropriate control group makes it impossible for this study to contribute meaningful information to the question of whether or not evening primrose oil is effective. For example, were women with unripe cervices more likely than women with ripe cervices to have been recommended evening primrose oil? If so, a similar outcome in initially dissimilar groups could reflect a therapeutic benefit of evening primrose oil. Did some midwives prefer recommending nipple stimulation to evening primrose oil and, therefore, might some of the "untreated" women actually have received an unreported intervention? If so, this study might be unknowingly comparing evening primrose oil with another treatment rather than with nontreatment. Did some women in the untreated group use evening primrose oil on their own, without reporting it to their health care providers? If so, treated women would have been misclassified as untreated, biasing the results toward the null hypothesis.
Given the limitations of the study design, the use of sophisticated analytic techniques and post hoc analysis of secondary endpoints can hardly be expected to yield meaningful results. In addition to the common error of using a parametric statistical test to compare Apgar score (which is a rank and not a continuous metric), there is the inexplicable use of tests that assume a normal distribution for data that are clearly skewed, such as length of labor. The authors miss an important analytical point here: The treatment being tested is designed to address the upper end of a skewed distribution, that is, women at the overlong extremity of pregnancy or labor length. (See figure 1.) The treatment works if some of these women are brought from the tail of the distribution closer toward the center. Testing of the mean and standard error of a distribution may be too insensitive a measure of a treatment that is aimed at the tail. The authors even go so far as to eliminate from the analysis one woman who was an "outlier" with a labor length of greater than 80 hours, when it is precisely this outlier at whom the intervention trial should have been directed.
The authors additionally concluded that evening primrose oil was associated with a "trend" toward an increase in arrest of labor and vacuum extraction, although the data did not demonstrate such an association and trend testing was not performed.
The word "trend" in this case was apparently used to mean "not statistically significant, but pretty close." This statistical offense is more serious because it is being used for post hoc analyses in an apparent effort to dredge up some findings. Given the exploratory nature of the analyses of labor complications (and the rather small number of women with such complications), the requirements for calling attention to a finding should be a more rigorous rather than less rigorous demonstration of significance.
To be fair, the authors conclude that "the need for further research is apparent." Too often, though, the call for further research is used as a way of legitimizing a study that failed to provide meaningful information. There is concern that this study might be cited by academicians or by authors of review articles as evidence that evening primrose oil is not effective. The problem of "garbage in, gospel out" could give a continuing life to this study, which does not contribute useful information on the effectiveness of evening primrose oil for facilitating labor.
July 2000; Volume 2; 53-55
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content