Letters
Volume 163 Number 2
693
Table I. Recalculated data of Milunsky et al.
RR and 95 % CI Prevalence (%) Sensitivity (%) Specificity (%) PTL+ (%) PTL- (%)
All outcomes
Outcomes with RR> 1.0
Outcomes with RR > 1.0 (excluding neural tube defects and chromosomes)
1.69 (1.48-1.92) 13.4 11.8 93.3 21.6 12.8
2.10 (1.80-2.45) 9.0 14.3 93.3 17.6 8.4
1.85 (1.56-2.19) 8.7 12.5 93.3 15.3 8.2
RR, Relative risk; C1, confidence interval; PTL +, positive posttest likelihood; PTL - , negative posttest likelihood.
We do not question the value of maternal serum 0:fetoprotein screening in the prenatal diagnosis of fetal neural tube defects and chromosome abnormalities. However, we question its value as a general screening tool for prediction of "high-risk" pregnancy. With the figures calculated for those outcomes associated with increased relative risk, maternal serum o:-fetoprotein screening detects only one in seven and misses six in seven of the pregnancies that progress to the adverse outcomes listed in Table I of Milunsky et al. Further, an abnormal test outcome results in false labeling of a pregnancy as "high risk" 8 times in 10. We do not believe these figures justify the use of maternal serum o:-fetoprotein screening as a general test for "high-risk" pregnancy. Use of maternal serum o:-fetoprotein as a general screen for "high-risk" pregnancy could be evaluated more thoroughly in a doubleblinded prospective trial of management comparing a "high-risk" protocol versus "routine" obstetric care in women with abnormal maternal serum o:-fetoprotein levels. Dawna M. Gilchrist, MD,jan M. Friedman, MD, PhD, and R. Doug Wilson, MD Department of Medical Genetics, University of British Columbia, 4500 Oak St., Vancouver, British Columbia, Canada V6H 3N1
REFERENCES 1. Sheps SB, Schechter MT. The assessment of diagnostic tests. JAMA 1984;252:2418-22 . 2. Sackett DL, Haynes RB, Tugwell P. The interpretation of diagnostic data. In : Clinical epidemiology. 1st ed. Boston: Little Brown, 1985:59-138.
Reply To the Editors: Gilchrist et al. are correct. Maternal serum o:-fetoprotein screening cannot yet be used "as a general screen" for identifying high-risk pregnancy. However, elevated maternal serum o:-fetoprotein values, in the absence of a detectable fetal defect, indicate a breach in the integrity of the maternal-fetal interface. Therefore our observations of increased relative risks for specific complications in a very carefully followedup cohort of 13,846 women should alert the obstetrician to exercise anticipatory surveillance. It would seem
injudicious to ignore a relative risk of 2 to 4.7 for oligohydramnios, abruptio placentae, preeclamptic toxemia, premature ruptured membranes, low birth weight, neonatal death, and other major congenital defects. In association with unexplained maternal serum o:-fetoprotein elevations, others have noted a significant increased incidence of low birth weight, I fetal loss, neonatal death, and various non leaking congenital defects. 2 The majority of pregnancies with elevated maternal serum o:-fetoprotein levels conclude with a normal child. Nevertheless, such observations may serve as an early warning and hence maternal serum o:-fetoprotein levels could also be used as an "important adjunctive tool." Aubrey Milunsky, MB, BCh, DSc, DCH, Hershel jick, MD, and Walter Willett, MD Center for Human Genetics, Boston University School of Medicine, 80 E . Concord St., Boston, MA 02118.
REFERENCES 1. Brock DJH, Barron L,Jelew P, et al. Maternal serum alphafetoprotein measurements as an early indicator oflow birth weight. Lancet 1977;2:267-8. 2. Burton BK. Outcome of pregnancy in patients with unexplained elevated or low levels of maternal serum alphafetoprotein. Obstet Gynecol 1988;72 :709-13.
Treatment of cervical pregnancy with methotrexate
To the Editors: I read with interest the report by Palti et al. (Palti Z, Rosenn B, Goshen R, Ben-Chitrit A, Yagel S. Successful treatment of a viable cervical pregnancy with methotrexate. AM J OBSTET GYNECOL 19R9: 161 : 1147-8). Although I appreciate the importance of this therapeutic modality, I would stress that a prerequisite is, of course, correct ultrasonographic diagnosis before severe hemorrhage occurs. Cervical suture, as described by Bernstein et al. I and referenced by the authors, still remains in the obstetrician's armamentarium as a potential conservative surgical aid to control massive hemorrhage in an acute situation usually undiagnosed before presentation (as originally described). With reference to the figure in the article, although Fig. 1, A, clearly represents a transabdominal ultrasonogram, it is not clear whether Fig. 1, B, was obtained transabdominally or by endovaginal ultrasonography.
694
Letters
August 1990 Am J Obstet Gynecol
Coronal section (as defined in Gray's Anatomy2 as vertical planes at right angles to the median-sagittal plane) of the pelvic organs is only obtainable by the endovaginal route. In fact a recent publication has suggested that the term trans-pelvic scan is more appropriate for the plane imaged when the sound beam is directed from side to side in the pelvis during endovaginal sonography.' Therefore, if Fig. 1, B , was performed transabdominally (as it appears to have been), it should be correctly termed a transverse section rather than a coronal section. Inasmuch as methotrexate has rarely been applied to viable pregnancies, it would have been interesting to note the histopathologic findings of this case, which presumably reflect the effect of methotrexate on this viable pregnancy. David M. Sherer, MD Department of Obstetrics and Gynecology, Division of Maternal-Fetal Medicine, Box 668, Strong Memorial Hospital, University of Rochester School of Medicine and Dentistry, 601 Elmwood Ave. , Rochester, NY 14642
REFERENCES 1. Bernstein D, Holzinger M, OvadiaJ, Frishman B. Conservative treatment of cervical pregnancy. Obstet Gynecol 1981 ;58:741-2. 2. Williams PL, Warwick R. Introduction. In: Williams PL, Warwick R ,eds. Gray's anatomy. 36th ed. Philadelphia: WB Saunders, 1980:XV. 3. Dodson MG, Deter RL. Definition of anatomical planes for use in transvaginal sonography. JCU 1990;18:239-42.
Forty-six percent of the study population to whom the test was applied (N = 41) had ectopic pregnancies. This proportion is much higher than that seen in the typical outpatient American population followed to rule out ectopic pregnancy. If the proportion of patients with true ectopics was only 20%, for example, then the positive predictive value of the three tests (absolute slope, relative slope, and human chorionic gonadotropin score) would be 82%, 82 %, and 83.7%, respectively. This is a clinically relevant decrease from the values for a 46% ectopic prevalence of 94. 1%, 94.1%, and 94.7%, respectively. (Note that the values in Table II are actually sensitivity for the test applied to diagnose intrauterine pregnancy and ectopic, not predictive value, as stated). Second, in contemporary American medical practice it is not possible to hospitalize for evaluation so many patients who ultimately prove not to have ectopic pregnancies (indeed almost 30% were not even pregnant). Thus the American physician will not always have the luxury of waiting 1 to 6 days with a patient in the hospital to gather information necessary to apply this test. Finally, it seems unusual that five of the patients with ectopic pregnancies but none of the patients with intrauterine pregnancies were resampled on day + 1 and only two (versus 11 patients with intrauterine pregnancies) were resampled after day + 3. Could this represent some bias in the application of the test? Susan T. Haas, MD
Reply To the Editors: We appreciate the letter by Dr. Sherer concerning our article. However, we disagree with him that coronal section of the uterus and cervix is obtainable only by the endovaginal route. With some experience , transabdominal coronal section of the uterus and cervix can be achieved by pressing the woman's lower abdomen by the sector probe, as clearly demonstrated in Fig. 1, B. With regard to his second question, histopathologic examination revealed nonspecific missed abortion pathologic changes. Z. Palti Department of Obstetrics and Gynecology, Hadassah University, P.O.B . 24035,}erusalem 91240, Israel
Caution in use of new test for ectopic pregnancy To the Editors: The article by Lindblom et al. (Lindblom B, Hahlin M, Sjoblom P. Serial human chorionic gonadotropin determinations by fluoroimmunoassay for differentiation between intrauterine and ectopic gestation . AMJ OBST ET GVNECOL 1989; 161 :397-400) contributes to the difficult problem of diagnosing ectopic pregnancy when serial human chorionic gonadotropin titers are rising. However, several issues relevant to interpretation of a new diagnostic test are not addressed in the article. I believe that understanding more about the value of tests under different testing conditions might influence significantly readers' use of these three new tests.
Department of Obstetrics and Gynecology, Brigham and Women's Hospital, 333 Longwood Ave., B oston, MA 02115
Reply To the Editors: We appreciate the considerations in Dr.
Haas' letter that highlight important principles of test evaluation. Her comments concerns three items: prevalence of the disease, hospitalization time, and sample interval. Concerning the prevalence of ectopic pregnancy in the study population (46 %), we agree that this figure may seem high. However, it was the real proportion of ectopic pregnancy in the group to which the human chorionic gonadotropin score was applied. We believe that the following factors account for the high prevalence: (1) liberal use of endovaginal ultrasound, which accurately (and indeed rapidly and easily) recognized viable intrauterine pregnancies in a large proportion of early pregnant women ; (b) the widespread knowledge in our community of risk determinants 1. 2 implies that few ectopic pregnancies were "lost" for serial human chorionic gonadotropin determination because of late presentation and emergency surgery; (3) exclusion of cases with decreasing human chorionic gonadotropin levels, which are not compatible with viable pregnancy. From the results presented, it is easy to extract the sensitivity, specificity, and positive/negative predictive values. We used the positive predictive value to compare the performance of the human chorionic gonadotropin score with the other methods applied (abso-