Predictive value sensitivity, and specificity of ultrasonic targeted imaging for fetal anomalies in gravid women at high risk for birth defects

Predictive value, sensitivity, and specificity of ultrasonic targeted imaging for fetal anomalies in gravid women at high risk for birth defects Rudy E. Sabbagha, Zubie Sheikh, Ralph K. Tamura, Sharon DalCompo, Joe Leigh Simpson, Richard Depp, and Albert B. Gerbie

Chicago, Illinois In this report the predictive value of ultrasonic targeted imaging for fetal anomalies (TIFFA) is defined. Six hundred fifteen pregnant women at high risk for birth defects were scanned from January, 1980, to December, 1983. Follow-up evaluation was available on 569 fetuses. The pregnancies were classified into five groups according to the indications used for ultrasonic targeted imaging studies. The largest number of women were placed in group 1 and were referred because of a variety of abnormalities in previous or ongoing pregnancies. The women classified in the other four groups were examined because of maternal or fetal reasons related to specific craniospinal (29%), urinary (7.9%), gastrointestinal (6.7%), and skeletal (3.7%) defects. In our series the predictive values of abnormal and normal ultrasonic targeted imaging studies were 95% and 99%, respectively. A detailed breakdown of the accuracy of ultrasonic targeted imaging in relation to each anatomic category is presented; these data are useful in counseling gravid women with anomalous fetuses . (AM J OSSTET GVNECOL 1985;152:822-7.)

Key words: Accuracy of ultrasound, fetal anomalies Detailed ultrasonic targeted imaging for fetal anomalies (TIFFA) or Stage II ultrasound is not routinely performed in most of the diagnostic ultrasound centers in the United States. Its use is basically intended for pregnant women at high risk for birth defects. The reasons are as follows: The cost-to-benefit ratio of routine use is not established; prolonged exposure to ultrasound for targeted imaging examinations is deemed unnecessary in the face of a low yield of anomalies in normal pregnancy; the accuracy of targeted ultrasonic imaging is not fully established. This study is designed to determine the predictive value of targeted ultrasonic imaging examinations to : (I) enable us to better counsel patients referred for this specialized procedure and (2) serve as a basis for comparison with the findings of other, similar departments.

Patients and methods We conducted a prospective study from January, 1980, to December, 1983, to compare the ultrasonic results with those of the neonates. Six hundred fifteen pregnant women at high risk for anomalies had an ultrasonic examination in the form

From the Department of Obstetrics and Gynecology, Northwestern University and Northwestern Memorial Hospital. R eceivedfor publication December 14,1984; revised April 4, 1985; accepted April 25, 1985. R eprint requests: Rudy E. Sabbagha, M .D. , Prentice Women's Hospital, Ultrasound Department, Suite #110, Chicago, IL 60611.

822

of detailed targeted imaging for fetal anomalies. They were referred from a large pool including the Northwestern Perinatal Center as well as primary providers of care. The referrals represented a predominantly midwestern population of middle socioeconomic level but of different ethnic backgrounds. During the period of the study the number of referrals increased sharply from 162 in 1980 and 1981 to 453 in 1983 and 1984. Information on outcome was obtained by direct correspondence with the parents or by direct telephone communication with either the parents or the pediatrician(s) involved. Outcomes were available in 596 pregnancies, which form the data base of this report. Patients were classified into five groups according to the indications used for initiating ultrasonic targeted imaging studies. Three hundred fourteen patients were classified in group 1. They were referred for a variety of abnormalities including: (1) high level of a-fetoprotein in amniotic fluid; (2) suspicion of abnormality on a standard or basic (Stage I) scan, including polyhydramnios and oligohydramnios; (3) fetus presenting as breech at term ; (4) intrauterine growth retardation ; (5) insulin-dependent maternal diabetes mellitus ; (6) other reasons, such as exposure to teratogenic agents, preceding cervical cerclage, balanced translocations, lack of family medical history due to adoption, and anxiety. Additionally, we included women at high risk for fetal congenital heart disease in group 1. However, in these patients ultrasonic targeted imaging examina-

Ultrasonic targeted imaging for fetal anomalies 823

Volume 152 Number 7, Part I

Table I. Indications and results of ultrasonic targeted imaging studies performed on pregnant women at high risk for anomalies involving the central nervous, genitourinary, gastrointestinal, and skeletal systems Diagnosis by

Anomaly

Central nervous system Anencephaly H ydroce phalus Spina bifida Encephalocele Holoprosencephaly Microcephaly Total Genitourinary system Polycystic kidney Multicystic kidney Renal agenesis Ureteropelvic junction obstruction Ureterovesical obstruction Urethral obstruction Total Gastrointestinal system Duodenal atresia Gastrointestinal obstruction below duodenum Omphalocele Gastroschisis Diaphragmatic hernia Hirschsprung diseasett Total Skeletal system Limb reduction defect Split hand syndrome Arthrogryposis:j::j: Total

Positive family history

36

32 43

CorTeet diagnosis Basic scan

Other

Normal

6 28

7

48

5*

41 9 3

37

It

9 3

35

13

26 4 2

Abnonnal

Nonna l

5 18 5 1

1:j:

140

30

2:j:

24

2

I§

2

I§

4 2

1 1 I

1

3

7

3

35

II

3

2 19 I

I

False diagnosis

I

Abnonnal

I 1

2

2

125

Result of study

7~

12 2 4

I

2#

9

1**

18 I 3

10

3

22

16

15

5

2

111

I

24

13

20

1 1

22

I

15

6

*High a-fetoprotein level in four patients and hydrocephalus in one. t Hydrocephalus. :j:Same patient had false normal diagnosis of hydrocephalus and encephalocele. -i §Oligohydramnios. IIResolution of hydronephrosis, ~Mesoderm defect of muscle (two fetuses) . # H ydramnios. ** High a-fetoprotein level. ttEquivocal finding by ultraso und , deleted from results. :j::j:Mild disease, deleted from results.

tions were directed only at the four-chamber anatomic structures to rule out primary and secondary atrial septal defects, single atrium or ventricle, hypoplastic ventricle(s), and large ventricular septal defect. Patients were informed that in utero diagnosis of heart defects was still in a developing phase and only information regarding the structural and dynamic components noted in the four-chamber view of the heart could be offered, The remaining 282 pregna ncies included women at

high risk for specific anomalies involving the central nervous system (group 2), urinary system (group 3),

gastrointestinal system (group 4), and skeletal system (group 5), Classification in groups 2 through 5 was based on a positive history and suspicion of an anomaly on a basic scan (Table I), Although scanning was done before 24 weeks of pregnancy in many patients, some were not referred until the third trimester. Additionally, in patients at high risk for anomalies that may not be apparent by 24 weeks' gestation (for instance, infantile polycystic kidney disease and heterozygous achondroplasia) repeat scans were specifically requested by 28 and 32 weeks of pregnancy. Our definition of "correct abnor-

824

Sabbagh a et al.

August I, 1985 Am J Obstet Gynecol

Table II. The overall predictive value, sensitivity, and specificity of ultrasonic targeted imaging examinations performed on 594 pregnant women at high risk for congenital anomalies Congenital anomaly Study results

Abnormal Normal Total

Present

78 3 81

I

Absent

Total

4 509 513

82 512 594*

Predictive value of abnormal examinations = 78 of 82 or 95%. Predictive value of normal examinations = 509 of 512 or 99%. Sensitivity = 78 of 82 or 95%. Specificity = 509 of 512 or 99%. *Two fetuses with equivocal results (Hirschsprung disease and arthrogryposis) were deleted from analysis of the total group, reducing the total number to 594 pregnancies.

mal" included: (1) Accurate ultrasonic diagnosis of the defect made prior to 24 weeks' gestation. The diagnosis was also considered accurate if the abnormality was not initially noted (for instance, at 18 weeks) but was subsequently visualized on a repeat scan that we scheduled. (2) The anomaly under consideration was visualized in the third trimester of pregnancy simply because the patient was not referred at an earlier date. These guidelines were also used for the definition of "correct norma!." Incorrect diagnoses were classified as "false normal," "false abnormal," or "equivocal" (Table I). The ultrasonic consultation for ultrasonic targeted imaging studies required an average of 1 hour and encompassed: (1) documentation of the history and reason for referral; (2) performance of a basic ultrasound scan (Stage I), usually by a non physician professional trained in obstetric and gynecologic sonography; and (3) detailed imaging of fetal structural anatomy from a variety of sagittal, transverse, coronal, and oblique planes by a physician ultrasonographer (sonologist).

Results In the study population of 596 women, two fetuses with equivocal results were deleted from the predictive value analysis. In the remaining group the predictive values of abnormal and normal targeted ultrasonic imaging studies (without reference to the prevalence of disease) were 95% and 99%, respectively (Table II). Group 1. The largest number of women (314 of 594 or 53%) were classified in group 1. Of these pregnancies, 21 of 594 or 3.5% had polyhydramnios, diagnosed by the presence of amniotic fluid in excess of 7 to 8 cm, measured in a plane vertical to the sagittal axis of the uterus, and six of 594 or 1% had oligohydramnios,

Table III. Abnormalities detected by targeted imaging for fetal anomalies in high-risk pregnant women with a variety of findings in previous or ongoing pregnancies not pertaining to the central nervous, urinary, gastrointestinal, or skeletal systems (group 1) No.

Outcome

Polyhydramnios

21

Oligohydramnios

6

Cystic hygroma Nonimmune hydrops Sacrococcygeal teratoma

6 3

Mass in chest

2

Normal (16/21 or 76%) Tracheoesophageal fistula (1121) Trisomy 18 (1121) Endocardial fibroelastosis (1/21) Intestinal obstruction (2/21)* Severe growth retardation (2/6) Spontaneous abortion (2/6) Renal agenesis (2/6)* Pregnancies terminated Fetal/neonatal death Neonatal surgical procedure Cystic adenomatoid malformation (I) Iatrogenic pneumomediastinum and false abnormal (I) Premature labor Neonatal surgical procedure (benign)

Abnormality

2

Chorioangioma Ovarian mass Total

42*

*Of 42 positive findings, 20 were associated with amniotic fluid volume abnormalities and normal fetuses; of the remaining 22 fetuses, three fetuses had conditions not demonstrable by ultrasound, one fetus with a mass in the chest had a false abnormal result, one fetus with a false normal diagnosis of renal agenesis was counted in Table I, two fetuses with correct abnormal diagnoses of gastrointestinal obstruction and one fetus with a correct abnormal diagnosis of renal agenesis were counted in Table I, and 14 had correct abnormal diagnoses.

diagnosed by a paucity of amniotic fluid and overcrowding of fetal parts. The anomalies noted in this group are listed in Table III. No anomalies were detected in 30 insulin-dependent diabetic gravid women. Similarly, no cardiac defects were noted in 20 patients at high risk for congenital heart disease. The outcomes in four fetuses with chromosomal translocations were as suggested by ultrasound, norma!' In eight pregnancies the amniotic fluid a-fetoprotein level was >3 SDs above the mean and the sonographic diagnoses were: spina bifida (five fetuses), omphaloceIe (one fetus), cystic hygroma (one fetus), and normal (one fetus). Group 2: Central nervous system anomalies. One hundred seventy-three fetuses were scanned for possible craniospinal defects (Table I). In this series the

Ultrasonic targeted imaging for fetal anomalies

Volume 152 Number 7, Part 1

825

Table IV. Predictive value, sensitivity, and specificity of ultrasonic targeted imaging examinations in the diagnosis of spina bifida Spina bifida Study result

Present

Abnormal Normal Total

5 1 6

I

Absent

Total

1 41 42

6 42 48

Predictive value of abnormal examinations = five of six or 83%. Predictive value of normal examinations = 41 of 42 or 97.6%. Sensitivity = five of six or 83%. Specificity = 41 of 42 or 97.6%.

predictive values of abnormal and normal results were 96.7% and 98.6%, respectively. In the subset of fetuses

with spina bifida the predictive value of an abnormal result was lower than that of the group with craniospinal defects (Table IV). Group 3: Urinary system anomalies. In fetuses examined sonographically for possible urinary abnormalities (Table I) the predictive values of abnormal and normal results were 92% and 97%, respectively. Two fetuses with abdominal muscle deficiency anomaly were incorrectly diagnosed as having urethral obstruction (see Comment section). Group 4: Gastrointestinal system anomalies. In 40 high-risk pregnancies ultrasound studies were performed for possible fetal gastrointestinal anomalies (Table I). The predictive values of abnormal and normal results were 94% and 100%, respectively. In one fetus at high risk for Hirschsprung disease the ultrasonic findings were equivocal and the patient was excluded from the predictive value calculation. Group 5: Skeletal system anomalies.~Twenty-two fetuses were examined for the possibility of short limb dysplasia (Table I). A correct diagnosis was made in all six affected fetuses. The findings in one fetus at high risk for arthrogryposis were equivocal and the patient was excluded from the predictive value calculation. Comment

In this series ultrasonic studies involving detailed targeted imaging for fetal anomalies were performed on 596 pregnant women at high risk for birth defects. Two fetuses were excluded because of equivocal results. In three fetuses the diagnosis could not have been made by ultrasound (Table III). In the remaining group there were 81 fetal anomalies (13.6%), four false abnormal results (0.6%), and three false normal results (0.5%). By comparison, in a large number of women examined ultrasonically by Campbell,' 17% had fetal anomalies and the rates of false abnormal and false normal results were 0.3% and 0.8%, respectively.

Fig. 1. Scan showing cross section of fetal trunk in sacral area. The iliac wings are shown by large arrows and the open spina bifida by small arrows. Neural tissue (two white echoes) is seen in the center of the defect. The bladder (b) is anterior.

Analysis of subsets in 173 fetuses at high risk for craniospinal defects (group 2) showed that all cases of anencephaly were correctly diagnosed, for a detection rate similar to that reported by other studies. 2 . 3 Hydrocephalus was missed in one fetus who also had absence of skull bones anteriorly, encephalocele by definition. The fetus was examined only once and although the sonogram was suspicious in that ventricular dilation was borderline the report was signed out as normal. Although a variety of craniospinal defects were suspected by basic scans, performed elsewhere, in 35 fetuses none of these diagnoses included spina bifid a (Table I). The latter remains one of the most difficult abnormalities to diagnose even by ultrasonic targeted imaging examinations.' The predictive value of abnormal ultrasonic targeted imaging examinations in the diagnosis of spina bifida was 83%; by contrast the specificity or predictive value of a normal result was 98% (Table IV). The latter statistic is particularly useful in counseling women with an elevation in the level of amniotic fluid o:-fetoprotein, even when acetylcholinesterase is also elevated. The reason is that these biochemical tests are associated with false normal and false abnormal results. 3 • 4 In one report by Hobbins et al.,3 in 14 of 28 or 50% of pregnancies with elevated o:-fetoprotein levels the offspring were devoid of overt anomalies. The fetus in whom a spina bifida was missed in this study was also hydrocephalic and the amniotic fluid volume was very small. Because of oligohydramnios the scan was suboptimal. In the one fetus in whom a false abnormal diagnosis of spina bifida was made (Table IV)

826

Sabbagha et al.

August I, 1985 Am J Obstet Gynecol

Fig. 3. Echogram showing cross section of fetal trunk in the lower chest area (H = heart). T he spine is anterior (large arrow). The fetal bladder (BL) is dilated and appears outside the trunk (delineated by small arrows) as in omphalocele. The fetus, however, had an abdominal mu scle deficiency anomaly in which mesoderm is lacking in the abdominal wall and urinary tract musculature.

Fig. 2. Photograph of open spina bifida in the sacral area measuring 1.8 cm and visualized by ultrasound (see Fig. I).

the scan was performed early in the series. Retrospective review showed that the image produced was slightly oblique resulting in an artifactual V-shaped or cuplike spine simulating a thoracolumbar spina bifida. The sensitivity for detectio n of spina bifid a in this series of 83% fell between those reported in two other studies, namely, 79% and 93 %. 1. 2 However, it should be appreciated that in these studies data were collected about 5 years prior to the publication of any reports, when resolution of the ultrasonic equipment was less optimal and experience of sonologists in this area not extensive. Recently, with the equipment now available it was possible for one of us to visualize a spina bifida <2 cm in size and situated in the lumbosacral area (Figs. 1 and 2). Such small lesions were previously reported to be ultrasonically undetectable.2 In one fetus at high risk for infantile polycystic kidney disease, an autosomal recessive disorder, the anomaly did not manifest itself until after the twenty-fourth week of pregnancy.' The earliest reported diagnosis of infantile polycystic kidney disease was at 18 weeks' gestation. 6 In one pregnancy complicated by oligohydramnios renal agenesis was missed at term (Table I). Retro-

spective review of the scans showed that kidne y-like echoes were indeed noted paraspinally. Apparently at term such echogenicity can arise from fat deposition in the pararenal spaces 7 and/or fetal adrenal glands." Thus in suspected cases of renal agenesis ultrasonic examination should be directed toward outlining the fetal bladder. Toward this end intravenous administration of furosemide (60 mg) to the mother may be considered. s However, the beneficial effect ofthis approach particularly in the presence of intrauterine growth retardation is doubtful." A false abnormal diagnosis of unilateral ureteropelvic junction obstruction was made in one case; this was attributed to spontaneous resolution, a phenomenon reported by other investigators. 1o A diagnosis of ureterovesical obstruction was made in two cases. Marked convolutions of the ureters differentiated this condition from that secondary to urethral obstruction. Both fetuses exhibited marked enlargement of the heart (without cardiac structural defects) and died in the neonatal period, despite successful in utero shunting. Of interest is that two of seven patients referred for possible in utero shunts to drain a dilated urinary system (secondary to possible urethral obstruction) had an abdominal muscle deficiency anomaly (Fig. 3)." Garris et al. 12 described this disorder clearly and classified it as prune-belly syndrome, subgroup 3. Insertion of in

Volume 152 Number 7, Part I

utero diversion shunts for this condition may not be necessary. In two fetuses the diagnosis was made after 24 weeks' gestation. Nelson et al. I3 doubt that this condition can be diagnosed prior to the twenty-fourth week of pregnancy and conjecture that the amniotic fluid swallowed will only exceed the resorptive capacity of the gut in the early part of the third trimester of pregnancy. The prenatal diagnosis of this condition is important, regardless of when it can be made, because it leads to investigation of the fetal karyotype and permits detailed planning for delivery at an optimal site as well as anticipatory aspiration of the gastric contents and rapid neonatal confirmation of the defect. \1 In four fetuses with gastrointestinal obstruction below the duodenum, the correct diagnosis was made. Differentiation from simple distention, noted in some large-for-gestational age fetuses at term, was made by a marked increase in bowel size, thinning of the walls of dilated segments, and the presence of mild ascites. The false abnormal diagnosis of omphalocele in one case (Table I) may have been the result of contraction of the abdominal musculature, which created an artifactual image of a small omphalocele. The differentiation between omphalocele and gastroschisis was described by Jassani et al.l< It should be remembered that up to 50% of fetuses with omphalocele will have other associated abnormalities, particularly cardiac and chromosomal. In the diagnosis of skeletal dysplasia (Table I) the accuracy of ultrasound was enhanced because five of six fetuses were severely affected and the anomaly readily manifested. In the fetus with heterozygous achondroplasia (a product of parents with the same condition) the diagnosis of limb shortening was not apparent at 16 weeks of pregnancy," ·but the femur length at 24 weeks was 3.5 cm, well below 2 SDs of the mean. Detailed ultrasonic targeted imaging examinations in our center by experienced physician ultrasonographers failed to verify the presence of anomalies in 23 fetuses examined ultrasonically elsewhere. In these referrals false diagnoses were made of hydrocephalus (n = 17), anencephaly (n = 1), urethral obstruction (n = 2), duodenal atresia (n = 1), omphalocele (n = 1), and diaphragmatic hernia (n = 1). These observations attest

Ultrasonic targeted imaging for fetal anomalies

827

to the importance of properly conducted ultrasonic targeted imaging studies. In this report the predictive values of abnormal and normal ultrasonic targeted imaging examinations were 95% and 99%, respectively. However, in the general population the accuracy will depend on: (1) expertise of the examiner; (2) prevalence of the anomaly in specific geographic locations; (3) whether the population undergoing examination was screened by history, biochemical tests, or basic ultrasound scans or for other pregnancy complications; and (4) period of follow-up, since subtle, nonlethal anomalies may be undetected early after birth. REFERENCES 1. Campbell S. Ultrasound in obstetrics and gynaecology: recent advances. Clin Obstet Gynaecol 1983;10:475. 2. Robinson HP, Hood VD, Gibson AAM, et al. Diagnostic ultrasound: early detection of fetal neural tube defects. Obstet Gynecol 1980;56:705. 3. Hobbins jC, Venus I, Tortora M, et al. Stage II ultrasound examination for the diagnosis of fetal abnormalities with an elevated amniotic fluid alpha-fetoprotein concentration. AMj OBSTET GYNECOL 1982;142:1026. 4. Milunsky A, Sapirstein VS. Prenatal diagnosis of open neural tube defects using the amniotic fluid acetylcholinesterase assay. Obstet Gynecol 1982;59: I. 5. Simpson jL, Sabbagh a RE, Elias S, et al. Failure to detect polycystic kidneys in utero by second trimester ultrasonography. Hum Genet 1982;60:295. 6. Shenker L, Anderson C. Intrauterine diagnosis and management of fetal polycystic kidney disease. Obstet Gynecol 1982;59:385. 7. Dubbins PA, Kurtz AB, Wapner Rj, et al. Renal agenesis: spectrum of in utero findings. JCU 1981 ;9: 189. 8. Hadlock FP, Deter RL, Carpenter R, et al. Sonography of fetal urinary tract anomalies. AjR 1981;137:261. 9. Romero R, Cullen M, Grannum P, et al. Antenatal diagnosis of renal anomalies with ultrasound. III. Bilateral renal agenesis. AM j OBSTET GYNECOL 1985; 151 :38. 10. Sanders R, Graham D. Twelve cases of hydronephrosis in utero diagnosed by ultrasonography. j Ultrasound Med 1982; I :341. 11. Smith DW. Recognizable patterns of human malformation. Philadelphia: WB Saunders, 1976. 12. Garris j, Kangarloo H, Sarti D, et al. The ultrasound spectrum of prune-belly syndrome. JCU 1980;8: 117. 13. Nelson LH, Clark CE, Fishburne jI, et al. Value of serial sonography in the in utero detection of duodenal atresia. Obstet Gynecol 1982;59:657. 14. jassani MN, Gauderer MWL, Fanaroff AA, et al. A perinatal approach to the diagnosis and management of gastrointestinal malformations. Obstet Gynecol 1982;59:33. 15. Filly RA, Golbus MS, Carey jC, et al. Short-limbed dwarfism: Ultrasonographic diagnosis by mensuration of fetal femoral length. Radiology 1981;138:653.