Effect of incorrect gestational dating on Down s syndrome and neural tube risk assessment

Transcription

1 Original Article Ann Clin Biochem 2001; 38: 230±234 Effect of incorrect gestational dating on Down s syndrome and neural tube risk assessment S N Millner From the Department of Pathology, University of Illinois at Chicago, Chicago, IL 60612, USA SUMMARY. Down s syndrome risks are estimated between 15 and 20 completed weeks gestation (cgw) using an algorithm involving maternal age and serum a- fetoprotein (AFP), chorionic gonadotrophin and unconjugated oestriol levels, each expressed as a multiple of the median level (MoM) at the cgw. The AFP MoM itself is the basis for screening for open neural tube defects (ontd). Because medians change during this period, gestational dating must be accurate so that appropriate medians are used. A calculated Down s syndrome risk >1:380 at term is generally considered to indicate a `high-risk pregnancy. This study focused on 378 patients with reported risk 41:500 based on physician-supplied cgw (and hence considered at `low risk for Down s syndrome) to determine the effect of common 1±2-week dating errors on risk estimates. Using the original analytical data, each patient s risk was recalculated for each week over the 15±20 weeks, and classi ed into three categories: <1:380 `low ; 1:380±1:100 `moderate ; and >1:100 `high. Advancing originally `low-risk patients by one week increased the risk by 1 09±14 1 times (median 3 18, mean 3 60); 46 (12 2%) became `moderate and 2 (0 5%) became `high risk. Advancing by two weeks increased risks 1 58±60 5 times (median 10 03, mean 12 04); 131 (36 5%) became `moderate and 39 (10.9%) became `high risk. Predictably, ontd screening results also were affected. Although 1±2 week differences in AFP medians had little effect on most patients in this study sample, some who originally were ontd negative became ontd positive, whereas others who had been ontd positive became screen negative. Thus, in many cases, a 1±2 week dating error may have only minimal effect on the estimated risks for chromosome or neural tube defects, but in other cases the effect of such an error would be signi cant. INTRODUCTION Elevated maternal serum a-fetoprotein (AFP) levels in pregnancy are generally regarded as an indicator of open neural tube, ventral wall and other developmental defects in the foetus, and also of various problems of pregnancy. AFP, combined with serum levels of human chorionic gonadotrophin (hcg) and unconjugated oestriol (ue3), constitutes the `triple test panel for the assessment of a baby s having a trisomy or triploidy. The most general applications are as an estimate of the risk of Down s syndrome due to trisomy 21, the most common congenital Correspondence: Dr S N Millner. smillner@vic.edu cause of severe mental retardation and, more recently, Edwards syndrome due to trisomy 18. Each of the three analyte levels is reported in terms of its concentration as a multiple of its median level (MoM) in populations of unaffected singleton pregnancies at the physicianspeci ed completed week of gestation (cgw). Maternal age and the three MoM are incorporated into algorithms for calculating the Down s syndrome risk, the one most generally used being that of Cuckle, Wald and coauthors. 1,2 It has long been known that the incidence of chromosome abnormalities is signi cantly affected by the mother s age. Although the risk increases with age, particularly above age 35, more babies with these defects are born to younger mothers owing to the greater number of pregnancies in the younger age groups. 230

2 Incorrect dating and Down s syndrome and neural tube risk 231 Nevertheless, in accordance with accepted screening practice, a patient with a risk 5 that of a 35 year old, i.e. approximately 1:270 at midtrimester and 1:380 at term, is considered `at risk. Less than 1:380 at term is considered `not at risk or, more reasonably, `low risk. It is not always appreciated that (1) the risks determined using the generally followed protocols are statistical in nature and do not indicate with certainty that a particular baby will or will not have Down s syndrome; (2) the protocols involve MoM which are `adjusted for maternal diabetes and race using factors that do not relate to the individual patient but which are based on trends recorded in large, speci cally de ned and characterized populations; (3) strong reliance is placed on the accuracy with which the patient s cgw is determined. Ultrasound dating is most accurate in the rst trimester; however, for many reasons an initial ultrasound may not be performed until the second trimester. Ultrasound accuracy in the second trimester ± the period in which this test is used ± has been estimated as from +7 days to +10 days, a range of two to almost three weeks. 3,4 Although many laboratories use day- rather than week-speci c medians the most common practice appears to use week-speci c medians, as only week of gestation (assumed to be completed week rather than truncated, i.e. rounded) is generally supplied by the obstetrics service. As a result, the analyte levels of, for example, a patient 15 weeks+6 days are related to the same 15-week medians as a patient 15 weeks+1 day, and not to 16-week medians. Clearly, a dating error of one day or a few days (within the accuracy level of ultrasound determinations pointed out above) could result in the use of 14-, 15- or 16-week medians, with substantial differences in reported risk. The less than optimal accuracy of dating is also evident from the fact that the laboratory is frequently asked to recalculate a reported risk following a reassessment of a patient s gestational dating. In other cases a recalculation may not be requestedat all if a change of 1±2 weeks is assumed to have little effect on a previously reported low risk. It is also not unusualfor a second specimen to be submitted with a cgw inconsistent with that indicated for the rst specimen. Williams et al. 5 have shown that the choice of algorithm and method of establishing medians may cause a signi cant variation in reported Down s syndrome risk. The present study was an attempt to determine how risks calculated using one algorithm might be affected by inaccuracies of 1±2 weeks or more in gestational age. This entailed re-entering each patient s analytical data into the algorithm and determining the Down s syndrome risks using the medians, not only at the patient s original cgw but at each cgw from 15 to 20 weeks. Whereas an algorithm involving serum AFP, hcg and oestriol is used to assess the risks of chromosome abnormalities,an elevation of serum AFP MoM alone is the basis of the most common biochemical screening test for open neural tube defects (ontd). In pregnancy, serum (and amniotic uid) AFP is also increased in ventral wall, renal and numerous other developmental defects. The generally accepted `cut-offs distinguishing `screen-positive and `screen-negative non-insulin dependent patients are a serum AFP 52 0 or 52 5 MoM, and 52 0 for patients who are insulin dependent. As AFP medians increase approximately15% per week during this period, it should be expected that ontd screening results also would be affected by what might otherwise be considered insigni cant dating errors. METHODS AFP was determined using an Abbott IMx, and more recently with an Abbott Axsym (Abbott Laboratories, Abbott Park, IL 60064). hcg originally was determined using an Abbott IMx, and more recently using a Bayer Immuno-1 (Bayer Corporation, Tarrytown, NY 10591). ue3 was determined by radioimmunoassay (Diagnostic Systems Laboratories, Webster, TX 77598). (Changes in analytical instruments were an administrative decision, not based on instrument accuracy or reliability. Analyte medians were calculated in accordance with accepted practice and were consistent despite instrument changes which affected neither the results reported to physicians nor the data collected for this study.) All original routine reports of analyte levels, MoM and Down s syndrome risks had been generated and stored using this author s `PACE software, an acronym for `Prenatal a-fetoprotein, chorionic gonadotrophin, oestriol. 6 In accordance with accepted procedure, the program reports MoM ± appropriately adjusted for the patient s weight, race and insulin dependence ± and risk of trisomy 21. It also indicates graphically the likelihood of various conditions contributing to the patient s AFP results and reports the possibility of other conditions,

3 232 Millner including trisomy 18 and triploidy. Because it stores patient data, the report also includes the results of previous submissions and notes dating inconsistencies. Data stored as PACE les were copied directly into a program that used the same risk algorithms to recalculate and tabulate each patient s risks at the original cgw, and at each week from 15 to 20 weeks, using the laboratory s established 15±20-week medians. From patient data routinely processed during the most recent period prior to this study, the 378 were selected who had indicated Down s syndrome risks 41:500. Because in general a 1:380 term risk distinguishes between `low and `high, 1:500 or less would generally be accepted as low risk for Down s syndrome. Although patient ages ranged from 14 to 45 years, the mean was 24 9 (median: 24; mode: 23) years. Racial makeup was 54% black, 4% oriental, 42% caucasian; 4% were insulin dependent. All were reported to be singleton pregnancies. Patients in this group were also ontd screen positive or screen negative. The small screenpositive group consisted of 15 non-diabetic patients with AFP MoM 52 5, and ve insulin-dependent patients with AFP MoM (Excluded were cases with extremely elevated AFP MoM, ranging from 5 1 to 8 6, which were associated with severe malformations and fetal demise.) Mean age was 27; median 26 years. The screen-negative group consisted of 80 non-diabetic patients with MoM 1 5±2 4 and 10 insulin-dependent patients with MoM 1 0±1 9. The mean age was 25 years, median 24. Patients with lower MoM were excluded, as it would be expected that only an unreasonable dating error would increase their AFP MoM suf ciently to become screen positive. For this study, this group was also limited to patients 516 weeks, as its purpose was to evaluate the effects of recalculating one- and two-week changes in cgw during the 15±20-week period for which the panels are valid. RESULTS AND DISCUSSION Many obstetricians consider any Down s syndrome risk <1:380 at term to be low risk and those >1:380 to be high risk. All the patients studied here ± with risks at the supplied cgw of <1:500 ± would therefore have been considered at low risk. For the present study, however, recalculated risks have been categorized into three groups: low risk, <1:380; moderate risk, 1:380±1:100; high risk, >1:100. Figure 1 shows the results obtained after recalculation of risks for two patients originally estimated as 17 weeks cgw. Patient F (Fig. 1a) was originally reported as having a risk of 1:1980. Even if she were actually as far along as 20 weeks, she would still be in the low-risk category. In contrast, Patient P (Fig. 1b) was at even lower risk (1:2300) if her original cgw of 17 weeks was correct. However, if Patient P were actually 18 weeks, i.e. one week later, her risk would be 1:308, shifting her into the moderate risk category. Her risk would be signi cantly higher ± 1:70 ± if she were actually 19 weeks. With regard to the entire group of originally low-risk patients studied here, advancing the original cgw by one week caused risks to increase from between 1 09 and 14 1 times (n=378, median increase 3 18 times, mean increase 3 60 times). Speci cally, 46 of these originally low-risk patients (12 2%) entered the moderate risk category; the lowest risk in this group had been 1:937. Two (0 5%) became high risk; the lowest risk of this pair had been 1:619. Advancing the cgw by two weeks caused risks to increase 1 58±60 5 times (n=359, median increase times, mean increase times). One hundred and thirty-one became moderate risk; the lowest risk in this group had been 1: Thirty-nine entered the high-risk category; the lowest risk in this group had been as low as 1:3500. No one factor, such as maternal age or insulin dependence, appeared to determine the rate at which risks were changed by changes in dating. It is evident that if the actual cgw is later than that originally determined, some patients would remain in the same risk category whereas others would clearly be at higher risk for Down s syndrome. It follows that the converse also would be true: if the cgw of a patient determined to be high risk is really less than that originally estimated by even 1±2 weeks, she might actually be at lower risk. Figure 2 shows the original term risk for Patient K, reported to be at 15 weeks. With a risk as originally reported of 1:590, she would be screen negative for Down s syndrome. As an insulin-dependent diabetic, however, her AFP MoM of 2 6 was ontd screen positive (i.e. MoM >2 0). If she were actually 16 weeks, i.e. one week later, her risk would become 1:160 and AFP MoM 2 2, i.e. she would still be at high risk for ontd and she

4 Incorrect dating and Down s syndrome and neural tube risk 233 FIGURE 2. Patient K: 41-year-old insulin-dependent caucasian, 172 lb; originally reported risk approximately 1:590 at 15 weeks. Plot drawn as described for Fig. 1. FIGURE 1. Originally calculated and recalculated Down s syndrome risks for two patients with originally supplied completed weeks of gestation (cgw) of 17 weeks. (a) Patient F: 20-year-old insulin-dependent diabetic, caucasian, 134 lb; based on age and analyte levels, her Down s syndrome risk was originally calculated as approximately 1:1980. (b) Patient P: 33-year-old non-diabetic oriental, 215 lb; originally reported risk approximately 1:2300. Plotted, in oneweek steps (Ð), are their risks (right ordinates) if they were actually at 15±20 weeks. In Figs 1a, 1b, and 2 a vertical dotted line is drawn at the original cgw supplied to the laboratory and a horizontal line (Ð - - Ð) separates `low from increased risk, 1:380. Curves showing a-fetoprotein (ÐÐ), human chorionic gonadotrophin (± ±) and unconjugated oestriol (- - -) multiples of the median (MoM) (left ordinates) also are shown at the respective cgw. (Curves are not smooth, as MoM were recalculated to only one decimal place.) also would enter the conventional high-risk category for Down s syndrome. At cgw 17 weeks she would be borderline in ontd screening with an AFP MoM of 1 9, but the risk of her baby having Down s syndrome would increase to 1:70. In general, the effect on ontd screening of changing cgw would be expected to depend on how close the initial MoM was to the screening cutoff. Advancing the cgw by one week allowed ve of six non-diabetic patients with original AFP MoM of 2 5±2 9 to become <2 5 and one insulin-dependent patient originally at 2 0 to become <2 0. One non-diabetic with MoM 2 8 required a two-week increase in cgw to become screen negative, as did two insulindependent patients with MoM originally 2 4 and 2 6, respectively. The cgw had to be advanced by two weeks or more for the ve patients with AFP MoM above 3 0 to become screen negative. As might be expected, three patients with AFP MoM 2 4 became screen positive when the cgw was moved back by one week, and 18 patients with AFP MoM 1 8±2 4 became screen positive when the cgw was moved back two weeks. Two of the insulin-dependent patients also became screen positive when the cgw was moved back two weeks. The 52 patients with AFP MoM 1 8 or less would have become screen positive only if their dating error was more than two weeks. In many cases, one specimen is submitted for analysis and no subsequent, timely recalculations are requested when dating is reassessed by one to two weeks. It is clear that a decision regarding how to proceed with a pregnancy may be made on the basis of an assessment of risks that may be incorrect due to erroneous dating. Dating error may put a patient on either side of

5 234 Millner a risk cutoff, so that she may be informed that she is at low risk whereas she is actually at higher risk; or she is told she is at high risk and only a subsequent dating reassessment shows that she had been at much lower risk. Because the chromosome defect test panel provides a statistical assessment of risks based on a number of statistical parameters and, along with ontd screening, assumes that dating is accurate, and in view of the margin of error of gestational dating, particularly in cases where it is done initially in the second trimester, it may be advisable to report risks routinely not only at the supplied cgw, but for a range of, for example, two weeks. This procedure might avoid undue complacency or concern on the part of both patient and physician. REFERENCES 1 Cuckle HS, Wald NJ, Thompson SG. Estimating a woman s risk of having a pregnancy associated with Down s syndrome using her age and serum alphafetoprotein level. Br J Obstet Gynaecol 1987; 94: 387±402 2 Wald NJ, Cuckle HS, Densem JW, Nanchalal K, Royston P, Chard T, et al. Maternal serum screening for Down s syndrome in early pregnancy. Br Med J 1988; 297: 883±7 3 Chervenak FA, Gabbe SG. Obstetric ultrasound: assessment of fetal growth and anatomy. In: Gabbe SG, Niebyl JR, Simpson JL, eds. Obstetrics: Normal and Problem Pregnancies. New York: Churchill Livingstone, 1996: 279±326 4 Stebbins B, Jaffe R. Fetal biometry and gestational age estimation. In: Jaffe R, Bui T-H, eds. Textbook of Fetal Ultrasound. New York: Parthenon, 1999: 47±57 5 Williams KL, Nix BAJ, Dunstan FDJ. Effect of screening algorithm, parameter values and median smoothing on patient-speci c risk estimates for Down s syndrome screening. Ann Clin Biochem 2000; 37: 165±73 6 Millner SN. Graphic reporting format for maternal serum alpha-fetoprotein and a comprehensive `P.A.C.E. pro le. Clin Chem 1991; 37: 953 [Abstract #208] Accepted for publication 27 December 2000