## Confounding Variables Influencing AFP Concentration

Fig. 4. Relative changes in the median AFP values in maternal serum and AF at each week during the early second trimester of pregnancy.

14 15 16 17 18 19 20 21 22 Gestational Weeks

Fig. 4. Relative changes in the median AFP values in maternal serum and AF at each week during the early second trimester of pregnancy.

per week (Fig. 4). In order to eliminate the need to have normative data for each week of gestation, patient results are expressed as the ratio of the median concentration in unaffected pregnancies at that gestational age. The median value is used instead of the mean value to avoid the influence of substantially elevated results that would skew the mean in small population samplings. The MoM result is potentially independent of the gestational age, allowing patients of all gestational ages to be assessed together on the basis of their MoM values. In order to achieve this independence of gestational age, the median values must be selected in a manner that avoids any bias throughout the range of reportable gestational ages.

Obtaining Normative Median Data. Medians are calculated from the AFP concentrations recently measured in a group of specimens that span the gestational age range to be used in screening. If the patient population being served is relatively static, acceptably robust median values can be calculated from 50 to 100 samples for each week of gestation. Knowledge about the normalcy of the pregnancies is not necessary because unusually high concentrations are balanced by low results—an inherent quality of the median rather than the mean of a population. Frozen specimens are also acceptable and particularly useful for establishing AFAFP medians, because AFP is quite stable in both amni-otic fluid and serum.

In a screened population, a gestational age of 16 wk is common, whereas 20 wk is uncommon. In order to lessen the inaccuracy of medians at outlying gestational ages, the observed medians are regressed (logarithms of concentration regressed against gesta-tional age, weighted by the number of observations in each category), thus making use of the known natural increasing (MSAFP) or decreasing (AFAFP) pattern of concentration with increasing gestational age. Accuracy in both measures—AFP concentration and gestational age—is required. Gestational ages can be measured in completed weeks (e.g., 16 wk 5 d is 16 completed weeks), but it is preferable to express the gestational age to the nearest day, or in decimal weeks, despite the fact that this implies a greater accuracy than is likely present within any individual patient. The median regression equation is then used to interpolate gestational age medians to the nearest day or decimal weeks for subsequent patients, until such time as the analytical conditions change, requiring a new assessment of medians.

The method of estimating the gestational age in patients should also be known because it imparts a small bias to the median values. Medians for AFAFP are virtually always based on ultrasound biometry. MSAFP medians will differ slightly with different methods of gestational age assignment (30). If the gestational age is inaccurate in a population, the true rise in MSAFP will be partially masked by that inaccuracy. Therefore, medians based on last menstrual period (LMP) rise less steeply than medians based on biometry (Fig. 4). Depending on the size of a screening program and its mix of LMP- and biometry-assigned gestational ages, some programs use different median regressions for these two methods of assignment (30).

Although LMP dating is sufficiently accurate to sustain screening, ultrasound biometry improves performance. At least two pieces of information are required to assign a gestational age: the date of the estimation and the estimated age or biometry measurement. The methods of assigning gestational age ranked in declining accuracy are crown-rump length (CRL) before 12 wk, biparietal diameter (BPD) after 12 wk, age based on composite multiple biometry measurements, and LMP dates. Physical examination and expected date of delivery (EDD) are less reliable. Ultrasound biometry is considered accurate within 8% of the assigned gestational age (31) or within 9-10 d at early midtrimester. Because different algorithms exist for converting biometry measurements to gestational age, it is preferable to collect the actual measurements and use an accepted algorithm for the entire screened population. The algorithms by Daya (32) for CRL and the 1982 BPD data of Hadlock (33) are an example of a pair of biometry algorithms with good concordance that span from 6 to at least 30 wk.

If more than one ultrasound examination is performed, a good first trimester CRL is preferred over a later BPD because biological variation in fetal size will start to become a factor. However, ONTD screening is improved if a BPD is used, for two reasons: first, anencephaly can be ruled out if a BPD measurement is reported by the ultrasonographer; and second, OSB-affected fetuses have smaller BPD measurements (34). A smaller BPD will cause the assignment of a lower than true gestational age and median value. The MSAFP MoM will therefore be, on average, elevated more than if another dating method had been used, increasing the likelihood of a screen-positive result in the presence of OSB.

Demographic and Clinical Factors That Influence the MoM Value

There are a number of patient demographic and clinical factors other than gestational age that influence AFP concentrations and are recommended for adjustments.

Racial Origin. Prenatal screening for fetal ONTD was first developed in largely white populations. Subsequently, it was found that the black (African American) population has 10-22% higher concentrations of AFP in both maternal serum and amniotic fluid in women with and without pregnancies with OSB (35-38). The effects of this racial difference, and possible remedies, will vary with the size of the populations being screened. In a population equally distributed by race, the overall population medians would be higher than those seen strictly in the white women, and lower than those in the black women. Screen-positive rates at any cutoff would be lower, and higher, than expected, respectively. This difference can be compensated either by the use of separate MSAFP

median data for the black and white populations, or by the use of a median adjustment factor (commonly 1.10 or 1.15) to modify the white medians for use in black patients. Even though AFAFP concentrations are also 12% higher in the black population (37), adjustment of AFAFP medians is not commonly performed because of the very small overlap between the affected and unaffected populations with this diagnostic test. The Asian population (e.g., Japanese, Chinese, Filipina) have the same, or only slightly higher, weight-corrected MSAFP concentrations compared to the white population (38,39), and median correction for this group is uncommon.

Maternal Weight. A relationship between maternal weight and MSAFP was first reported (40) in 1981 and then confirmed in the same year (41). The developing fetus is the source of MSAFP, and because fetal size in early pregnancy does not reflect maternal size, the circulating volume of the mother will indirectly affect the MSAFP concentration. Maternal weight is an easily accessible index of circulating volume. Maternal weight has no effect on AFAFP, nor is maternal weight different between OSB and unaffected pregnancies (42). Two weight correction formulae for MSAFP have been reported: one that fits the observed MSAFP and weight data to a log-linear relationship (43), and one that uses an inverse relationship (44). Published weight correction parameters are appropriate only for screened populations with the same average maternal weight as the published study (44); therefore, individual screening programs should calculate their own weight correction parameters for either of the two formulae. Screening without weight correction will result in a wider distribution of MSAFP MoM values, a higher screen positive rate in lower-weight women, and probably a slight loss in detection of OSB (42).

Number of Fetuses. Multiple gestation pregnancies have MSAFP (but not AFAFP) concentrations that are commensurate with the number of developing fetuses. Twin pregnancies have on average 2.16 times the MSAFP concentration of singleton pregnancies (45-47). Monozygotic twin pregnancies (one-third of all twin pregnancies) probably have slighly higher concentrations than dizygotic pregnancies (48,49). No adjustment is made to the MSAFP medians, but the MoM cutoff for serum samples from known twin pregnancies is selected at a higher value to maintain approximately the same FPR as in singleton pregnancies (see section on Twin Pregnancies).

Insulin-Dependent Diabetes Mellitus. Pregnant patients with a prior diagnosis of insulin-dependent diabetes mellitus have approx 20% lower concentrations of MSAFP (50,51). A factor is usually applied to the MSAFP MoM result to adjust for this. There is insufficient evidence for a similar adjustment in gestational diabetes, even if the patient is receiving insulin. AFAFP concentrations are also decreased in the presence of diabetes mellitus (52); however, adjustment for this is not common for the same reasons as cited for maternal race.

Use of Multiple MoM Correction Factors in Combinations. Although published data are scarce on the coexistence of multiple conditions that affect median MSAFP values, the assumption is that each factor is independent. Factors can therefore be com-bined—for instance, in a black diabetic pregnancy of a particular weight. Similarly, at present it is assumed that although black patients tend to have larger body frames, and Asians are smaller than whites, it is generally assumed that they all share approximately the same weight correction formula, populating the higher or lower ends of the weight correction curve, respectively.

## Post a comment