What are truncation limits?

"Truncation limits" is a term used to define the upper and lower values of a distribution of a variable beyond which more extreme values are taken to be equal to the value of the relevant truncation limit in the calculation of LIKELIHOOD RATIOS and risk. This is usually because there are too few data available to be sure that the modelled distributions are accurate in their tails.

What is meant by capping a risk estimate?

In estimating risks occasionally very high and very low risks can be generated. It may be that there is a degree of uncertainty as to whether these risks are accurate. For example, if a woman had a risk estimate of having a pregnancy with Down's syndrome of 1 to 100,000 there may be too few empirical data to validate this. In this event it may become appropriate to, say, cap the risk as being less than or equal to 1 in 50,000. Similarly with a high risk of say, 20 to 1 one might wish to limit this to say equal to or greater than 5 to 1.

How do I estimate centile values?

If all the values of a variable are ranked from the lowest value to the highest, the middle value (or if there are an even number of values the average of the two middle ones) is the median or 50^th centile. Other centile values can separate the distribution into n% above the n^th centile and (100-n) % below the n^th centile.

People are often unsure of the best method for estimating the detection rates and false positive rates in Down’s syndrome screening.

There are two main methods, the study specific method and the general method. The general method is the standard and preferred method because it provides more stable estimates of screening performance than the study specific method, which is subject to random sampling error. Also, while the study specific method is specific to the age distribution of the women in that particular study, the general method is not, an advantage that allows comparisons between studies.

For a detailed description of the two methods, see the article on the Wolfson Institute of Preventive Medicine website.

How do I check my false-positive rate?

Determining the false-positive rate itself is not always possible, since it requires knowledge of the outcome of all pregnancies. However, affected pregnancies are relatively rare, and the SCREEN-POSITIVE RATE is, for practical purposes, a good approximation of the false-positive rate.

To examine the screen-positive rate, use MS Report Summary in the Statistics menu. Enter the time period required, and specify Include all codes (unless you want the summary restricted to specified doctor or address codes). A table similar to the following is shown. The screen-positive rate for Down's syndrome is the percentage with increased risk of Down's, while for neural tube defects (NTD) it is the percentage with raised AFP.

How can I check if my false-positive rate is what it should be?

It is often thought that a false-positive rate (FPR) of 5% is somehow 'correct', because a 5% FPR is commonly used in the literature as a standard when comparing the performance of different screening tests. However, the correct FPR for your population will not necessarily be the same as the published estimates. It will depend on a number of factors, including the age distribution, which may differ from the age distribution of the standard populations used in scientific papers.

When assessing whether the FPR you observe in your screening programme is correct, that is, the one that would be expected for the population you screen, it is important to take into account the following factors which influence the FPR:-

• the age distribution of the screened population
• the combination of screening markers used
• whether an ultrasound scan is routinely used to date pregnancies
• whether MoM values are corrected for maternal weight
• the risk cut-off used

alpha provides a way of answering this question, using the Age distribution facility on the Tabulations menu. As well as showing the maternal age distribution, it provides tables showing the expected screening performance for the population screened, using different combinations of markers. For example, a centre using the quadruple test (AFP, uE3, total hCG and inhibin) might see a table like the following:-

In this example, where the risk cut-off is 1 in 300, and ultrasound dating and weight correction are used in most cases, the expected screening performance is given at the intersection of the row and column indicated in the table above. The expected performance is expressed in terms of  the DETECTION RATE (DR), the FALSE-POSITIVE RATE (FPR), and the ODDS OF BEING AFFECTED GIVEN A POSITIVE RESULT (OAPR).

How can I check that my normal median values are accurate?

alpha provides a broad range of quality control and monitoring features, to help you determine the performance expected from your screening programme, and ensure that the expected screening performance is achieved.

An important routine monitoring task is examining the normal median values of the screening markers, so that any drift can be identified and corrected. The Graph Monthly Medians feature in alpha displays the median reported MoM values from month to month in graphical form.

If the median MoM is consistently above or below 1.0 MoM, this may be an indication that the normal median equation does not accurately estimate the median concentration of the screening marker, and needs updating.

The Tabulation options, in the Statistics menu, allow you to tabulate screening markers by gestational week over a specified time period, and print a regression equation for the tabulated data. The coefficients of the new regression are calculated automatically.

LMS offers an alphaCheck service free of charge to licensed alpha users. We will guide you through the steps involved in monitoring screening performance, and provide expert advice in correcting any problems identified. Please contact us if you would like an alphaCheck.

Yes. If you change the reagents used to assay the screening markers, you will need to establish normal median values for the new reagents. Ideally, this is done by assaying approximately 200 samples using the new kit, in parallel with the old one.

Samples should preferably be uniformly distributed between the weeks, say 50 each in weeks 15-18 for second trimester markers, or weeks 10-13 for first trimester markers. Smaller numbers are acceptable in other weeks.

Having collected the measurements for the new assays, the data should be tabulated by gestational week. For each week, record the median gestational age in days, the number of samples, and the median concentration of the screening marker.

You can use a spreadsheet or database software to help with calculating the weekly medians.

For example, a tabulation for AFP might look similar to the following:-

Having tabulated the data, use the Regression option, on the Statistics menu in alpha, to derive a regressed median equation for the new kit. The regression for the AFP data above is shown below. The A and B values printed below the graph are the coefficients of the regressed median equation. In order to use the new normal medians, you need to enter the coefficients, using Add coefficients on the System menu.

Regression Line Equation

Median MS-AFP = A*B**GA (days)

A = 2.972833
B = 1.020696

This represents a 15.4% increase in median MS-AFP per week

How can I see if predicted risk estimates in my screening programme are accurate?

Recent validation studies have shown that risk estimates for Down's syndrome produced by alpha are remarkably accurate.

The validation technique works by assigning each woman in the screened population to one of a number of predicted risk categories, chosen so that each category contains approximately the same number of cases of Down's syndrome. The mean predicted risk in each category is then compared to the prevalence of Down's syndrome among the women in the same category.

alpha users can perform this validation for themselves, providing a sufficiently large number of women have been screened, using alphaOutcome.

Click here for a tool to show an example of calculations of confidence intervals in the detection rate.

Yes. You can use the alpha Marker Library to add extra screening markers, which can be used instead of, or in combination with, the triple test markers (AFP, uE3 and total hCG).

The alpha Marker Library is supplied free of charge with alpha, and offers alpha users unparalleled flexibility in the choice of screening markers. You can add first and second trimester serum markers, and the first trimester ultrasound marker nuchal translucency.

To add a new marker, such as inhibin-A, you need to:-

Contact us if you need information on installing a new marker in alpha.

A diabetic woman has an AFP level of 2.3 MoM - why does alpha say her AFP is raised, when my cut-off is 2.5 MoM?

AFP levels are, on average, lower in women with insulin-dependent diabetes (IDDM) than in non-diabetic women. The reduction is about 23% if AFP MoM values are not adjusted for maternal weight, and about 12% if the MoMs are weight-adjusted (this is because women with IDDM are, on average, heavier than other women, and part of the reduction in their AFP levels will be due to increased blood volume associated with their greater weight).

To account for the reduced levels, alpha divides the AFP MoM in a diabetic woman by the appropriate correction factor (0.77 without weight adjustment, or 0.88 with weight adjustment), and compares the adjusted MoM with the cut-off. In this example, the adjusted MoM would be 2.3/0.77 = 2.98 MoM without adjustment for weight, or 2.3/0.88 = 2.61 MoM with adjustment for weight, both of which are higher than the cut-off of 2.5 MoM.

The unadjusted MoM is printed on the report.

How does alpha interpret nuchal translucency (NT) and first trimester serum markers in a twin pregnancy ?

In a monozygotic twin pregnancy, the expectation is that both twins are affected or both are unaffected, but in a dizygotic twin pregnancy, whether one twin is affected will be unrelated to whether the other is affected. Because of the close association between chorionicity and zygosity, and because monozygotic twins are less common, it is reasonable to adopt the following approach (see reference 1):-

If the chorionicity is unknown, alpha assumes that the pregnancy is dichorionic.

In a dichorionic pregnancy, alpha estimates the risk separately for each twin, based on half the maternal age-specific risk, and the NT measurement for each twin. The risk for the pregnancy is then the sum of the two twin risks.

In a monochorionic pregnancy, alpha calculates the risk for the pregnancy based on the woman's age and the average of the two NT MoMs, because in expectation each twin will have the same risk.

If biochemical markers are measured, the risk based on the NT measurements and the maternal age becomes the new background risk for the pregnancy, and alpha combines this in the usual way with the biochemical marker levels to give a PSEUDO-RISK for the pregnancy.

1. Wald NJ, Rish S (2005). Prenatal screening for Down syndrome and neural tube defects in twin pregnancies. Prenat Diagn 25, 740-745

I want to analyse my screening data using Microsoft Excel. Can I transfer the data from alpha into a spreadsheet?

Yes, alpha provides  a facility called Analyze-it which exports data directly to an Excel spreadsheet.

In addition, alpha provides "Data Transfer", which makes it easy to send screening data from alpha to other applications, such as spreadsheets and databases. The data transfer facility is useful if you want to examine your data in a way that alpha does not cater for. The data transfer options are on the Statistics menu in alpha.

• You select the columns in the alpha database which you want to export
• If desired, you can restrict the exported records according to doctor, report address and report date
• You can export data in a variety of formats suitable for spreadsheets and databases
• Your export specification can be saved for later reuse

Can alpha be interfaced with other computer systems or laboratory equipment?

Yes. Transferring data electronically between two systems avoids the need for manual data entry on two systems, and reduces the chance of data entry errors. alpha provides several options which allow you to so this:-

• patient data and test results can be imported from other systems into alpha by means of ASCII files
• a worksheet option allows biochemical test results to be collected automatically from an analyser
• screening results can be exported from alpha to other systems in a variety of formats

Linking alpha with other systems may require additional interfacing software. We would be pleased to discuss your needs with you. Please contact us to let us know your requirements.

Why is the risk of trisomy 18 not shown on reports when it is below the cut-off?

(i) Uncertainty of risk estimation

There is no complete set of established distribution parameters with correlations and satisfactory early mid-trimester prevalence estimates. Most of the studies used to estimate parameters were interventional which can introduce bias. It has however been shown that risk estimates above about 1/200 to 1/300 are reasonably accurate but less so for lower risks. Accordingly, ?lpha does identify such pregnancies as being at high risk with an appropriate comment ("Increased risk of trisomy 18") but is otherwise silent on an interpretation.

(ii) Effect on the false positive rate

In Alpha the trisomy 18 risk is displayed only when the risk is above the cut-off recognizing the uncertainty described in (i) above. If a trisomy 18 risk were printed when it was below the cut-off it could cause unnecessary anxiety (reference 1). For example, a risk estimate of 1 in 350 or 1 in 400 could cause concern, particularly on realizing that the background risk is about 1 in 7000. Such women may be sufficiently worried to have an amniocentesis with the attendant risk of losing healthy fetuses for each affected trisomy 18 fetus detected. The trisomy 18 amniocentesis rate could reach 1%. The problem is made worse now that, with current tests, false positive rates for Down's syndrome can be as low as 1 to 2%. It would be unsatisfactory if there were nearly as many amniocenteses for trisomy 18 as for Downs Syndrome with little or no increase in the trisomy 18 detection. This means that issuing risk estimates for all is difficult to defend medically as well as scientifically.

(iii) Lack of knowledge of the ratio of viable trisomy 18 fetuses detected to healthy fetal losses from invasive diagnostic tests

A risk analysis is needed in respect of live born trisomy 18 infants versus additional amniocentesis (or CVS) required to detect them. Trisomy 18 is ten times less common than Down's Syndrome and most cases miscarry. There should not be too many amniocenteses (and consequent loss of healthy fetuses) per viable trisomy 18 pregnancy detected. Knowledge of this ratio is needed in determining screening policy (and it will vary according to risk cut-off). While this is approximately known in women with very high risks it is not known at all levels of risk.

1. Wald NJ and Canick JA (2002) Seeking other disorders within antenatal serum screening programmes for Down's syndrome J Med Screen 9:145-146

Can a warning be given if the screening result appears to be heavily influenced by a single marker measurement?

Alpha can identify cases where a single marker has a very large influence on the risk estimate raising the possibility that the risk estimate may be incorrect (reference 1). In these cases, Alpha will notify the user of the anomalous marker measurement and give them the opportunity of removing it from the risk estimate in the screening report.

Alpha does this by calculating the risk of a Down's syndrome pregnancy using all the markers, and then recalculated for screen-negative women with each marker omitted in turn. Screen-positive women do not have their risk recalculated, thereby avoiding the possibility of Down's syndrome pregnancies being reclassified as screen negative. If the risk changes by a factor of more than three hundred, the screening result is flagged as anomalous. If the risk changes substantially when only one marker is omitted, appropriate corrective action can be taken, such as re-assaying the serum sample or reviewing the hard copy of the NT image. If no errors are found, then the risk estimate could be recalculated without the marker value concerned, and this is reported with an appropriate comment.

1. Wald NJ, Bestwick JP, Barnes IM, Kellner LH (2007) Anomalous marker patterns in Down syndrome screening Prenat Diagn 27:185-186

Why is the screening report positive when a previous pregnancy is affected with Down's syndrome or neural tube defects?

A screening report is classified as Screen Positive when a pregnancy was previously affected with Down's syndrome because screening for Down's syndrome and neural tube defects is not simply based upon risk estimation. It is also based upon having previously had an affected pregnancy. Although the absolute risk may still be low, the fact that a woman has had an affected pregnancy is itself, in effect the result of a "positive screening test". The test or question is "Have you had a Down's syndrome or neural tube defect pregnancy?" Yes is positive, and No is negative in this screening enquiry.

The practical reason for treating a previously affected pregnancy in this way is to alert medical staff to the need for specific counselling. Otherwise, the fact could be overlooked and the woman may not receive appropriate counselling, may not have an amniocentesis when she would have wished to have one and there would be the possible view that medical care offered was not what it should have been.

In practice, a woman's risk of having an affected pregnancy can be presented to her in the context of her having had a previously affected pregnancy. She can then make a decision as to whether she wants an amniocentesis. With appropriate explanation the policy should be clear and understood.

Why is screening for neural tube defects not carried out at 14 weeks?

Although second trimester Down's syndrome screening can start at 14 weeks gestation, neural tube defect screening only starts at 15 weeks. AFP does not discriminate well between open spina bifida pregnancies and unaffected pregnancies before 15 weeks (reference 1). The parameters used in Alpha are published from 15 to 22 weeks gestation (reference 2)

1. Report of UK Collaborative Study on Alpha-fetoprotein in relation to neural tube defects (1977). Maternal serum alpha-fetoprotein measurement in antenatal screening for anencephaly and spina bifida in early pregnancy. Lancet June 1977,1323-1332

2. Wald NJ, Cuckle HS, Densem JW, Kennard A, Smith D. (1992). Maternal serum screening for Down's syndrome: the effect of routine ultrasound scan determination of gestational age and adjustment for maternal weight. Br J Obstet Gynaecol 99,144-149

What is the "scan update rule"?

In Alpha the scan update policy controls the reinterpretation and reclassification of screening tests from "screen positive for increased risk of Down's syndrome" to "screen negative" after the addition of ultrasound scan information. You can choose to reinterpret the test always, or to do so only if the new scan estimate of gestation differs from the 'dates' estimate by at least a specified number of days (between 1 and 28 days). It is desirable to set a limit (say 7 days) so converting a true-positive to a false-negative will be very rare. Tests that are not reinterpreted (because the difference in gestation is too small) are reported with the same screening result ("screen positive") as the original report, with the addition of a message indicating that the original interpretation remains unchanged.

Restricting the reclassification of screening results in this way helps to avoid giving false reassurance to women with affected pregnancies who have been given a positive screening result based on dates (reference 1). Such revision of the test result who had initially been told they were screen positive is particularly distressing. The remedy is to perform a scan on all women at the time of screening, thereby avoiding the need to revise the gestational age estimate after a woman has been screened and found to be positive.

1. Wald NJ, Kennard A, Densem JW, Cuckle HS, Chard T, Butler L (1992). Antenatal maternal serum screening for Down's syndrome: results of a demonstration project. BMJ 305, 391-394

What is the best gestational age for first trimester tests?

The median nuchal translucency, PAPP-A and hCG MoM levels change with gestational age during the first trimester.

The screening performance for NT is greater at 11 weeks than at 12 or 13 weeks, for PAPP-A it is better at 10 weeks and for hCG at 12 - 13 weeks. The best time to take the measurements is therefore a compromise and it is reasonable to target 11 weeks as the best time for obtaining first trimester measurements (references 1 and 2)

1. Wald NJ, Rodeck C, Hackshaw AK, Walters J, Chitty L, Mackinson AM (2003). First and second trimester antenatal screening for Down's syndrome: the results of the Serum, Urine and Ultrasound Screening Study (SURUSS). J Med Screen 10, 56-104.

2. Wald NJ, Rodeck C, Hackshaw AK, Rudnicka AR (2004). SURUSS in perspective. Br J Obstet Gynaecol 111, 521-531

Why does Alpha not show separate risks for each fetus in a twin pregnancy?

In the case of a twin pregnancy where both biochemical measurements are available and NT measurements have been taken from each twin, Alpha calculates the risk estimate for the pregnancy not for each fetus (reference 1). This is because the biochemical measurements relate to the pregnancy and can not be allocated to one twin or the other and because if an amniocentesis were done both amniotic sacs would normally be tapped. If biochemical measurements are available it would not be correct to loose information and calculate a risk for each twin based on their NT measurements alone.

1. Wald NJ, Rish S (2005). Prenatal screening for Down syndrome and neural tube defects in twin pregnancies. Prenat Diagn 25, 740-745

Can I specify which set of normal medians to use?

Alpha will normally calculate the expected marker level based on the regression equation coefficients in place on the date of the screening report. There may be cases when it is not correct to use the coefficients in place on this date. For example, when the first trimester markers in an Integrated test were assayed and the medians subsequently updated before the screening report was made, then the medians in place on the date when the measurement was taken should be used.

For cases such as this, Alpha provides the assay date field in the data entry screen. The assay date field is next to the field where the marker level is entered on the Alpha data entry screen. When a date is entered into the assay date field, Alpha will use the regression equation coefficients in place on that date, otherwise it will use those current at the time of issuing the screening report.

What maternal age related risk equation does Alpha use?

The risk of an individual woman having a Down's syndrome pregnancy is given by (1,2,3) :

risk=1/((1+e^((7.330- 4.211/((1+ e^((-0.2815 × (age -37.73)))))))))

Where age is the woman's age in years at the expected date of delivery. The age is expressed as a decimal fraction, so for example 25 years 6 months is 25.5 years.

The following table shows the risk at term corresponding to the age at expected date of delivery.

 1. Morris JK, Mutton DE, Alberman E. (2002). Revised estimates of the maternal age specific live birth prevalence of Down's syndrome. J Med Screen 9, 2-6

2. Morris JK, Mutton DE, Alberman E (2005). Corrections to maternal age-specific live birth prevalence of Down's syndrome. J Med Screen 12, 202

3. Morris JK and Wald NJ (2007). Estimating the risk of Down's syndrome in antenatal screening and the gestation at which this risk applies. J Med Screen 2007 14, 5-7

Why does Alpha not use a second maternal weight with a second sample in the Integrated test?

For an Integrated test, the maternal weight recorded at either the first or second trimester test can be used because the difference in weights recorded at these times will be small and have no significant effect on the risk estimate. The weight gain between the first and second trimester test will be around 10 pounds (5 kg). This will have a very small effect on risk and does not justify the complication of two adjustments.

As an example, in an Integrated test for a 35 year old with MoM values of AFP = 1, uE3 = 1, total hCG = 1, inhibin-A = 1, NT = 1 and PAPP-A = 1, the risk estimate is 1 in 29,223. If we assume her weight was 120 pounds at the first trimester and 130 pounds at the second trimester, using a typical weight adjustment factor for PAPP-A the PAPP-A MoM value will change to 0.92 and the risk estimate to 1 in 26,787. The difference in risk estimate when accounting for the change in weight is around 1%.

What does it mean when a MoM value falls outside the 95% confidence interval around 1.0 MoM?

The 95% confidence interval is the range of values within which one can be 95% certain that the true value occurs. If the same trial were repeated 100 times, in 95 trials the confidence interval would include the true value, and in 5 trials it would not.

When a MoM value falls outside the 95% confidence interval around 1.0 MoM it may indicate that the current estimates of the gestation-specific medians are not accurate, and therefore need revising.

When a MoM value falls outside the 95% confidence interval, Alpha will show this on marker tabulations (using asterisks) or graphs of monthly medians (using a red circle instead of black).

How does Alpha make adjustments for in-vitro fertilization pregnancies?

This general question often relates to several specific questions:

1. Why are adjustment factors used and what are they?

Serum marker levels may differ, on average, between women who undergo in-vitro fertilisation (IVF) and those who conceive naturally.  If this is not taken into account, it can lead to differences in the false-positive rate between these two groups.  To avoid this, Alpha uses adjustment factors when interpreting reports for women who have undergone IVF.  The adjustment factors are the median marker MoM levels in IVF pregnancies (which are 0.94, 1.14 (reference 1) and 0.91 (reference 2) in ue3, total hCG and PAPP-A respectively).   Alpha divides the observed MoM values for an individual by the adjustment factors and these adjusted MoM values are used to calculate the risk estimate.

 2. When a woman receives a donated egg, what maternal age is used in the risk calculation?

When eggs are donated from a third party, the donor’s date of birth is entered into Alpha.  Alpha uses the donor’s age at the expected date of delivery to calculate the maternal age related risk of the pregnancy being affected with Down’s syndrome.

3. When a frozen embryo is implanted should the date of egg collection or embryo implanatation be used to calculate the maternal age?

When a frozen embryo is implanted the date on which the eggs were collected (which could be several years prior to implantation) is entered into Alpha rather than the date of implantation.    This date is used to determine the age of the mother or donor at the expected date of delivery and hence the maternal age related risk of the pregnancy being affected with Down’s syndrome.

 4. How is the date of embryo transfer used to estimate gestational age?

The date on which an embryo is transferred into the mother is used to estimate gestational age.  The gestational age is taken to be 14 days longer than the time since embryo transfer because the LMP date is about 14 days before the date of conception.   If the date of embryo transfer is recorded, this takes precedence over other estimates of gestational age.  

 1. Wald NJ, White N, Morris JK, Huttly WJ, Canick JA. (1999). Serum markers for Down's syndrome in women who have had in vitro fertilization: implications for antenatal screening. Br J Obstet Gynaecol 106, 1304-1306

2. Lambert-Messerlian G, Palomaki GE and Canick JA  (2009) Adjustment of serum markers in first trimester screening J Med Screen 16, 102-103

Why is the screening result for neural tube defects based on the AFP MoM value and not on the risk estimate?

With only one marker (AFP) “risk screening” is not needed so the AFP MoM value alone is used.  Also, the risk estimate for neural tube defects (NTDs) depends on accurate values for the local prevalences of anencephaly and of spina bifida.   These are often uncertain, for example because of differences in the use of periconceptional folic acid and because ultrasound scans may have identified anencephaly before serum testing excluding such cases from the population screened.    Therefore, using the risk estimate for NTD is less accurate than it is for Down’s syndrome. 

Why does Alpha not draw attention to markers which are individually high or low?

Alpha takes into consideration all the measured screening markers when calculating the Down’s syndrome risk estimate since this provides the most accurate determination of risk.   While it is known that very high or very low levels of the individual screening markers can be associated with adverse outcomes of pregnancy, such as miscarriage or still birth, the predictive screening performance is usually poor, and there is, in most cases no remedy.  It is accepted that it is poor medical and screening practice to look for disorders for which there is no remedy.

When should women who are diabetic be identified as such in Alpha?

Only patients who are receiving insulin treatment for diabetes should be entered as diabetic into Alpha.  Therefore diabetic women who are not insulin-dependent diabetics should not be entered as diabetic in Alpha.