Guidance

Software requirements for Down’s syndrome (T21), Edwards’ syndrome (T18) and Patau’s syndrome (T13) screening

Updated 2 December 2020

1. Scope

This document presents definitions and requirements for chance calculation software used for Down’s syndrome, Edwards’ syndrome and Patau’s syndrome screening within the NHS in England. The definitions and requirements are intended to provide a minimal specification for T21 and/or T18 and T13 chance calculation software within the NHS in England.

2. General principles

The main elements covered in this specification are:

• the methods used to compute multiple of median (MoM) values and chances
• the quality control and monitoring tools available for users of the software
• data and audit requirements

The methods specified for MoM values and chances involve algorithms specified by a set of parameters. These parameters must be configurable and auditable.

2.1 Requirement: configurability audit and validation

All parameters associated with the calculation of chance must be fully configurable under restricted access according to the requirements of the screening laboratory. Parameters defining the distribution of MoM values will be specified by the screening laboratory. Configuration should be possible by electronic import of a file.

2.2 Requirement: audit trail

All changes to the software parameters must be automatically recorded and auditable.

2.3 Requirement: validation

For the purpose of validation, according to laboratory standard operating procedures, the software must be capable of running test cases to facilitate checks on MoM values and chances.

2.4 Requirement: CE marking

The software must comply with CE statutory regulations.

3. Gestational age from ultrasound measurements

3.1 Requirement: calculation of gestational age from crown rump length (CRL)

For CRL in the range a to b inclusive, the software must be able to compute gestational age g in days as described below. If the CRL lies outside of the range a to b, the gestational age from this formula must not be used for chance calculation.

3.2 Singleton pregnancies

For singleton pregnancies, gestational age g in days at the time of the scan must be obtained from ultrasound measurements of CRL using the following formula:

3.3 Twin pregnancies

In twin pregnancies, the gestational age must be obtained from the larger of the 2 CRL measurements using the above formula.

3.4 Requirement: calculation of gestational age from head circumference (HC)

Singleton pregnancies

For HC in the range a to b inclusive, the software must be able to compute the HC gestational age g, in days as described below. If the HC lies outside of the range a to b inclusive, the gestational age from this formula must not be used for chance calculation.

3.5 Twin pregnancies

In multiple pregnancies, the gestation derived from the largest HC measurement must be used to determine gestational age.

3.6 Requirement: calculation of gestational age in days at the time of serum sampling

The software must be able to calculate gestational age at the time of serum sampling using the formula:

where g_serum​ is the gestational age at the time of serum sampling, g_scan​ is the gestational age at the time of the scan and days is the number of days from date₁​ to date₂​. If date₂​ occurs before date₁,​ this will be negative.

4. Maternal age m (years)

4.1 Requirement: calculation of maternal age

The software must be able to calculate maternal age at the expected date of delivery as described below.

4.2 Maternal age m (years) at the expected date of delivery (non-IVF pregnancies)

Where date of birth (DoB) is the maternal date of birth, use the equation:

​​

4.3 Maternal age m (years) at the expected date of delivery (IVF pregnancies)

where date_collection​ is the date of egg collection, DoB_donor​ is the date of birth of the egg donor and A is a configurable parameter.

For the calculation of prior chance, m should be used with no rounding or truncation.

5. Standardisation: MoM values

The median marker value for pregnancies with gestational age g (days), maternal weight w (kg), smoking status (s) and ethnicity (e), … is denoted by the function m(g,w,s,e,…)

For a marker value yy, in an individual pregnancy, the multiple of the median or MoM value is then given by:

​

where a is a configurable factor. In current software systems m(g,w,s,e, …) is specified by defining separate mathematical functions m(g) and m(w) for gestation and weight together with factor adjustments for smoking, ethnicity and other factors.

5.1 Requirement: median regression equations

The software must be capable of incorporating log-cubic regression relationships for both gestation and weight.

It should be possible to apply lower and upper truncation limits to gestation and weight in the above equations. These limits should be marker specific.

5.2 Requirement: adjustments other factors

The software must be capable of incorporating gestational age (by day) factor adjustments for:

• ethnicity – configurable levels
• smoking – configurable levels
• twins – monochorionic and dichorionic
• vanishing twins – factor and time since demise effect for those with a measurable CRL and factor for those with an empty pregnancy sac
• diabetes – configurable levels
• method of conception – configurable levels
• analyser – configurable levels

6. Localisation and updating

In some circumstances, it is necessary to make changes to accommodate differences between populations, laboratories, equipment or reagent lots. Functional forms can be updated by changing parameter estimates in m(g) and m(w). This can be achieved through the incorporation of regression fitting functionality within the software or externally by export of data and analysis using standard statistical software. The inclusion of regression fitting software is not a requirement for the NHS in England.

6.1 Factor updates

Often the changes required are simple factor updates. This means that all median levels for the particular marker are multiplied by the same constant factor, which can be achieved by defining a local factor.

​

In the reference configuration a = 1 and during an update a←a×f where f is an update factor. For example, if the overall median MoM value were 1.1, a would be multiplied by 1.1.

6.2 Requirement: factor updates

Software must be capable of making factor updates using a median MoM value entered manually by the laboratory. The acceptable range of median MoM values should be from a to b inclusive.

7. Prior chance models

7.1 Requirement: prior term chance of trisomies at maternal age m at expected date of delivery in decimal year

The software must be capable of calculating the prior chance of Down’s syndrome, Edwards’ syndrome or Patau’s syndrome using any of the following:

​

​

​

​

​

​

​

​

​

​

​

​

The coefficients A, B, C, D and E, above, are specific to each trisomy.

8. Likelihoods and chance calculation

Chances are obtained by combining the prior chance of trisomies with a multivariate Gaussian likelihood for the log MoM values. These multivariate Gaussian distributions are specified in terms of a set of marker means, a correlation matrix and a set of marker standard deviations for each condition (for example, does not have one of the conditions screened for, T21, T18 or T13). The software must be capable of calculating the chance of:

1. T21.

2. Joint chance for T18 and T13.

Depending on the individual’s choice, 1, 2 or both 1 and 2 should be reported.

8.1 Requirement: truncation limits

The software must be capable of applying condition-specific lower truncation limits a and upper truncation limits b to each marker MoM value to produce a truncated MoM value for use in chance calculation. Values below the lower truncation limit must be replaced by the lower truncation limit. Values above the upper truncation limit must be replaced by the upper truncation limit. In cases where MoM values are truncated, the original MoM value and the truncated value must be stored and it must be possible to display the original MoM value and truncated MoM value on the chance report.

8.2 Requirement: mean log₁₀ MoM values

It must be possible to specify mean log₁₀ MoM values in trisomies by gestational day for singleton pregnancies, for monochorionic twin pregnancies and for dichorionic twin pregnancies.

8.3 Requirement: fetus specific chances in twins

In situations where one or more markers (for example, NT) are measured on each twin, and the twins can be assumed to be dizygotic, the software must be able to produce fetus specific chances for T21 and for T18 and T13.

9. Recalculation of chances

9.1 Requirement: recalculation of chance

In cases where information is updated and the chance recalculated by the user, the most recent information and chance information should be displayed. All previous information should be kept in the database for the purposes of auditing.

10. Reporting chance information

10.1 Requirement: minimum and maximum chance

The software must be capable of imposing truncation limits on the reported chance, so that the minimum reported chance is less than 1 in a and the maximum reported chance is greater than 1 in b.

For example, less than 1 in 5,000; greater than 1 in 2.

The database should contain both the reported chance and the chance to full precision as calculated.

10.2 Requirement: application of chance cut-off

Chances close to the cut-off should be reported by rounding the reciprocal of the decimal chance upwards to the nearest integer value. For example, a decimal chance of 0.00663 that equates to 1 in 150.83 would be reported as 1 in 151. A decimal chance of 0.00667 that equates to 149.93 would be reported as 1 in 150. Chances greater than or equal to the chance cut-off should be reported as ‘at higher chance’. The classification should be consistent with the reported chances.

NB 1 in 150.1 becomes 1 in 151 and is not at higher chance.

11. Diagnostics

11.1 Requirement: MoM diagnostics

It must be possible for the user to select a subset of data based on any combination of date range, ethnicity, smoking status, singleton or twin pregnancy, analyser, reagent lot or sonographer.

The software must then be able to produce the following diagnostic plots:

• CUSUM plots of log₁₀ MoM values
• plots of the standardised screen positive rate, with 95% confidence intervals, by calendar month overall and separately for trisomy 21 and trisomies 18/13
• estimated median MoMs with 95% confidence intervals by:
  • gestational week maternal weight group
  • smoking status
  • ethnicity
  • diabetes status
  • analyser

11.2 Requirement: event log

The software must be able to maintain an event log to record events, such as lot change, analyser service, recalibration and median updates. The software must be capable of indicating these events on the CUSUM chart.

12. Export facility

12.1 Requirement: export

It should be possible for the user to specify a list of database fields and a date range to obtain a Microsoft Excel copy of the records over the date range containing the data in the specified fields. The fields must include all data used in the calculations of MoM values and chances, including all ultrasound measurements. Identification codes for sonographers should also be included.