Analytical goal setting in aneuploidy screening: within person biological variability of first trimester biochemical markers

Transcription

1 DOI: /pd.4019 ORIGINAL ARTICLE Analytical goal setting in aneuploidy screening: within person biological variability of first trimester biochemical markers Kevin Spencer* and Nicholas J. Cowans Prenatal Research Unit, Department of Clinical Biochemistry, King George Hospital, Barley Lane, Goodmayes IG3 8YB, UK *Correspondence to: Kevin Spencer. ABSTRACT Objective To determine the average within person biological variability of free-b human chorionic gonadotrophin (free hcgb), intact hcg and pregnancy-associated plasma protein A (PAPP-A), and to establish analytical goals for the measurement of these markers when used in first trimester screening. Methods Free hcgb, PAPP-A and intact hcg were measured on paired first trimester samples collected during the same pregnancy. Results were converted to Multiple of the Median (MoMs).The overall total variation at each day log was determined from a correlation of the marker MoMs in the log domain. Biological variation was calculated after taking into account analytical variation. Results The within person biological variability for free hcgb varied from 1.30% at 2 days separation to 5.25% at 5 days. For PAPP-A this was 1.96% and 5.03%, respectively, and for intact hcg this was 14.59% and 21.09%. All markers exhibit a rapid increase in biological variability as the time separation increased. Conclusions Setting analytical goals for precision of measurement of first trimester biochemical markers from within person biological variability would suggest that free hcgb and PAPP-A needs to be measured with a precision of 2.5%, targets close to those set empirically by the Fetal Medicine Foundation and achieved in practice by some analytical system in routine use. Funding sources: None Conflicts of interest: Kevin Spencer is a consultant for Brahms AG. Department has research funding from PerkinElmer. INTRODUCTION Screening for trisomy 21 and other aneuploidies in the first trimester by means of the combined test achieves higher detection rates and lower false positive rates than second trimester alternatives. The combined test consists of the measurement of pregnancy-associated plasma protein A (PAPP-A) and free-b human chorionic gonadotrophin (free hcgb) in maternal serum and the measurement of fetal nuchal translucency (NT) by a trained ultrasonographer. The results of these are combined in a screening algorithm to produce a likelihood ratio, which is multiplied by the a priori maternal age related risk, resulting in an individual pregnancy specific risk. 1,2 This test has been demonstrated to achieve detection rates of 90% and above with false positive rates of 5%. 3 5 Wide sale implementation of such a screening program is now evident in many countries across the world. The biochemical variables PAPP-A and free hcgb are known to be influenced by a number of maternal or pregnancy variables such as gestational age, ethnicity, maternal weight and maternal smoking status, and correction for these variables are taken into account when calculating the Multiple of the Median (MoM) prior to use in the risk algorithm. 6 In clinical chemistry (biochemistry) the idea of analytical goal setting is not a new concept. Indeed the idea of setting analytical performance criteria based on biological variability dates back to the early 1970s from the work of Cotlove et al. 7 and developed by Stockl et al., 8 Holding 9 and Spencer 10 through the 1980s and 1990s. Although there are other methods that have been used to define analytical goals such as (1) state of the art, (2) views of experts, (3) based on quality control performance, (4) view of clinicians 8 it is generally considered that analytical goals derived from within person biological variability are the most appropriate to use in clinical decision making. 8 In this context the concept first proposed by Cotlove et al. 7 was that analytical variation (CV A ) should be less than one half the average within-subject biological variation (CV I ). In the context of screening for aneuploidy, very little is known about the within-subject biological variability of the various biochemical markers. Because the vast majority of markers change significantly with advancing gestational age, it would seem appropriate to look at within-subject biological variability after correcting for gestational age effects by converting the measured marker levels to MoMs. Although

2 Biological variability and analytical goal setting in aneuploidy screening 125 there is a growing body of literature that has established that there is a significant between pregnancy biological variability, 9 13 little is known about the within-subject biological variability in the same pregnancy. With respect to second trimester screening using either the triple marker or quad marker approach, it was shown that for some markers there was a high degree of correlation between a first sample and a repeat sample taken between 2 and 16 days later. 14,15 For first trimester markers however one small study 16 of 261 patients has also confirmed a high degree of correlation between a first sample and a repeat sample taken 14 days apart and the estimated average within person biological variability for free hcgb was 21% and 32% for PAPP-A. The aim of this current study was to extend the evaluation of within-person biological variability in pregnancy to a larger data set and to also include a measure of intact hcg, which in some jurisdictions is used as an alternative to free hcgb despite being widely known to result in poorer detection rates. 17 MATERIALS AND METHODS Study population In Barking Havering and Redbridge University Hospital NHS Trust, since June 1998 all pregnant women are offered first trimester (11 weeks 0 days to 13 weeks 6 days) screening for trisomy 21 and trisomy 13/18 by a combination of fetal NT thickness and maternal serum free hcgb and PAPP-A. 1 6 Free hcgb and PAPP-A are measured by a KRYPTOR Analyzer (Brahms AG, Hennigsdorf, Germany) a random access immunoassay analyzer using time-resolved amplified cryptate emission technology, which provides an accurate and highly reproducible result within 20 min. Fetal NT and crown rump length are measured by standardised techniques by sonographers who had completed the certificate of competence in the First Trimester Scan issued by the Fetal Medicine Foundation. Demographic characteristics, ultrasound findings and the results of biochemical testing are entered into a computer database (ViewPoint) at the time of assessment. Data on pregnancy outcome are obtained from the maternity unit and are also entered into the database. Women who subsequently on ultrasound examination were found to have a gestational age prior to 11 weeks 0 days were rebooked at the appropriate time and a further blood sample was taken at this visit. 18,19 All prenatal screening serum samples were stored at 20 C immediately after analysis and were available for further studies. We searched the ViewPoint database to identify pregnancies between 1999 and 2006 in which more than one sample had been collected and located these samples from our sample archive store. Analyte measurement Paired samples were reanalysed after thawing for serum free hcgb and PAPP-A using the routine Kryptor method along with three level internal quality control samples twice daily as for routine clinical practice. Serum intact hcg (BO82-101) was measured using the AutoDelfia (PerkinElmer, Turku, Finland) system and two level internal quality control samples were run at the beginning and end of every plate. Statistical analysis Pregnancy-associated plasma protein A and free hcgb concentration levels were converted to MoMs, corrected for gestational age, maternal weight, smoking status, ethnic origin, parity and assisted conception using methods outlined previously. 6 Intact hcg concentration levels were converted to MoMs corrected for gestational age based on a previously established gestational age curve. 20 Correction for other covariables as in the case of serum free hcgb and PAPP-A have not been established; the exclusion of this correction is unlikely to impact on the overall results because each pair of samples were from the same pregnant women in which these effects would be the same in each sample. The percentage between run analytical coefficient of variation (CV A ) was determined from the various control values by expressing the standard deviation of the results for each level as a percentage of the mean result and averaging this across each level for each of the markers. To establish total coefficient of variation (CV TOTAL ), we first grouped the data by the number of days between the two samples. Second, for each group we performed Pearson s correlation analysis for each paired marker MoMs in the log 10 domain and established the total variation from 1 r 2 as a percentage, where r is the Pearson correlation coefficient. 16 We subsequently established for each group the average within individual biological variation (CV I ) from the formula rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi h i CV TOTAL ¼ ðcv I Þ 2 þ ðcv A Þ 2 as described by Fraser and Harris. 21 Linear regression analysis of CV I against time between repeat sampling was performed to assess any trends. RESULTS Interrogation of the ViewPoint databases revealed 1326 patients in which a serum sample was available for analysis on two separate occasions in the first trimester of pregnancy. Table 1 summarises some of the clinical features of this study population. Table 2 shows the between run analytical performance at three concentration levels for free hcgb and PAPP-A, and at two Table 1 Characteristics of the study population Characteristic Total number of cases 1326 Median maternal age (years) [range] 30.1 [17 49] Median gestation sample 1 (days) [range] 72 [56 76] Median gestation sample 2 (days) [range] 89 [78 94] Median maternal weight (kg) [range] 67 [36 150] Proportion of nonsmokers (%) 73.6 Ethnicity (%) Afro-Caribbean 2.3 Asian 2.8 Caucasian 80.2 Oriental 1.4 Other/Unknown 13.2

3 126 K. Spencer and N. J. Cowans concentrations for intact hcg. The average CV A for free hcgb, PAPP-A and intact hcg was therefore 2.0%, 2.5% and 8.0%. Table 2 Between run analytical performance (CV A %) Marker Concentration CV % Number of runs Free hcgb IU/L IU/L IU/L PAPP-A 0.32 IU/L IU/L IU/L Intact hcg IU/L IU/L Table 3 summarises the Pearson correlation coefficient (r), total variation (CV TOTAL ) and the individual biological variation (CV I ) for the three biochemical markers investigated. The results show that free hcgb and PAPP-A have quite low biological variability with levels in the range of 1% to 5% when repeated within 5 days but that for intact hcg is considerably larger at 14% to 21%. The table also shows that the biological variation increases with time between the first and second sample. Figures 1 3 show that there is a significant relationship between the individual biological variation and time between repeat sampling and the data for each marker fitted a second-order polynomial regression with correlation coefficients very close to 0.9. DISCUSSION The results of this study have shown that the within person biological variability of the MoMs of the two main first Table 3 Total and individual biological variation for free b-hcg, PAPP-A and intact hcg (r = Pearson correlation coefficient) Free b-hcg PAPP-A Intact hcg Time interval (days) r CV Total % CV I % r CV Total % CV I % r CV Total % CV I % Number

4 Biological variability and analytical goal setting in aneuploidy screening 127 Figure 1 Relationship between days between sampling and average individual biological variation (CV I ) for free b-hcg MoM. Solid line is the second-order polynomial fit (y = x x 2 ; r = 0.891); dashed line is 95% CI Figure 3 Relationship between days between sampling and average individual biological variation (CV I ) for intact hcg MoM. Solid line is the second-order polynomial fit (y = x x 2 ; r = 0.883); dashed line is 95% CI Figure 2 Relationship between days between sampling and average individual biological variation (CV I ) for PAPP-A MoM. Solid line is the second-order polynomial fit (y = x x 2 ; r = 0.907); dashed line is 95% CI trimester markers free hcgb and PAPP-A is quite small and are considerably less than 5% when assessed over short time periods of 5 days or less. For intact hcg the within person biological variability is much greater than for free hcgb with values of around 20% when assessed over the same short time period, which might also add further to the argument that Intact hcg should not be used in first trimester screening. 17 A previous small study that looked at data from 261 pairs of samples collected on average 14 days apart measured CV I s of 21% and 32% for free hcgb and PAPP-A, respectively. 16 This was in close agreement with that found at around 14 days in the current study. The results of this study also show a surprising increase in CV I as the time period between the two blood tests increase, such that the CV I appears to follow a second-order polynomial increase within increasing time difference. It is difficult to understand from a biological perspective why this should be because gestational age effects should have been taken into account by converting the results to MoMs, although other underlying factors such as time of day in sampling cannot be excluded. The low CV I of less than 5% may have implications for screening if the concept of setting analytical goals based on within-person biological variation is to be utilised. Cotlove et al. 7,22 24,8 Stockl et al., 7,22 24,8 Fraser and Petersen, 7,22 24,8 Klee 7,22 24,8 and Fraser 7,22 24,8 have proposed that the analytical variation (CV A ) should be one half of the CV I, in this instance then the analytical variation for both free hcgb and PAPP-A should be less than 2.5%. In the analytical system used in this study (Brahms Kryptor) analytical variation of this order is attainable and can be demonstrated for both markers over a long period of time (in excess of 10 years). The impact of analytical variation on risk estimation in screening for Down syndrome has received limited study. Benn and Collins 25 showed that in screening in the second trimester imprecision in the measurement of analytes becomes amplified when risk is calculated, and put forward formulae to evaluate the effect of analytical variation. They found that when the assay CVs increased the confidence interval of the risk estimate widened considerably. Spencer 26,27 showed a similar effect when screening in the first trimester using free hcgb and PAPP-A as summarised in Figure 4. In practice the current state of the art for analytical performance is best summarised from performance in external quality assurance (EQA) schemes. The United Kingdom National External Quality Assurance Service (UKNEQAS) First Trimester Down Syndrome scheme has been operational since 2001 and covers not only 31 centres in the UK but includes over 400 European and worldwide centres. Overall performance in

5 128 K. Spencer and N. J. Cowans Figure 4 Influence of analytical precision on confidence interval of risk. Free b-hcg= 2.50 MoM; PAPP-A 0.55 MoM; nuchal translucency 2.2 mm; crown rump length 55 mm; age 24 years. (1) Marker CV 2% 2.5%, (2) marker CV 1% 1.5%, (3) marker CV 4% 5%, (4) marker CV 6% 7%. Adapted from Spencer 27 this scheme is summarised in the 2010 Annual Report. 28 Of the four major systems used for first trimester biochemical screening, the within method between laboratory CV% (which is a worst case scenario) is around 4% for both markers for the Brahms Kryptor and the Roche platforms, slightly higher at 6% for the PerkinElmer platforms and around 8% for the Siemens Immulite platforms. On the basis of our data on biological variability, analytical performance below 5% is required to minimize increased variability in the confidence interval of the risk. This fits with our previous simulations on the impact of different levels of imprecision performance on risk estimates. Expert groups such as the Fetal Medicine Foundation 29 have also advised that analytical performance between day in screening laboratories should be less than 3.5%. Clearly such advice fits with the now established biological variation of the first trimester biochemical markers and has a sound scientific basis. However, the overall error in risk estimation has not been thoroughly investigated in the context of screening for trisomy 21 and there are many components to the screening program, which in themselves may add significant error, some of these errors include analytical variability, sample stability/ degradation, gestational dating errors, assay bias, kit lot changes and between-person and within-person biological variability. Although analytical variability is a known and easily estimated quantity as is (for some analytes) assay bias for many of these other variables the impacts are not know or are difficult to quantify therefore defining a Clinically Allowable error 22,23 or overall performance criteria (Total Allowable Error) becomes more complex. Bias in measurement may well be a greater factor than analytical variability, because a bias of only 10% for each marker in the direction of low PAPP-A and high free hcgb will increase overall screen positive rates by over 50%. 30 Further work is required to understand why the biological variability for each marker increases with the time separation between samples. Although biological variability is small when samples are collected very close together, using half the biological variation as an analytical goal may be too stringent to set aside other factors that may impact on bias. Whether a suitable Total Allowable Error and analytical goal taking into account some of these other factors is possible requires further study. WHAT S ALREADY KNOWN ABOUT THIS TOPIC? The within person biological variation of the first trimester biochemical markers free hcgb and PAPP-A measured 14 days apart was 21% and 32% in a previous study. WHAT DOES THIS STUDY ADD? For samples taken 5 days or less apart, the within-person biological variation is <5% for PAPP-A and free hcgb, but 20% for intact hcg. Analytical goals of less than 2.5% CV are required when measuring free hcgb and PAPP-A. REFERENCES 1. Wright D, Spencer K, Kagan KO, et al. First trimester combined screening for trisomy 21 at 7 14 weeks gestation. Ultrasound Obstet Gynecol 2010;36: Wright D, Kagan KO, Molina FS, et al. A mixture model of nuchal translucency thickness in screening for chromosomal defects. Ultrasound Obstet Gynecol 2008;31: Spencer K, Souter V, Tul N, et al. A screening program for Trisomy 21at weeks using fetal nuchal translucency, maternal serum free beta hcg and PAPP-A. Ultrasound Obstet Gynaecol 1999:13; Nicolaides KH, Spencer K, Avgidou K, et al. Multicenter study of first trimester screening for trisomy 21 in pregnancies: results and estimation of the potential impact of individual risk-orientated two-stage first-trimester screening. Ultrasound Obstet Gynecol 2005;25: Kagan KO, Wright D, Baker A, et al. Screening for trisomy 21 by maternal age, fetal nuchal translucency thickness, free beta-human chorionic gonadotropin and pregnancy-associated plasma protein-a. Ultrasound Obstet Gynecol 2008;31: Kagan KO, Wright D, Spencer K, et al. First-trimester screening for trisomy21byfreebeta-humanchorionicgonadotropinand pregnancy-associated plasma protein-a: impact of maternal and pregnancy characteristics. Ultrasound Obstet Gynecol 2008;31: Cotlove E, Harris EK, William GZ. Components of variation in long term studies of serum constituents in normal subjects. III. Physiological and medical implications. Clin Chem 1970;16: Stockl D, Baadenhuijsen H, Fraser CG, et al. Desirable routine analytical goals for quantities assayed in serum. Eur J Clin Chem Clin Biochem 1995;33: Holding S, Cuckle H. Maternal serum screening for Down s syndrome taking account of the results of a previous pregnancy. Prenat Diagn 1994;14: Spencer K. Between pregnancy biological variability of maternal serum alpha fetoprotein and free beta hcg: implications for Down syndrome screening in subsequent pregnancies. Prenat Diagn 1997;17: Spencer K. Between pregnancy biological variability of first trimester markers of Down syndrome: implications for screening in subsequent pregnancies. Prenat Diagn 2001;21:445 7.

6 Biological variability and analytical goal setting in aneuploidy screening Spencer K. Between pregnancy biological variability of first trimester markers of Down syndrome and the implications for screening in subsequent pregnancies: an issue revisited. Prenat Diagn 2002;22: Wright D, Syngelaki A, Birdir C, et al. First-Trimester Screening for Trisomy 21 with Adjustment for Biochemical Results of Previous Pregnancies. Fetal Diagn Ther 2011;30: Cuckle HS, Densem J, Wald NJ. Repeat maternal serum testing in multiple marker Down s syndrome screening programmes. Prenat Diagn 1994;14: Hackshaw AK, Densem J, Wald NJ. Repeat maternal serum testing for Down s syndrome screening using multiple markers with special reference to free a and free b-hcg. Prenat Diagn 1995;15: Spencer K, Cuckle HS. Screening for chromosomal anomalies in the first trimester: does repeat maternal serum screening improve detection rates? Prenat Diagn 2002;22: Evans MI, Hallahan TW, Krantz D, Galen RS. Meta-analysis of first trimester Down syndrome screening studies: free beta-human chorionic gonadotropin significantly outperforms intact human chorionic gonadotropin in a multimarker protocol. Am J Obstet Gynecol 2007;196: Spencer K, Spencer CE, Power M, et al. One stop clinic for assessment of risk for fetal anomalies: a report of the first year of prospective screening for chromosomal anomalies in the first trimester. BJOG 2000;107: Spencer K, Spencer CE, Power M, et al. Screening for chromosomal anomalies in the first trimester using ultrasound and maternal serum biochemistry in a one-stop clinic: a review of three years prospective experience. BJOG 2003;110: Spencer K, Cowans NJ, Uldberg N, et al. First trimester intact hcg as an early marker of trisomy 21: a promise unrecognised? Prenat Diagn 2008;28: Fraser CG, Harris EK. Generation and application of data on biological variation in clinical chemistry. Crit Rev Clin Lab Sci 1989;27: Fraser CG, Petersen PH. Analytical performance characteristics should be judged against objective quality specifications. Clin Chem 1999;45: Klee GG. Establishment of outcome-related analytic performance goals. Clin Chem 2010;56: Fraser CG. Biological Variation: From Principles to Practice. Washington, DC: AACC Press; Benn PA, Collins R. Evaluation of effect of analytical imprecision in maternal serum screening for Down s syndrome. Ann Clin Biochem 2001;38: Spencer K. Evaluation of effect of analytical imprecision in maternal serum screening for Down s syndrome. Ann Clin Biochem 2001;38: Spencer K. Risk, a QC parameter. Down s Screening News 2003;10: s%201st%20t% %20[SAG]%20Edited.pdf Biochemical%20Laboratories_new.pdf 30. Wright D, Abele H, Baker A, Kagan KO. Impact of bias in serum free beta-human chorionic gonadotropin and pregnancy-associated plasma protein-a multiples of the median levels on first-trimester screening for trisomy 21. Ultrasound Obstet Gynecol 2011;38: