Human Reproduction, Vol. 14, No. 5, 1175-1177,
May 1999
© 1999 European Society of Human Reproduction and Embryology
Precision and method bias of two assays for oestradiol: consequences for decisions in assisted reproduction
1 Department of Obstetrics and Gynaecology, The University of Western Ontario and 2 Gamma-Dynacare Medical Laboratories, London, Ontario, Canada
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Oestradiol monitoring in assisted reproduction is empirical, with no proof of benefit. Precise oestradiol estimation will be an essential pre-requisite for high quality evaluation of possible differences between combined and ultrasound-only monitoring. Objectives of the present trial were independent method comparison and bias estimation of chemiluminescent immunoassay (CLIA) versus radioimmunoassay for oestradiol. In a prospective comparison, 505 consecutive samples were split and assayed concurrently. Precision (reproducibility), relative bias and logistics were analysed and compared to manufacturers' findings. Correlation between CLIA and radioimmunoassay was excellent. Positive bias with CLIA necessitated altering decision points for therapy. Precision (reproducibility) was superior with CLIA, making it an appropriate candidate method for future randomized trials of the effectiveness of combined oestradiol/ultrasound monitoring for assisted reproduction.
Key words: bias/immunoassay/IVF/oestradiol/precision
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Oestradiol monitoring in assisted reproduction is empirical, with no proof of benefit (Golan et al., 1994
). Since information is inversely related to variance of measurement (M.N.Patel, University of Illinois, personal communication), advances in assay technology that increase precision (reproducibility) of oestradiol estimation have the potential to enhance clinical care. High quality evaluation of possible differences between ultrasound-only and combined oestradiol and ultrasound monitoring will depend on randomized design, precise estimation of oestradiol, and sufficient sample size for adequate power.
Oestradiol monitoring has a singular role in populations being treated by withholding gonadotrophin administration (Lee et al., 1998) in which the end-point of `coasting' is decreased serum oestradiol. It is unlikely that randomized trials of oestradiol monitoring will be performed in this population at risk for severe ovarian stimulation.
Advances in assay technology also offer logistic advantages of automation and the avoidance of radioisotopes. One alternative to radioimmunoassay is chemiluminescent immunoassay (CLIA), based on the principle of chemical production of light detected by a luminometer. However, differing assay technologies for oestradiol produce systematic disparities (Lee et al., 1991; Rojanasakul et al., 1994). The objectives of the present trial were independent method comparison and bias estimation of CLIA and radioimmunoassay for oestradiol (National Committee for Clinical Laboratory Standards, 1995) and the evaluation of the consequences of method bias on clinical decision making in assisted reproduction. Precision (reproducibility), relative bias and logistics for determining oestradiol using CLIA versus RIA were analysed and compared to manufacturers' findings.
|
Materials and methods |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Determinations of serum oestradiol using CLIA and radioimmunoassay were performed at Gamma-Dynacare Medical Laboratories from July 15 through September 30, 1996. Five hundred and five consecutive serum samples from patients undergoing assisted reproduction were split and analysed daily by both methods. Serum samples were obtained each day and measured concurrently. Technologists were blinded to results of the other method when performing analyses.
The CLIA was carried out using a commercially available assay, ACS Estradiol-6 (Chiron Diagnostics, Markham, Ontario, Canada) using the automated chemiluminescent immunoassay system ACSTM-180 (Chiron Diagnostics, Markham, Ontario, Canada), according to the manufacturer's protocol. The assay principle involves oestradiol in the patient sample competing with oestradiol labelled with dimethyl-acridinium ester (the light reagent) for a limited amount of rabbit anti-oestradiol antibodies. Rabbit anti-oestradiol antibodies are captured by mouse anti-rabbit IgG coupled to paramagnetic particles. There is an inverse relationship between the amount of oestradiol in the patient sample and the amount of relative light units detected. Samples in excess of 6000 pmol/l required manual dilution. All samples were analysed in duplicate.
The radioimmunoassay was also carried out using a commercially available assay, Coat-a-Count® Estradiol (Diagnostic Products Corporation, Los Angeles, CA, USA). This assay is a solid phase coated tube technology and a 60 min incubation was used in accordance with the manufacturer's protocol. Samples in excess of 10 000 pmol/l required manual dilution. All samples were analysed in duplicate.
Sample size and statistics
The sample size of 500 was chosen arbitrarily and this number was reached during the study period. Statistical analysis was performed using Winstar statistical software (Anderson-Bell Corp., Arvado CO, USA).
The mean of duplicates for CLIA and radioimmunoassay were compared using paired t-tests. Correlation was determined by linear regression and calculated overall and in ranges previously described (Loumaye et al., 1997). Imprecision (non-reproducibility) was expressed as coefficient of variation (CV) [(standard deviation/mean)x100] and compared using independent t-tests. Reliability of agreement was assessed by 
. Results are expressed as mean ± SD or mean ± 95% confidence intervals (CI). In analyses, P < 0.05 was considered significant.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Correlation
Overall correlation between CLIA and radioimmunoassay was excellent (r = 0.99). Figure 1
is a scatterplot of CLIA and RIA results. At lower concentrations, correlations declined (r = 0.87 in 100–500 pmol/l range and r = 0.45 in the 500–750 pmol/l range).
|
Precision
Precision was estimated as CV at three levels of quality control material (Lyphochek, Biorad Laboratories, Mississauga, Ontario, Canada). The CV was calculated for CLIA and radioimmunoassay at low, medium and high standards as shown in Table I
. When results were below detectable limits, the lowest value detectable was assigned as the value. Minimal detectable oestradiol was 36 pmol/l for CLIA and 32 pmol/l for radioimmunoassay. At all three levels, CV was greater with radioimmunoassay compared to CLIA.
|
Bias estimation and reliability of agreement after adjustment for bias
Results were higher with CLIA than with RIA (Figure 2
). In the present trial, the relationship between CLIA and radioimmunoassay was CLIA = 1.04 (radioimmunoassay) + 75 pmol/l, higher than determined by the manufacturer [CLIA = 1.03 (radioimmunoassay) + 19 pmol/l]. Positive bias with CLIA was 110 (92–132) pmol/l overall, 57 (47–67) pmol/l in the range 100–500 pmol/l and 110 (75–145) pmol/l in the range 500–750 pmol/l.
|
Due to positive bias the cut-point for ovarian suppression,
150 pmol/l using RIA (Yuzpe et al., 1994), was adjusted to 200 pmol/l for CLIA. The cut-point for increasing ovarian stimulation with gonadotrophins after 6 days, 250 pmol/l using radioimmunoassay, was adjusted to 300 pmol/l for CLIA.
Reliability of agreement was compared using the statistic for categorical variables of `suppressed' versus `not-suppressed' and `increase' versus `no increase'. Results of agreement by category are shown in Table II. Kappa indicated good agreement for `suppression' and strong agreement for `increase stimulation'.
|
Efficiency: time to perform assays
Time to perform a typical run of eight patient samples per day was 30 min using CLIA and 120 min using radioimmunoassay. This difference reflected the automated nature of CLIA and the reduced incubation time.
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Independent comparison of methods and estimation of bias is recommended to evaluate actual performance of new assays (National Committee for Clinical Laboratory Standards, 1995). Different immunoassay methods for steroids often yield highly variable results (Lee et al., 1991
; Boots et al., 1998).
While other investigators have compared CLIA to radioimmunoassay (Rojanasakul et al., 1994; Dancoine et al., 1997), the present trial evaluated the consequences of method bias on clinical decisions in assisted reproduction. In this prospective comparison, CLIA was superior in precision (reproducibility) to radioimmunoassay in all ranges. Correlation between CLIA and radioimmunoassay was excellent overall, r = 0.99 (Figure 1
). Results did not correlate so well in lower concentration ranges important for clinical decisions (Figure 1). Positive method bias with CLIA was higher in this independent comparison than that reported in the product monograph for CLIA. Positive bias with CLIA necessitated upward modification of clinical decision points. After adjustment for bias, reliability of agreement was `good' for ovarian suppression and `strong' for need to increase stimulation. Due to precision and efficiency superior to that obtained with radioimmunoassay, CLIA has been adopted as the standard oestradiol assay for assisted reproduction at Gamma-Dynacare Medical Laboratories.
In some centres ultrasound-only monitoring is standard care (Wikland et al., 1994; Tsirigotis et al., 1995; Hurst et al., 1997). Golan and co-workers performed a randomized comparison of ultrasound-only and combined oestradiol and ultrasound monitoring and concluded that both approaches were safe and effective (Golan et al., 1994). There were fewer than 60 subjects in each group and the no difference result may have been a type II error. If all cycles in the Golan trial had had cryopreservation and results for cryopreservation consistent for the entire study group, then the outcome with combined monitoring would have been superior (t = 2.12, P = 0.02).
The custom of oestradiol monitoring is based on tradition and the need for vigilance in preventing a serious iatrogenic syndrome: ovarian hyperstimulation syndrome. More evidence regarding oestradiol monitoring is needed. Precise oestradiol estimation will be an essential prerequisite for high quality evaluation of possible differences between combined and ultrasound-only monitoring. Chemiluminescent immunoassay for oestradiol is a candidate assay method for future randomized trials.
|
Acknowledgments |
|---|
The authors thank Gayle Waite, MLT, Gamma-Dynacare Laboratories, London, Ontario, Canada for performing the assays. Alice Deutsch, PhD, Bioscreen Inc., New York, NY, USA, Colin Howles, PhD, and Ernest Loumaye, MD, PhD, from Ares Services S.A., Geneva, Switzerland are thanked for editorial advice.
|
Notes |
|---|
3 To whom correspondence should be addressed at:339 Windermere Road, London, Ontario, Canada N6A 5A5
|
References |
|---|
|
TopAbstractIntroductionMaterials and methodsResultsDiscussionReferences |
|---|
Boots, L.R., Potter, S., Downing Potter, H. and Azziz, R. (1998) Measurement of total serum testosterone levels using commercially available kits: high degree of between kit variability. Fertil. Steril., 69, 286–292.[Web of Science][Medline]
Dancoine, F., Couplet, C., Buvat, J. et al. (1997) Analytical and clinical evaluation of the Immulite estradiol assay in serum from patients undergoing in vitro fertilization: estradiol increase in mature follicles. Clin. Chem., 43, 1165–1171.
Golan, A., Herman, A., Soffer, Y. et al. (1994) Ultrasonic control without hormone determination for ovulation induction in in-vitro fertilization/embryo transfer with gonadotrophin-releasing hormone analogue and human menopausal gonadotrophin. Hum. Reprod., 9, 1631–1633.
Hurst, B.S., Tucker, K.E., Awoniyi, C.A. and Schlaff, W.D. (1997) Preprogrammed, unmonitored ovarian stimulation reduces expense without compromising the outcome of assisted reproduction. Fertil. Steril., 68, 282–286.[Web of Science][Medline]
Lee, C., Tummon, I., Martin, J. et al. (1998) Does withholding gonadotrophin releasing administration prevent severe ovarian hyperstimulation syndrome? Hum. Reprod., 13, 1157–1158.
Lee, C.S., Smith, N.M. and Kahn, S.N. (1991) Analytic variability and clinical significance of different assays for serum estradiol. J. Reprod. Med., 36, 156–160.[Web of Science][Medline]
Loumaye, E., Engrand, P., Howles, C.M. and O'Dea, L. (1997) Assessment of the role of serum luteinizing hormone and estradiol response to follicle-stimulating hormone on in vitro fertilization treatment outcome. Fertil. Steril., 67, 889–899.[Web of Science][Medline]
National Committee for Clinical Laboratory Standards Guideline (1995) Method comparison and bias estimation using patient samples; approved guideline. NCCLS document EP9-A (ISBN 1-56238-283-7), NCCLS, 940 West Valley Road, Suite 1400, Wayne, PA 190087.
Rojanasakul, A., Udomsubpayakul, U. and Chinsomboon, S. (1994) Chemiluminescence immunoassay versus radioimmunoassay for the measurement of reproductive hormones. Int. J. Gynaecol. Obstet., 45, 141–146.[Medline]
Tsirigotis, M., Hutcheon, S., Yazdani, N. and Craft, I. (1995) The value of oestradiol estimations in controlled ovarian hyperstimulation cycles. Hum. Reprod., 10, 972–973.
Wikland, M., Borg, J., Hamberger, L. and Svalander, P. (1994) Simplification of IVF: minimal monitoring and the use of subcutaneous highly purified FSH administration for ovulation induction. Hum. Reprod., 9, 1430–1436.
Yuzpe, A.A., Nisker, J.A., Kaplan, B.R. et al. (1994) Nafarelin acetate for down-regulation in in vitro fertilization. J. Reprod. Med., 40, 83–88.
Submitted on July 28, 1998; accepted on December 4, 1998.
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  A. S. Dighe and P. M. Sluss Improved Detection of Serum Estradiol after Sample Extraction Procedure Clin. Chem., April 1, 2004; 50(4): 764 - 766. [Full Text] [PDF]
|
|
|
|
|
|
J. Taieb, C. Benattar, R. Diop, A. S. Birr, and A. Lindenbaum Use of the Architect-i2000 Estradiol Immunoassay during in Vitro Fertilization Clin. Chem., January 1, 2003; 49(1): 183 - 186. [Full Text] [PDF]
|
|
|
|
 |
|
 E. H. Yu Ng, W. S. B. Yeung, E. Yee Lan Lau, W. W. K. So, and P. C. Ho High serum oestradiol concentrations in fresh IVF cycles do not impair implantation and pregnancy rates in subsequent frozen-thawed embryo transfer cycles Hum. Reprod., February 1, 2000; 15(2): 250 - 255. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||