Limitations of Direct Immunoassays for Measuring Circulating Estradiol Levels in Postmenopausal Women and Men in Epidemiologic Studies Free

This article is featured in Highlights of This Issue, p. 899

Estradiol (E₂) in serum serves as an important diagnostic marker in a variety of clinical conditions in both men and women. Clinically, serum levels of E₂ are used to assess ovarian function in women with menstrual disorders, precocious or delayed puberty, and assisted reproduction, as well as to monitor the effect of aromatase inhibitor treatment in breast cancer patients and determine postmenopausal status. In men, serum E₂ is used to assess gynecomastia. In epidemiologic studies, circulating levels of E₂ are often used to assess the etiologic role of estrogen in hormone-related conditions, including cancers of the breast, ovary, prostate, and liver. Accurate and reliable E₂ assays are essential for the validity of such studies. This is especially important when measuring E₂ levels in postmenopausal women or elderly men whose E₂ levels are low (<30 pg/mL). The optimum assay for such measurements is gas or liquid chromatography–tandem mass spectrometry (GC- or LC-MS/MS) or RIA with preceding extraction and chromatography steps (often referred to as conventional or indirect RIA). However, direct immunoassays, which include RIA, and chemiluminescent and enzyme immunoassays without a preceding purification step, are often used in epidemiologic studies to measure E₂ levels in postmenopausal women and elderly men, although in endocrinology literature it has been established that direct immunoassays are not valid tools for the measurement of low E₂ levels (1-11). The lack of sensitivity and specificity of direct immunoassays contributes to inconsistent results in the literature.

Although use of mass spectrometry assays for measurement of steroid hormones is growing rapidly, immunoassays will continue to be used widely in studies and clinical diagnostic testing. The original gold standard RIA used to quantify serum E₂ levels was first described by Abraham in 1969 (12, 13). This particular method requires an initial purification of E₂ before quantification by RIA. The purification is achieved by organic solvent extraction and Celite or Sephadex column chromatography. The organic solvent extraction step removes conjugated steroids, such as estrone sulfate, which is present in high concentrations in women and men, whereas the chromatographic procedure separates out most of the interfering E₂ metabolites (e.g., estrone), which could cross-react with the antibody in the assay. Use of a highly specific E₂ antibody in the RIA step further increases the specificity of the assay. The extraction/chromatographic E₂ RIA (conventional RIA) was shown to be sensitive, precise, specific, and accurate (12, 13).

The conventional RIA was applied to other steroid hormones, and the method has remained essentially the same for ∼40 years, except for use of an iodinated marker (¹²⁵I), instead of tritium, to provide greater assay sensitivity. Because the extraction and chromatography steps are cumbersome, time consuming, and costly, direct RIA methods without the purification steps were developed in the late 1970s to quantify steroid hormones directly. The ¹²⁵I-labeled marker used in RIAs was replaced with a chemiluminescent or enzymatic label to allow direct immunoassays to be automated, thereby substantially increasing the throughput of samples for steroid hormones.

Direct RIAs lack sensitivity in measuring serum levels of E₂ in postmenopausal women and men. Assay sensitivity is defined as the minimal detection limit of an assay. From the practical standpoint, it is the lowest concentration of a standard (on the standard curve) that can be distinguished from the zero standard and is based on the confidence limits of the measurements. A major deficiency is that manufacturers of kits used to measure steroid hormones seldom report how the sensitivity of their assay was determined. Most direct RIAs have a reported sensitivity ranging from 5 to 20 pg/mL, whereas E₂ levels in postmenopausal women and elderly men are usually <30 pg/mL and are often <5 pg/mL in very old women and men. Although the manufacturer of one direct E₂ RIA kit (Siemens Medical Solutions Diagnostics) indicates that their E₂ RIA has a sensitivity of 1.4 pg/mL, it should be noted that the lowest E₂ concentration used for the standard curve in the assay is 5 pg/mL. Measurements below the lowest point on a standard curve are not reliable and should not be accepted as valid data for statistical analysis.

Direct E₂ RIAs also lack specificity in measuring low E₂ levels. In the body, E₂ is converted to >100 conjugated and unconjugated metabolites. Some of the conjugated estrogens, such as estrone sulfate, have very high concentrations in the circulation and may cross-react with the antibody in the RIA, resulting in falsely high E₂ values. The lack of specificity also explains why direct assays are not valid for measuring serum E₂ levels in postmenopausal women receiving oral estrogen therapy, such as micronized E₂ or conjugated equine estrogens. Because the concentrations of the metabolites of these administered estrogens are very high, some of the metabolites will be recognized by the antibody, and if they are not removed by a purification step in the assay, this will result in gross overestimation of the E₂ levels.

Matrix differences between serum samples and the pure solutions of varying amounts of E₂ used to generate the standard curve in a direct E₂ RIA may also affect the validity of the results. This is especially true of hemolyzed and lipemic samples, in which there may be interference by RBCs and lipids, respectively, with the binding of the antigen (E₂) to the antibody.

To illustrate the point that direct RIAs are unreliable and invalid in measuring samples with low E₂ levels, we carried out a study comparing serum E₂ levels measured by direct RIA (Siemens Double Antibody Estradiol RIA kit) and our conventional RIA (10) in 38 female samples with low E₂ levels (<30 pg/mL) and 15 samples from postmenopausal women treated with estrogen. As shown in Table 1, 24 (63%) of the 38 E₂ measurements by direct RIA were below the detection limit (<5 pg/mL). This shows the importance of choosing a highly sensitive assay to measure very low E₂ levels for epidemiologic studies, such as a conventional RIA or GC- or LC-MS/MS, all of which have a sensitivity of ≤2 pg/mL (11).

Of the 14 (37%) samples with detectable E₂ levels by direct RIA (Table 1), the measurement by direct RIA in six samples was lower than that by conventional RIA, whereas eight were higher, suggesting a mixed pattern of cross-reactivity and misclassification. Thus, it should be noted that when assaying samples with low E₂ levels, sensitivity and reliability are not the only concerns. Specificity should be a concern because at low levels of E₂ cross-reactivity with E₂ metabolites may result in falsely high E₂ levels that are difficult to interpret. It is also unclear whether the extent of cross-reactivity is independent of the “true” E₂ concentration and whether there is differential cross-reactivity in cases and controls.

Table 2 shows E₂ values determined by the direct and conventional RIAs in serum samples obtained from the postmenopausal women treated with estrogen (micronized E₂ or conjugated equine estrogen). As shown, a large overestimation by the direct E₂ RIA is evident. For each sample, E₂ levels measured by direct RIA are much higher than those by conventional RIA, although the magnitude of the difference is not related to baseline levels. In some samples, the measured concentrations by direct RIA are 10 times higher than those by conventional RIA, suggesting large cross-reactivity of the assay antibody with the E₂ metabolites following treatment with micronized E₂, and with metabolites of E₂ and equine estrogens (equilin, equilenin, and their metabolites) following treatment with conjugated equine estrogens. The correlation coefficient between direct and conventional RIA measurements was −0.33 (P = 0.24; Fig. 1). The manufacturer of the direct E₂ RIA kit does recommend that an organic solvent extraction step be used before measuring E₂ by the direct RIA in samples obtained from women using estrogen therapy. However, the extraction step removes only conjugated estrogens; there are still >30 unconjugated estrogens that remain in the organic layer after extraction, some of which may cross-react with the antibody in the direct RIA. Therefore, a chromatographic step should be added to the purification procedure to remove the interfering estrogen metabolites before the RIA. For similar reasons, a preceding chromatographic step is also essential when measuring E₂ levels by RIA in serum samples obtained from women with breast cancer who are undergoing treatment by aromatase inhibitors. Certain aromatase inhibitors (e.g., fulvestrant and/or their metabolites) may cross-react with the antibody in the assay when the chemical structure of the drug resembles that of E₂.⁴

An important consideration in the validation of any steroid hormone RIA is assay accuracy. This is usually carried out by spiking stripped serum samples with a known amount of steroid and then measuring the steroid in the samples by RIA. One should expect the recovery of the steroid to be 90% to 110%. In our study, in stripped serum samples containing 15, 41, and 121 pg/mL of E₂ measured by our conventional RIA, the concentrations measured in the same samples by the Siemens direct E₂ RIA were <5, 10, and 50 pg/mL of E₂, respectively. The reason for the underestimated E₂ values measured by the direct RIA is not known but may be due to several factors, such as the presence of different populations of antibodies in the antiserum used for the RIA.

In conclusion, because of the poor sensitivity and specificity of direct immunoassays, especially at low levels of E₂, we recommend the use of conventional RIA or a GC- or LC-MS assay in epidemiologic studies measuring serum E₂ levels in postmenopausal women and men to improve the validity of the study. Use of inappropriate E₂ assays, such as direct RIA without purification, may compromise the validity of the epidemiologic research, leading to erroneous conclusions.

No potential conflicts of interest were disclosed.