Falsely elevated cortisol serum levels in preterm infants due to use of immunoassay

To the Editor,

In the first weeks of life, many preterm infants fail to mount an appropriate cortisol response for the degree of illness or stress, known as relative adrenal insufficiency (RAI) [1]. This may manifest as refractory hypotension, defined as hypotension that is unresponsive to volume replacement and inotropes, but responds readily to systemic corticosteroids [1]. RAI may predispose to a pro-inflammatory state [2], but it is unknown at which cortisol level RAI is likely to manifest [3]. Studies in preterm infants with refractory hypotension have shown that serum cortisol level was unable to predict hydrocortisone treatment response [4]. Therefore, clinical practice guidelines do not recommend determination of cortisol before initiating corticosteroid treatment in hypotensive infants.

Low activity of immature adrenal cortex enzymes, mainly 11β-hydroxylase, is considered one of the mechanisms behind RAI [5]. This leads not only to a low cortisol by also to a relative abundance of the precursor corticosteroids 17-hydroxyprogesterone (17-OHP) and 11-deoxycortisol. Furthermore, in preterm infants the cortisol-cortisone shuttlefavors cortisone, reflecting the developmental timing of 11β-hydroxysteroid steroid isoenzyme expression similar to what is found in fetal tissues [1, 6]. Cortisol is often measured using immunoassays, which are prone to cross-reactivity. Studies have demonstrated that in some conditions cortisol levels were overestimated by immunoassays [7, 8]. In patients with congenital adrenal hyperplasia (CAH) due to 21alpha-hydroxylase deficiency, this was caused by cross-reactivity with elevated precursor corticosteroids [7]. Such cross-reactivity does not occur with liquid chromatography-tandem mass spectrometry (LC-MS/MS), which is highly specific and is therefore considered superior for steroid hormone analysis [7].

We hypothesized that immunoassays may overestimate cortisol levels in preterm infants, considering that their precursor corticosteroid levels may be elevated like in patients with CAH. To our knowledge, evidence of cross-reactivity in cortisol measurement has not been previously investigated in preterm populations. The aim of this study was to assess whether two widely employed immunoassays, i.e., the Cobas assay and the Alinity assay, are able to determine cortisol levels as accurately as LC-MS/MS in preterm infants.

We collected serum samples for measurement of cortisol and other corticosteroids in preterm infants born <30 weeks gestational age, who participated in the PRIDICT-BPD (PulmonaRy Inflammation and glucocorticoiD sensitivity for the prediCTion of BronchoPulmonary Dysplasia) study. This feasibility study aimed to test the role of various biomarkers for the prediction of BPD. The serum samples were collected directly after birth from cord blood, and at days 3, 7, 14 and 28 from capillary or arterial blood. For each time point four pools of samples were formed, each of which had an equal amount of leftover serum samples from six infants, resulting in a total number of 20 pools. Serum samples from infants who received postnatal corticosteroids were excluded for this specific study. The PRIDICT-BPD study was approved by the Medical Research Ethics Committee of the Vrije Universiteit Amsterdam (protocol number 2019.371).

Cortisol concentrations were measured using two immunoassays, i.e., the Cobas Cortisol II Assay (Roche Diagnostics, Rotkreuz, Switzerland), and the Alinity Cortisol Assay (Abbott Diagnostics, Chicago, IL, USA), and with our in-house developed LC-MS/MS method (as described earlier [9]). Our LC-MS/MS method has a lower limit of quantitation of 0.5 nmol/L, an intra-assay coefficient of variation (CV) of <5 %, and an inter-assay CV of 4.5 and 3.9 % at cortisol levels of 125 and 650 nmol/L, respectively. This method is well comparable to other LC-MS/MS methods [9]. All analyses were performed at the Endocrine Laboratory of the Amsterdam UMC (ISO15189 accredited).

We calculated the absolute difference in measured cortisol level of each immunoassay relative to the LC-MS/MS method. Agreement across the measurement range was tested using Passing and Bablok regression analysis, and correlation was tested using Pearson correlation.

Figure 1 shows the absolute difference in measured cortisol levels of each immunoassay relative to the LC-MS/MS method for each pool separately. For the Alinity assay, as compared to the LC-MS/MS method, the absolute difference in cortisol level was <50 nmol/L at all time points. For the Cobas assay, as compared with the LC-MS/MS method, the absolute difference in cortisol level was largest at day 3, with a maximum difference of 131 nmol/L (Cobas 394 vs. LC-MS/MS 263 nmol/L), and at day 7, with a maximum difference of 155 nmol/L (Cobas 491 vs. LC-MS/MS 336 nmol/L).

Figure 1:

Absolute difference in cortisol level (nmol/L) of each immunoassay relative to the LC-MS/MS method by postnatal age in pooled samples of preterm infants. Cortisol levels were measured with two immunoassays (Alinity and Cobas) and with the LC-MS/MS method in preterm infants during the first four weeks of life. The zero line indicates no difference in cortisol levels between assay methods.

Passing and Bablok regression analysis showed a slope of 1.01 (95 % CI 0.95–1.08) and an intercept of 0.79 (95 % CI −9.1 to 15) nmol/L for the Alinity assay, indicating good agreement with the LC-MS/MS method (Figure 2A). For the Cobas assay, analysis showed poor agreement with the LC-MS/MS method, as shown by a slope of 1.39 (95 % CI 1.21–1.55), and an intercept of −3.13 (95 % CI −26.5 to 17.6) (Figure 2B). Pearson correlation coefficients were 0.984 and 0.947 for the Alinity assay and the Cobas assay, respectively.

Figure 2:

Passing and Bablok regression analyses for serum cortisol measurements in pooled samples of preterm infants. On the x-axis, the cortisol concentrations measured using the LC-MS/MS method and on the y-axis the cortisol concentrations measured using the respective immunoassays are shown. (A) Alinity assay; (B) Cobas assay.

At last, Table 1 shows, in the same pooled samples, the results of also by LC-MS/MS measured levels of corticosteroids that could possibly lead to cross-reactivity. Concentrations varied between 143 and 188 nmol/L for cortisone, between 0.4 and 2.4 nmol/L for 11-deoxycortisol and between 8 and 41 nmol/L for corticosterone.

Our study shows that the Cobas assay, unlike the Alinity assay, severely overestimates cortisol levels in preterm infants. As such, it confirms previous observations demonstrating that in conditions characterized by elevated levels of precursor corticosteroids, immunoassays may be prone to cross-reactivity with hormones that share the general structure of the hormone of interest [8]. An implication of our findings is that previous studies among preterm infants addressing associations between serum cortisol levels and clinical outcomes should be interpreted cautiously, because of the possible risk of overestimated cortisol levels.

In healthy adults without medication use, cross-reactivity is unlikely to play a major role in immunoassays for cortisol. This was corroborated by analyses at our laboratory, demonstrating good agreement of both immunoassays with our LC/MS-MS method, with slopes of 0.94 (95 % CI 0.88–0.98) and 0.96 (95 % CI 0.93–0.98), intercepts of −5.9 (95 % CI −16 to 8.1) and 1.3 (95 % CI −4.9 to 8.1) and correlation coefficients of 0.986 and 0.996 for the Alinity assay and the Cobas assay, respectively [10].

The adrenocortical function of preterm infants is, similar to CAH, characterized by an abundance of corticosteroid precursors relative to cortisol, and by high cortisone levels. Although this biochemical pattern likely offers an explanation for overestimated cortisol levels by use of immunoassays, it is unclear why the degree of cross-reactivity differs between immunoassays, in spite of absence of evidence for substantial cross-reactivity according to both kit inserts. According to the kit inserts, cross-reactivity was estimated at 4.9 and 1.9 % for 11-deoxycortisol, at 0.08 and 0.6 % for 17-OHP, at 6.6 and 2.7 % for cortisone, and at 2.5 and 0.9 % for corticosterone in the Cobas assay (after addition of 10 μg/mL and the Alinity assay (after addition of 100 μg/dL 11-deoxycortisol and 1,000 μg/dL 17-OHP and cortisone), respectively. Levels of cortisone, 11-deoxycortisol and corticosterone measured by LC-MS/MS, only marginally contributed to cross-reactivity in cortisol measurement (see Table 1). However, other steroid hormones could contribute to cross-reactivity too. Another mechanism which could lead to falsely elevated results in the cortisol immunoassay are differences in the serum protein composition between preterm infants and term infants or adults. Among these differences is the concentration of cortisol-binding globulin (CBG), which is lower in preterm infants [11, 12]. Low CBG levels were previously shown to overestimate cortisol levels in certain immunoassays [13]. Nonetheless, it should be acknowledged that reported cross-reactivity may vary between patient populations, and, therefore, we would like to call upon manufacturers to improve their cortisol immunoassays for patient populations with a different composition of precursor corticosteroids and/or corticosteroid metabolites than healthy adults.

One of the limitations of this study is that due to the relatively large serum volume required for immunoassays in relation to the limited volume available, we had to pool samples from multiple infants, and we were unable to test other cortisol immunoassays being used in clinical practice. However, it can be expected that more cortisol immunoassays are susceptible to overestimation in preterm infants, encouraging further testing.

In conclusion, some immunoassays may overestimate serum cortisol levels in preterm infants in their first weeks of life, necessitating the use of a LC-MS/MS method or an immunoassay that has proved to be accurate in preterm infants.