Abstract
Interference in immunoassays is a widely recognized problem, which could potentially lead to unnecessary investigations and treatment. We describe a case where interference in a cortisol immunoassay led to a falsely low serum cortisol concentration and interference in the free thyroxine assay led to falsely elevated serum thyroxine concentrations, in the same patient. A 42-year-old woman with documented hypothyroidism underwent a synacthen test for suspected adrenal insufficiency. Previous thyroid function tests had been discordant and difficult to interpret, with elevated thyroxine and non-suppressed thyroid-stimulating hormone. The synacthen test showed a subnormal cortisol response and she was commenced on cortisol replacement. Clinically, it was doubted whether she had true adrenal insufficiency and it was thought that the cortisol results might be artefactually low due to assay interference. Cortisol was measured by an alternative immunoassay, before and after incubation in an antibody blocking tube (‘Scantibodies’), after heat treatment and also after treatment with Protein A. The results supported assay interference and cortisol ‘replacement’ was stopped. Thyroxine had been discontinued although thyroid function tests (TFTs) were significantly different between analytical platforms, also consistent with interference. Thyroxine replacement was restarted and once on a stable dose, the discrepancy in TFTs was also investigated by similar procedures as for cortisol. Good clinician–laboratory interface and laboratory work-up of patients with results that are discordant from the clinical findings can reduce unnecessary investigation and inappropriate treatment.
Introduction
Interference in immunoassay is a widely recognized problem, which could potentially lead to unnecessary, expensive and harmful investigations and treatment. 1–5 We describe a case where interference in a cortisol immunoassay led to a falsely low serum cortisol concentration and interference in the free thyroxine assay led to falsely elevated serum thyroxine concentrations, in the same patient. This case highlights the complexity of interpretation of test results and the importance of suspecting assay interference when laboratory results are discordant with clinical findings.
Case report
A 42-year-old psychiatric nurse on thyroxine treatment for previously diagnosed hypothyroidism was referred to the endocrinology clinic with ‘difficult to interpret’ thyroid function tests (TFTs) for further management. Four months prior to review her replacement thyroxine was discontinued, based on TFTs which showed normal FT3, high FT4 and non-suppressed thyroid-stimulating hormone (TSH). She had positive anti-thyroid peroxidase but normal anti-thyroglobulin antibodies. Ongoing blood tests since that time had all been abnormal with high FT4 and non-suppressed TSH. Her past medical history was otherwise unremarkable and she was not on any other medications. She was aware that, at the time of diagnosis of hypothyroidism, in Scotland 8 y prior, her samples for TFTs were referred to a reference laboratory because they were ‘difficult to interpret’. There was a very strong family history of thyroid disease with her mother and two sisters having been diagnosed with hypothyroidism. On examination she was clinically euthyroid with no evidence of thyroid eye disease. She was clinically euvolaemic with blood pressure being 120/70 mmHg and no postural drop.
Repeat TFTs (Roche Cobas 6000) showed elevated TSH of 69 mIU/L (reference range 0.4–4 mIU/L) with free T4 of 17 pmol/L (reference range 11–24 pmol/L) and free T3 of 2.4 pmol/L (reference range 2.5–5.5 pmol/L).
Random cortisol (Roche Elecsys 1010 analyser, Roche Diagnostics, New Zealand) was 47 nmol/L at 1400 h which was unexpectedly abnormally low. Follow-up synacthen test showed pre-dose cortisol 48 nmol/L increasing to 198 nmol/L, 30 min after 250 μg i.m. synacthen (normal response >550 nmol/L) consistent with adrenal insufficiency.
One month later, the patient felt unwell with vague symptoms including being fatigued, irritable and constipated. She also had difficulty sleeping, muscle pains across her chest and legs and had noticed her skin to be slightly browner than usual. Examination revealed regular pulse, no postural drop in blood pressure with stable weight. She had clinical signs of hypothyroidism with slow relaxing reflexes and increased puffiness of her face. She also seemed slightly hyperpigmented. At this time, she was started on hydrocortisone 10 mg daily, fludrocortisone 50 μg daily and was instructed to start thyroxine 50 μg daily after 3 d.
However, the electrolytes were normal with Na 140 mmol/L, K 3.9 mmol/L, creatinine 80 μmol/L, glucose 4.5 mmol/L and C-reactive protein <3 mg/L. Liver function tests and serum calcium were normal. Plasma renin activity was 0.7 nmol/L/h (reference range 0.4–2.3 nmol/L/h) with plasma adrenocorticotrophic hormone (ACTH) 0.3 pmol/L (reference range 1.0–12.0 pmol/L) both before steroid replacement, thus not supportive of primary adreno-cortical insufficiency. Other tests including FSH, LH, PRL, IGF1, PTH, testosterone, DHEAS were all normal. Adrenal antibodies were not detected.
The clinical features were thus not typical of adrenal insufficiency and the other biochemical results were also not supportive. The TFTs were also not consistent with the clinical findings of hypothyroidism. Given the uncertainties with the diagnosis, the possibility of an interference with the cortisol assay as well as the thyroxine assay was considered and confirmed. Steroid supplements were discontinued after a week's replacement and thyroxine replacement was recommenced.
Methods and results
The original serum cortisol result in the synacthen stimulation tests which showed subnormal response was measured by the Roche Elecsys 1010 analyser. To evaluate the possibility of interference in this cortisol assay we split the samples from a further ACTH stimulation test and assayed cortisol by the Roche Elecsys 1010 analyser as well as an ‘in-house’ enzyme-linked immunosorbent assay (ELISA). 6 Furthermore, we assayed these samples on the Roche Elecsys 1010 following incubation of the serum in heterophile blocking tubes (Scantibodies Laboratory, Inc, Sanlee, CA, USA) and following extraction with dichloromethane. The post-synacthen sample was subsequently assayed on the Roche Elecsys 1010 following heat treatment at 60°C for 60 min and protein A treatment. For extraction of cortisol from serum, 200 μL was extracted with 2 mL dichloromethane and 1.5 mL of the extract was dried and reconstituted in 150 μL of Roche low-cortisol calibrant and then assayed using the Roche analyser thereby mitigating matrix effects. For protein A treatment, 0.5 mL of Protein A Ceramic Hyper DF (Biosepra, Life Technologies) was washed twice each with 2 mL of phosphate-buffered saline, centrifuged, the supernatant removed after which it was incubated with 0.5 mL plasma for 60 min at room temperature. Tubes were then centrifuged and the plasma supernatant removed for cortisol analysis, by in-house ELISA and also Roche Elecsys 1010 analyser (Roche Diagnostics, New Zealand).
The initial TFTs were performed on Roche Cobas 6000 analyser (Roche Diagnostics, New Zealand). To evaluate the possibility of interference in the TFT assays, we re-assayed a split sample on an Abbott Architect ci8200 analyser (Abbott Laboratories, Abbot Park, Illinois, USA). Further samples were collected after the patient had been on a stable dose of thyroxine for a few months and the same experiments as for cortisol were carried out with the exception of extraction with dichloromethane and heat treatment.
Rheumatoid factor was also measured and found to be normal. Table 1 shows the synacthen cortisol results (upper panel) and the cortisol results on another single pre-synacthen sample following various treatments (lower panel). The thyroid function results while the patient was on a stable dose of thyroxine are shown in Table 2.
Upper panel: adrenocorticotrophic hormone stimulation test using 250 μg i.m. Synacthen® for various assay methods: normal response 30 min following synacthen = serum cortisol > 550 nmol/L). Lower panel: serum cortisol and IgG concentrations pre- and post-treatment with Protein A or heat on Roche Elecsys 1010
ELISA, enzyme-linked immunosorbent assay
Thyroid function tests and IgG concentrations pre- and post-treatment with Protein A and incubation with Scantibody tube while on a stable dose of thyroxine
nd, not detected; TSH, thyroid-stimulating hormone
Discussion
Heterophile antibodies are natural antibodies or autoantibodies, which are polyspecific and usually show low affinity, weak binding to a heterogeneous group of antigens in a non-competitive manner. 1,2 Interference by heterophile antibodies may lead to falsely elevated or falsely lowered analyte concentrations in one or more assay systems depending on the site of the interference in the reaction. 2 Although falsely elevated values are more commonly described, there are few reported cases of falsely low values due to heterophile antibody interference. 1,2
The prevalence of interference by heterophile antibodies in immunoassays varies with assay type, analyte and method of detection. 1,2 Approximately 40% of serum samples contain non-analyte substances that bind to the assay antibodies and most manufacturers minimize their impact by incorporating blocking agents into the reagents. Despite this, heterophile antibody interference continues to be a clinical problem with negative consequences of unrecognized heterophile antibody interference. 3–5 Heterophile antibodies can also be transient making interpretation of results even more difficult. 7
In our patient, cortisol concentrations were falsely low and the free thyroxine concentrations were disproportionately elevated leading to confusion in the interpretation of TFTs as well as ACTH stimulation tests. Incubation of the patient serum with protein A removed 95% of the IgG, yet the cortisol concentration was not corrected to the expected 450–500 nmmol/L. Incubation of the patient sample in anti-sheep serum also did not correct the cortisol concentration to the expected value. However, heat treatment corrected the cortisol more closely to the expected value, suggesting either interference by non-IgG antibody or multiple interferences.
For the thyroxine assays, incubation with Scantibody tube did not correct it, although an increase in FT3 was noted, which could possibly be due to a matrix effect. On the other hand, incubation with protein A completely corrected both FT3 and FT4 assays suggesting IgG interference in these assays. This is in contrast with the cortisol assay where incubation with Scantibody partially corrected it and incubation with Protein A had little effect. This suggests more than one interferent in the patient sample and different interferents affecting the two assays.
The Roche Elecsys 1010 chemiluminescence plasma cortisol assay is a competitive immunoassay, which uses a biotinylated antibody and a ruthenium complex-labelled cortisol derivative as a competitor for endogenous cortisol. In the second step, streptavidin-coated microparticles are used as a solid phase to bind the biotinylated antibody. It is likely that the interferent had bound endogenous cortisol leaving more of the biotinylated antibody to bind the labelled analogue complex giving a higher chemiluminescence signal, thus falsely low cortisol concentrations. As protein A treatment was ineffective and heat treatment ablated the interference, it is unlikely that the interference is caused by IgG, unless there is circulating auto-antibody to cortisol. In this remote possibility, Protein A treatment would remove the antibody–cortisol complex thus depleting cortisol in the sample. Both IgM and corticosteroid-binding globulin (CBG) are heat-sensitive and possible candidates, although rheumatoid factor was undetectable leaving the possibility of interference by CBG. A CBG variant with high affinity for cortisol could conceivably compete with the polyclonal antibody for cortisol thereby resulting in falsely low concentrations. Variants of CBG with altered affinity for cortisol have been described previously 8 and this possibility could be investigated using Scatchard analysis 9 and genotyping if warranted. Interference due to autocortisol antibody remains a possibility although incubation with Scantibody did not completely correct the result. Interestingly, the in-house ELISA, which has a different assay configuration, was not subject to interference. The ELISA uses an immobilized cortisol-thyroglobulin conjugate where cortisol in the standards or sample competes for the mouse monoclonal antibody binding sites. Following a wash step, immobilized cortisol-antibody is then detected using anti-mouse peroxidase antibody. An alternative postulate is that the Ka of the ELISA cortisol monoclonal antibody is greater than the Ka of the postulated variant CBG.
The discrepancy in interference between the Roche Cobas 6000 and the Abbott Architect ci8200 FT4 assays may be explained by the difference in the configuration of the assays; both using chemiluminescence measurements. The Abbott Architect ci8200 chemiluminescence immunoassay for FT4 uses a two-step design in which T4-specific sheep antibody-coated microparticles are used to capture endogenous T4. Unbound material is then removed by a wash step prior to the addition of analogue T4 (T3 acridinium-labelled conjugate), thereby precluding interaction between sample antibodies (interferent) and hormone analogue. The Roche Cobas 6000 adds the analogue (biotinylated T4) and the capture antibody (T4-specific polyclonal sheep antibody coated with Ruthenium complex) into the sample without a wash step in-between, thus enabling a possible interfering antibody to react with the analogue. We hypothesize that the interfering antibody binds to the labelled analogue leaving less of this to be bound to the capture antibody leading to low chemiluminescence signal; thus falsely elevated FT4 concentrations. This same biotinylated T4 analogue is used in the Roche Cobas 6000 FT3 assay, thus enabling interference by a similar mechanism in the FT3 assay.
In our patient, treatment with thyroxine was originally discontinued after suspected over-treatment, which in turn led to recurrence of clinical hypothyroidism. Addison's disease was also suspected with inappropriate commencement of steroid treatment, which was later withdrawn.
Difficulties in interpretation of TFTs may have been compounded further by the possibility of thyroid hormone resistance 10 and DNA was acquired for mutational analysis of the thyroid receptor-β (TR-β) gene. Once the interferences were better documented, we did not proceed with mutational analysis. There is also some evidence that cortisol deficiency promotes TSH secretion, 11 which confounded the interpretation of our results. This mechanism was entertained in the context of the subnormal ACTH stimulation test and elevated TSH with normal FT4 results. These highlight the complexity of interpretation of test results when faced with unrecognized heterophile antibody interference and the importance of suspecting it with discordant results or when laboratory results do not correlate with clinical findings. Confirmation of heterophile antibody interference in our case avoided expensive investigation including mutation analysis of the TRB gene and prevented potential adverse long-term effects of steroid over replacement.
Conclusion
Interference by heterophile antibodies can affect a variety of immunoassays and may lead to expensive further investigations and adverse outcomes for the patient. Early suspicion of interference when laboratory results are discordant with each other or with the clinical findings may prevent these. Close communication between the clinicians and laboratory staff is vital in early recognition and confirmation of suspected cases of analytical interference.