Long-term stability of thyroid peroxidase antibody (anti-TPO) in serum in the Danish General Suburban Population Study

Introduction

There is a considerable variation in stability and recovery among biomarkers measured on biobanked samples after long-term storage [1]. Reasons for biobanking samples are many, but often researchers decide to measure new biomarkers, which were not considered at the time of enrolment of participants. Thus, it is valuable to know, if biomarkers measured on long-term frozen biobanked samples reliably reflect the fresh values.

Concentrations of thyroid peroxidase antibodies (anti-TPO) vary with age and gender and are associated with hypothyroidism [2], [3], [4], [5]. There is only one previous study evaluating the effect of long-term storage on anti-TPO measurement [6]. That study evaluated storage for up to 23 years at −25 °C in 645 pregnant women. For approximately every second year of storage time, a sample of 50 frozen serum samples were compared with 50 frozen serum samples stored for six months, using samples from different participants at each time point. The concentrations of anti-TPO increased steadily with storage time with a significant difference after two years, and a considerable jump in values after 14 years. The authors concluded that results were likely attributed to long-term storage, because other pre-analytical conditions (sampling, vacutainers, transport, thawing, and storage temperature) did not change during the study.

In the Danish General Suburban Population Study (GESUS), we evaluated retrospectively the long-term stability of serum anti-TPO at −80 °C by repeating measurements on the same participants (n=70). Fresh samples were measured in 2010–2011 and biobanked samples were measured in 2022.

Materials and methods

Results

The average storage time was 11.9 years (SD: 0.43), and two of the 70 samples, which were originally above the LoD, were below LoD when remeasured in frozen samples (Table 1).

Comparing anti-TPO_frozen with anti-TPO_fresh using Bland-Altman revealed that the line of equality was within the confidence interval of the absolute mean difference [5.71 (95% CI: −0.32; 11.7) U/mL] (Figure 1C and Table 2). The line of equality was also within the confidence interval of the average percentage deviation [2.22% (95% CI: −3.89%; +8.34%)] (Figure 1D and Table 2). Thus, the average percentage deviation of 2.22% was not statistically significant and did not exceed the desired total allowable error of 37.1% for values below 60 U/mL or 23.4% for values above 60 U/mL (Supplementary Table 1). The Lin’s correlation coefficient was 0.87 (Figure 1A and Table 2). The Passing-Bablok regression [anti-TPO_frozen=−22.6 + 1.22*anti-TPO_fresh] (Table 2) revealed both a systematic and proportional difference between anti-TPO_frozen and anti-TPO_fresh. Based on visual inspection of the absolute difference plot, anti-TPO concentrations were generally lower in frozen samples than in fresh samples for fresh values below anti-TPO cut-off of 60 U/mL, but higher in frozen samples than in fresh samples for fresh values above anti-TPO cut-off of 60 U/mL (Figure 1A–D). The proportional difference was highest for low concentrations and lowest for high concentrations (Figure 1A–D). In total, 64 (91.4%) samples were correctly classified in anti-TPO_frozen compared with anti-TPO_fresh using an anti-TPO cut-off of 60 U/mL with a Kappa value of 71.8% (Table 3). The six incorrectly classified samples were all within 10 U/mL of the cut-off value (or within a 14.3% deviation from the cut-off value). Using an anti-TPO cut-off of 100 U/mL, 60 (85.7%) samples were correctly classified with a Kappa value of 71.4% (Table 3). Regressing the average difference per year, that is “spreading” the average percentage deviation of 2.22% out over approximately 12 years, the instability equation was PD%=0.20 [95% CI: −0.32 to 0.72] * time (years), p=0.45; thus, a statistically non-significant 0.2% difference in anti-TPO concentration per year.

Figure 1:

Comparison of anti-TPO measurement in fresh vs. frozen samples. (A–D) Fresh serum measured in 2010–2011 vs. frozen serum remeasured in 2022 in the same 70 participants. r_s, Spearman correlation coefficient; r_c, Lin’s concordance correlation coefficient. (A) Solid line, line of equality, dashed line=least-square regression line. (C, D) Solid line=line of equality, dotted lines=upper limit of agreement, mean of differences, and lower limit of agreement, dashed line=least-square regression line.

Discussion

This is the first stability study comparing anti-TPO measured on fresh and frozen serum samples from the same subjects before and after 12 years long-term storage. Although there was a systematic and proportional bias between frozen and fresh values, the line of equality was within the mean difference on both the absolute and relative scale. Classification of anti-TPO positivity was correct in 91% of frozen samples using an anti-TPO cut-off of 60 U/mL and 86% using an anti-TPO cut-off of 100 U/mL, which both indicated “good agreement” compared to the classification based on fresh serum samples. The Lin’s correlation coefficient also indicated agreement.

The evaluation of the impact of long-term storage should be compared to the expected analytical variability. The average percentage deviation of 2.22% was not statistically significant and the estimate was less than the desirable total allowable error of 37.1% for values less 60 U/mL and 23.4% for values greater than 60 U/mL. The desired total allowable error is based on a previous study on biological variability of anti-TPO in 24 healthy women between 23 and 46 years old and using thawed frozen serum samples stored for an unreported length of time [10]. Interestingly, the aspect of frozen storage was not discussed in that work, which, via the database provided at the Westgard website [14], probably provide the only evidence used to set desirable error rates for anti-TPO measurements in most laboratories around the world.

Other pre-analytical factors than storage can affect differences between frozen and fresh samples. In our study, the remeasurements were performed on samples thawed for the first time and there was no unintentional thawing of the material from 2010 to 2022 in the biobank. The preanalytical conditions (collection, vacutainers, processing, and storage conditions) at study enrolment were similar for all individuals in our study, although the serum for fresh anti-TPO was collected in a serum tube without separator while the biobanked serum was collected in a serum tube with separator.

Long-term storage of biobanked frozen plasma or serum may result in aggregation, precipitation, or biochemical degradation of proteins which can alter the tertiary structure and activity of the proteins [1]. The changes may be caused by ice damage to the proteins caused by water crystallization or dehydration of the samples and subsequently increased salt concentration resulting in osmotic damage of the proteins [1]. These factors may ultimately have caused changes in the measurements of anti-TPO when they were (re-)measured several years after collection.

The only other previous study of anti-TPO stability, used an unpaired design to compare the long-term anti-TPO stability over 23 years at −25 °C in 645 pregnant women from the Finnish Maternity Cohort [6]. The serum samples with the longest storage times were included from participants enrolled in the 1980s, whereas the serum samples stored for 6 months were included in the 2000s [6, 15]. The study did not provide the exact values for each year, but from our visual inspection of the presented boxplot, the median concentrations of anti-TPO in samples measured after 6–10 years storage time were doubled compared to baseline [6]. This is more than what we observed in our study after 12 years of storage. However, changes in population demographics and iodine sufficiency over time in this cohort could be explanatory factors for these changes instead of storage time per se [16]. In the 1980s, Finland had the highest iodine intake in Europe leading to the iodine fortification of salt no longer deemed necessary in 1986 [16]. However, since then iodine intake decreased and in the 2000–2010s the general Finish population displayed mild iodine deficiency [16]. Speculatively, the positive association between thyroid autoantibodies and storage time in the Finnish cohort, may be related to the substantial decrease in iodine intake observed during the same period.

Anti-TPO antibodies are predominantly subclasses of IgG antibodies [17]. A previous study, also using an unpaired design, assessed the stability of total IgG after one month, 2 and 25 years of storage at −25 °C comparing three different groups, each comprised of 130 healthy Norwegian male blood donors [18]. In that study, results were more similar to our data, and a non-significant percent difference of mean of 1.0% in total IgG was observed when comparing serum stored for 25 years with serum stored for one month.

Stability studies on long-term frozen samples are more prone to instrument change than short-term stability studies over hours, weeks, or months. This makes it challenging to separate the difference between frozen and fresh samples due to long-term storage or the difference between the instruments. Our study design is limited by the update of the Kryptor analyzer during the long-term storage. The verification study performed in the lab at the time of updating the equipment showed that while measurements were slightly higher on the updated Kryptor, the precision (CV), accuracy (bias based on external control samples), and correlation with the Kryptor Classic were satisfactory (Supplementary Tables 2–4). However, the verification study for the agreement between the two Kryptor analyzers was performed in the range 10–25,410 U/mL and included an insufficient number of samples in the range of 30–200 U/mL (n=4) to make a meaningful comparison in this range (Supplementary Table 2). Therefore, the exact agreement in the relevant range is unknown. While this is a limitation, most other variables were identical for fresh and remeasured samples including the assay, the reagents, the calibrator, and the participants. Ideally, a stability study for any biomarker should be planned already at the collection of the samples for biobanking and sufficient samples should be aliquoted to allow for remeasurements multiple times during biobanking to investigate stability [1, 19]. Due to the retrospective design, in which fresh samples were only run once at baseline, it was decided that frozen samples were run under the same conditions. Ideally, samples in both states should have been run in duplicate to increase precision. While the stability study was not ideally designed, we find the information presented of practical relevance.

In conclusion, anti-TPO serum samples in the range 30–198 U/mL were stable after 12-years of storage at −80 °C with an estimated nonsignificant average percentage deviation of +2.22%. Both a systematic and proportional difference between fresh and frozen values contributed to this relative bias. This comparison is based on Kryptor Classic and Kryptor Compact Plus, which used identical assays, reagents, and calibrator, but for which the agreement in the range 30–198 U/mL is unclarified.