Abstract Background Thyroglobulin (Tg), a glycoprotein that is synthesized in follicular cells of the thyroid gland, is the precursor of Thyroxin (T4) and Triiodothyronine (T3). Tg is a biomarker used for many clinical conditions, including monitoring for recurrence in patients with differentiated thyroid carcinoma. Typically, Tg levels drop to very low or undetectable levels after surgical excision and/or radioiodine ablation of the thyroid, with post-operative levels greater than 1-2 ng/ml indicating potential disease recurrence. Discordant Tg results in a patient sample were identified during a quality check of Tg results. This study investigates the root of cause of the discrepant Tg results. Given that: 1) Tg analyses were performed twice a week at our laboratory due to low daily test volume and patients’ samples received on an unanalytical day were stored at -20°C until analysis; 2) Tg analyses were performed by various medical technologists; 3) sample mixation prior to analysis was not routine practice for our automated laboratory, we hypothesized that the discrepancy was due to inconsistent sample handling, e.g., mixation vs unmixation prior to analysis. Methods Analysis of the daily QC and PT results in the period of the incident occurence ruled out causes in analytical phase. We split 52 de-identified patient serum samples into two equal aliquots and subsequently stored both aliquots at -20°C for 48 hours. Both aliquots of each sample were thawed at room temperature for 1 hour. One of the aliquots was analyzed for Tg without mixing, and the other was analyzed after mixing using a Vortexer for 4 seconds at 250 xg. To investigate if Tg concentrations form a gradient from the top to the bottom of a red-top vacutainer tube, five patient samples with Tg at 2.59, 2.79, 2.92, 52.37 and 144.12 ng/mL were thawed without mixing following storage at -20°C for 48 hours. Five layer (600 μl/layer) aliquots were carefully and sequentially taken from top to bottom for each specimen, and the Tg concentration in the aliquot of each layer was determined using the Beckman Coulter Access-2 Immunoassay. Results Of the 52 patients samples, 14 Tg results were excluded from statistical analysis due to exceeding the assay dynamic range (2 above and 12 below the upper and low limits, respectively). Analysis of the remaining 38 patient samples showed the mean of Tg concentrations was 22.8% (SD 18%) higher in mixed samples than the un-mixed samples. Regression analysis showed an R2 = 0.9976 (Mixed Tg = 1.1523 x Un-mixed + 0.4174, range 0.1 - 275.7 ng/mL, intercept 95%CI -0.721 to 1.556, slop 95%CI 1.1331 to 1.1715, p-value <0.001). Results of Tg gradient analysis showed significant higher Tg values in the 2nd, 3rd, 4th and 5th than the 1st layer by 26% (SD 11%), 63% (26%), 112% (57%), and 160% (76%), respectively. Conclusions False Tg values were found in samples without mixing before analysis after thawing from frozen state, due to formation of Tg concentration gradient in test tube. For obtaining reliable results, samples must be mixed before analysis when thawed from frozen storage.