Monitoring of SARS-CoV-2 antibodies using dried blood spot for at-home collection

Assessment of DBS analytical measurement range and imprecision

DBS limit of blank (LOB)

In order to assess the detection capability of the assay with DBS extracts, guidance from CLSI EP17-A2 was utilize²⁵. Two separate reagent lots were used to make 96 blank measurements on 6 contrived blood samples over a 4-day period. Two results were not included in the data analysis for each lot as the z-score for each of these results with respect to the remaining results were greater than 4.7 for both reagent lots. Using the mean and standard deviation of the remaining blank results (n = 94) as well as a normal distribution multiplier, the LOB for DBS extracts was determined to be 0.111 U/mL (Supplementary Sect. 1).

DBS limit of detection (LOD)

The limit of detection (LOD) was assessed using 5 contrived blood samples in accordance with CLSI EP05-A3 and CLSI EP17-A2 guidance (Supplementary Sect. 2)^25,26. The samples utilized had mean results (across two reagent lots) that were expected to be close to the clinical cutoff (0.146–0.531 U/mL). Pooled standard deviations were calculated for the two different reagent lots using CLSI EP17-A2 guidance²⁵. The first reagent lot produced a pooled standard deviation of 0.0419 U/mL while the second demonstrated a standard deviation of 0.0346 U/mL. Using the larger standard deviation (0.0419 U/mL), a normal distribution multiplier, and the reported LOB (0.111 U/mL) the LOD for DBS extracts was determined to be 0.180 U/mL.

DBS limit of quantitation (LOQ)

Measurements to determine the limit of quantitation (LOQ) of DBS extracts utilized 14 contrived blood samples covering an appropriate range of concentrations (0.0528–0.648 U/mL) These samples were extracted and measured in triplicate over a five-day period (15 total measurements for each level) using two different reagent lots on a single instrument. For this study, the target CV and bias were set to 25.0% based on FDA guidance for ligand binding assays at the lower limit of quantitation²⁷. Following collection of data, the imprecision profiles were analyzed using the Limit of Quantitation module in EP Evaluator (Supplementary Sect. 3). These results indicated an LOQ of 0.0873 U/mL for the first reagent lot while the second reagent lot demonstrated an LOQ of 0.0736 U/mL. In addition, acceptable biases less than ± 25.0% were observed for all levels greater than the DBS LOD. As both imprecision and bias results indicate an LOQ less than the observed LOD, the LOQ for DBS extracts is in practice equivalent to the LOD—0.180 U/mL²⁵.

Investigation of the DBS assay’s LOB, LOD, and LOQ indicates that a lower reporting limit can be achieved for DBS extracts. This may be attributable to the reduction of sample dependent matrix effects as a result of ~ tenfold extraction (dilution) with the approved assay diluent²⁸. As a result of the increased sensitivity, a clinical cutoff value below the EUA approved assay’s LOQ for serum and plasma can be implemented for DBS extracts, enabling assessment of categorical agreement between venous and DBS samples near the serum cutoff (0.8 U/mL).

DBS linearity

Linearity of the assay with DBS samples was assessed using 11 antibody concentrations made through serum sample admixtures in 10% increments followed by contrivance into DBS (Supplementary Sect. 4). Initial assessment of linearity indicated acceptable biases (± 20.0%) from 0.0667 to 147 U/mL with 147 U/mL being the highest concentration tested. This study was repeated with samples of higher concentration to extend the measurement range to 250 U/mL in order to match the reporting limit of un-diluted samples for the EUA approved assay. For this second study, samples with extracted concentrations that initially measured > 250 U/mL were manually diluted tenfold with Universal Diluent and re-measured as indicated in the assay’s package insert²³. Antibody concentrations for this second study ranged from 0.0974 to 323 U/mL and targets were determined by using a linear fit of all data points.

DBS clinical cutoff

The clinical cutoff, used to assign DBS results as negative or positive for antibodies, for DBS samples was created using results from self-collected samples from donors confirmed to be seronegative (n = 77, Supplementary Sect. 5). The standard deviation of these results (0.0405 U/mL) was multiplied by 3 and added to the mean (0.0625 U/mL) to give a value of 0.184 U/mL. The DBS clinical cutoff for positivity was then set to ≥ 0.185 U/mL, where all results less than 0.185 are reported as negative.

DBS analytical measurement range

Evaluation of the detection capability of the assay with DBS extracts indicates that sample concentrations can be measured from 0.180 U/mL (LOD/LOQ) to 250 U/mL (upper reporting limit of assay) with a clinical negative/positive antibody cutoff concentration of 0.185 U/mL. This DBS range represents a calculated serum measurement range of 2.6 to 3570 U/mL. Although dilution of samples is performed for venous serum and plasma samples for the EUA approved in order to extend the reporting limit, dilution of DBS samples is not currently utilized due to the relatively wider concentration range that can be reported for DBS samples (as a result of pre-dilution through extraction).

Imprecision of DBS samples

Assay imprecision was assessed for DBS extracts using two reagent lots over a 4-day period with 6 contrived blood samples that covered a range of antibody concentrations (0.504–156 U/mL). A total of 16 replicates for each sample were measured (Supplementary Sect. 6). For both reagent lots, the CVs observed for repeatability and within-laboratory imprecision were less than 15.0% which meets FDA specifications for ligand binding assays (Table 2)²⁷. All samples had a total categorical agreement of 100.0%.

Clinical correlation study

Following collection and measurement of samples, 6 of the 84 presumed negative donors had venous serum, self-collected DBS samples, and professionally collected DBS samples measure positive for SARS-CoV-2 antibodies despite having negative RT-PCR results at the time of serological specimen collection (Supplementary Sect. 7). The serum results for these donors ranged from 8.81 to 1170 U/mL where the clinical cutoff for serum samples is 0.800 U/mL. These results suggest previous (asymptomatic) infection or unreported vaccination. To confirm these results, the serum samples for these donors were measured using three additional EUA approved serology assays (Roche Elecsys Anti-SARS-CoV-2, DiaSorin Liaison SARS-CoV-2 S1/S2 IgG and DiaSorin Liaison SARS-CoV-2 IgM)²⁹. All six donors had serum results measure as positive on at least one additional assay and as such results were excluded from qualitative and quantitative analysis. All remaining results obtained were analyzed qualitatively using the established clinical cutoff for DBS extracts (0.185 U/mL) and quantitatively through correlative analysis.

Comparison of serum samples with DBS obtained through self-collection demonstrated a high degree of agreement (R = 0.9600) and Deming slope of 0.069 (n = 108), which is attributed to dilution of the sample through the extraction process as well as precise but incomplete recovery of antibodies (Fig. 2A,C). Two donors with negative venous serum results had self-collected DBS samples that measured positive for antibodies (results within 3.5× the DBS clinical cutoff, Supplementary Sect. 5). As a result of these two false positives as well as a false negative donor with unmeasurable antibody results for both serum and DBS, qualitative total categorical agreement of self-collected DBS samples compared to venous serum results was 98.1% (Table 3A). In addition, the negative percent agreement (NPA) and positive percent agreement PPA were found to be 97.4% and 100.0%, respectively. When compared to RT-PCR results, qualitative categorical agreement was 97.2% with an NPA and PPA of 97.3% and 97.1%, respectively (Table 3B). Results below and above the DBS measurement range were interpreted as negative and positive, respectively.

When quantitatively comparing professionally-collected DBS samples to serum results (n = 106), results were similar to self-collected results with a correlation coefficient (R) of 0.9888 and a Deming slope of 0.070 (Fig. 2B,D). Total qualitative categorical agreement (n = 111) to venous serum results was 100.0% (Table 3C). When compared to RT-PCR results, qualitative categorical agreement was 99.1% with an NPA and PPA of 100.0% and 97.0%, respectively (Table 3D). These results met FDA guidance at the time of these studies (NPA ≥ 95.0%, PPA ≥ 90.0%)²².

Robustness studies

In order to evaluate sample stability during the shipping process, simulated shipping studies were performed in accordance with ISTA 7D guidance as recommended by the FDA^22,30. Contrived DBS samples were prepared in triplicate and split between three storage conditions: room temperature (20–25 °C), a simulated winter and summer shipping excursions (Supplementary Sect. 8). Following completion of the excursions, samples were measured in a single measurement run where results from samples stored in parallel at room temperature were used as baseline results.

Additional robustness and analytical interference studies were performed to stress different aspects of the DBS sample collection process as well as the influence of several endogenous or exogenous interferences (Supplementary Sects. 9 and 10). For acceptance of results, a mean bias of ± 20.0% was used as quantitative acceptance following the FDA guidance for ligand binding assays²⁷. For qualitative assessment, a total categorical agreement of 95.0% was utilized based on guidance from the FDA’s Home Specimen Collection Serology Template²².

DBS shipping stability

For shipping excursion studies, contrived blood samples were created in triplicate and split into three shipping conditions: baseline ambient (20–25 °C), winter and summer excursions (Supplementary Sect. 8). Results less than 0.180 U/mL (DBS assay LOQ) were included in qualitative analysis but excluded from quantitative analysis. Overall, the sample results from winter and summer excursions both demonstrated total categorical agreement of 97.4% with mean biases of 6.0% and − 0.5%, respectively (Table 4). These results indicate that samples are stable from the time of collection in an individual’s home until received in the laboratory.

Stress testing the collection process

For robustness studies, results that were found to be less than the assay’s DBS LOQ were excluded from bias analysis but included in qualitative analysis. Investigation of drying times compared to the recommended drying time of 3 h (prior to sample storage) was performed. Samples that were immediately stored after contrived blood was added to the card had a categorical agreement of 95.0% and mean bias of − 32.5%. Results from samples that were dried for 1 and 22 h demonstrated categorical agreements of 95.0 and 100.0%, respectively with mean bias of 1.3% for both (Table 4). Effects of drying samples in a humid environment (40 °C, > 95% relative humidity) were also investigated²². All results observed had a categorical agreement of 95.0%, but the magnitude of the bias increased from − 13.3% for 1 h to − 44.6% for 22 h of humid drying.

Contamination as a result of potential errors in the collection process was investigated (Table 4, Supplementary Sect. 9). Exposure of DBS spots to alcohol, which may occur following finger sterilization without allowing the fingertip to dry, demonstrated a categorical agreement of 100.0% and mean bias of − 11.0%. Contamination of the DBS card by an unsterilized finger prior to collection demonstrated a categorical agreement of 95.0% and mean bias of 1.3%. Although all robustness studies had a total categorical agreement greater than or equal to 95.0%, biases less than ± 20.0% were observed in some instances. As these studies are not exhaustive of all possible contaminants, proper instruction materials must be provided to the patient in order to insure the collection of a sample of sufficient quality for measurement.

Analytical interference studies

Several studies were performed to assess the effects of different endogenous and exogenous interferents on the measurement of SARS-CoV-2 antibodies from DBS samples (Supplementary Sect. 10). Results from all interference studies had a categorical agreement greater than or equal to 95.0% to baseline measurements (Table 5). The exogenous interferents tested had mean biases less than ± 5.0% while most endogenous interferents tested had a mean bias result less than ± 10.0% with the exception of excess protein (17.7%).

Immunization study

Application of self-collected DBS was performed with serial monitoring of SARS-CoV-2 antibody levels following immunization in 8 donors. Donors periodically performed self-collection of DBS samples pre-vaccination through 19 weeks following initial vaccination. All donors reported receiving the Pfizer-BioNTech COVID-19 vaccine with the second dose occurring exactly 3 weeks following the first dose. As can be seen in Fig. 3A below, all donors had negative DBS results for samples collected within the first 9 days following initial immunization. Antibody levels for each donor rose above the DBS cutoff of 0.185 U/mL between days 10 and 16. Antibody levels increased rapidly following the second vaccination dose (Fig. 3B), with several donors reaching levels greater than the DBS reporting limit (250 U/mL which is calculated to be 3570 U/mL in serum)^31,32. Dramatically lower antibody levels were observed for Donor 7, likely a result of the immunosuppressive medication the donor reported taking for a chronic condition³³.