Concordance in COVID-19 serology, bone mineralization, and inflammatory analytes between venous and self-collected capillary blood samples exposed to various pre-analytical conditions

Abstract

Background

The COVID-19 has led to a significant increase in demand for remote blood sampling in clinical trials. This study aims to ascertain the concordance between venous versus capillary samples, processed immediately or exposed to various pre-analytical conditions.

Methods

Participants (≥12 years old) provided a venous blood sample (processed immediately) and capillary samples allocated to one of the following conditions: processed immediately or exposed to 12-, 24-, or 36-h delays at room temperature or 36-h delays with a freeze-thaw cycle. The analytes of interest included SARS-CoV-2 IgG, 25-hydroxy vitamin D (25(OH)D), alkaline phosphate (ALP), calcium (Ca), phosphate (Ph), and c-reactive protein (CRP). Paired samples were considered interchangeable if they met three criteria: minimal within-subject mean difference, 95% of values within desirable total errors, and inter-class correlation (ICC) > 0.90.

Results

90 participants (44.1% male) were enrolled. When comparing rapidly processed venous with capillary samples, 25(OH)D, ALP, and CRP met all three criteria; SARS-CoV-2 IgG met two criteria (mean difference and ICC); and Ca and Ph met one criterion (mean difference). When considering all three criteria, concentrations of 25(OH)D, CRP, and ALP remained unchanged after delays of up to 36 h; SARS-CoV-2 IgG met two criteria (mean difference and ICC); Ca and Ph met one criterion (mean difference).

Conclusion

These findings suggest that remote blood collection devices can be used to measure anti-SARS-CoV-2 IgG, 25(OH)D, CRP, and ALP. Further analysis is required to evaluate the interchangeability between venous and capillary testing in Ca and Ph levels, which are more sensitive to pre-analytical conditions.

Introduction

The COVID-19 pandemic has significantly impacted face-to-face clinical and research visits, leading to an increase in innovative virtual visits via telecommunication tools. ¹ As a result, self-sampling methods of blood collection have become more widespread in clinical trials. The most popular self-sampling method is through a finger prick, which collects a small volume of capillary blood that can either be analyzed at home, such as with a glucometer, or applied to a filter paper for later laboratory analysis. ² New blood sampling devices such as Loop (Loop Medical SA, Lausanne, Switzerland), Tap (YourBio Health, Boston, USA), and Tasso-SST (Tasso Inc, Seattle, USA)^3–5 promise easier usage, less pain, and more blood volume, although the volume varies across devices. The Tasso-SST blood self-collection kit is reported to collect up to 700 μl blood (250 μl on average), via the upper arm (adults) or lower back (children). ⁵ While the accuracy of traditional finger prick blood collection has been well-studied, there is limited evidence on the comparability of self-collected capillary blood samples from these devices to venous blood samples.

Venous blood is commonly recognized as the reference standard for clinical analysis due to its composition and stability. ⁶ However, capillary samples are a mixture of venous, arterial, and capillary blood ⁷ that can impact the accuracy of analyte results. ⁸ The collection site and suction method of the Tasso self-collection device, specifically, may affect this mixture, leading to potential discrepancies in analyte results. The concordance between capillary and venous samples collected by self-sampling devices and finger pricks varies for different analytes, and depends on factors such as measurement techniques, pre-analytical conditions, and processing delay.

With respect to our analytes of interest, namely, 25-hydroxy vitamin D (25(OH)D), calcium (Ca), phosphorus (Ph), alkaline phosphatase (ALP), C-reactive protein (CRP), and anti-SARS-CoV-2 immunoglobulin (IgG), previous studies have reported a high level of concordance between capillary and venous blood-derived sera for anti-SARS-CoV-2 IgG, even when samples were stressed by winter and summer conditions and a 72-h processing delay. ⁹ As for the measurement of 25(OH)D, automated ELISA have shown lower values in venous compared to finger prick capillary samples, whether processed immediately or stored for 72–96 h prior to serum extraction. ¹⁰ However, our group previously reported high concordance in 25(OH)D levels between finger prick capillary and venous samples, when measured via tandem-mass spectrometry and processed rapidly, which highlights the impact of the measurement technique and pre-analytical conditions.^11,12 In another study, 25(OH)D, ALP, and CRP ¹³ levels were found to be comparable between venous and finger prick capillary samples that underwent 24-h delayed processing at room temperature, although the measurement techniques were not specified. There is limited evidence on the concordance between venous and capillary samples for Ca and Ph concentrations, and results of these studies are generally inconclusive.^14,15 High concordance in anti-SARS-CoV-2 IgG levels between venous and capillary blood collected by Tasso-SST on the upper arm was reported in 2021, even when samples were subjected to stress from winter and summer conditions and incurred a 72-h processing delay. ⁹ Thus, the level of agreement between venous and finger prick capillary samples may vary for different analytes. However, it is unclear whether the same level of agreement holds true for capillary blood collected via other self-collection devices such as Loop, Tap, and Tasso OnDemand, and to what extent the delay in the transition of blood samples to the laboratory may affect test results. Additionally, temperature control during transport of specimens is a limitation of home testing and requires optimization to protect against freeze-thaw cycles that could alter sample stability, particularly during winter.

The co-primary objectives of this study were to determine the validity of Tasso-SST OnDemand device: first, as an alternative to venipuncture, and second, as a home-collection device, by testing the stability of capillary samples exposed to various conditions, mimicking potential shipping situations. Specifically, we aim to ascertain the concordance in serum levels of anti-SARS-CoV-2 IgG, 25(OH)D, CRP, ALP, Ca, and Phosphorous (Ph) between (i) venous and capillary samples collected via the Tasso-SST OnDemand device, both processed rapidly (i.e., using Tasso as an alternative to venipuncture) and (ii) rapidly processed capillary samples versus those exposed to various processing delays and a freeze-thaw cycle (i.e., using Tasso as home-collection device). Secondarily, we wish to make recommendations on the interchangeability of samples exposed to different conditions, given the observed stability or lack thereof, of various analytes.

Methods

Study design and participants

We conducted a prospective cross-sectional study with a randomized factorial design at the Sainte-Justine University Health Centre (UHC), Montréal, Canada, between September 2020 and November 2020. The study was approved by the Sainte-Justine UHC Human Research Ethics Committee (approval #2021-3067). Participants provided written informed consent (or assent in adolescents) for study participation.

Eligible participants were aged 12 and older and did not have a recent COVID-19 infection (≤15 days since positive test). However, we specifically aimed to also recruit individuals with a high likelihood of prior COVID-19 infection, including those with previous positive PCR test results or symptoms identified by a score developed by Menni et al. ¹⁶

Sample collection, processing, and allocation

A single blood draw collected two (3–5 mL) aliquots of venous blood: one for 25(OH)D, Ca, Ph, ALP, and CRP analysis (red cap vacutainer with silica clot activator), and one for IgG against the SARS-CoV-2 S-protein (gold cap vacutainer with silica clot activator and polymer gel for serum separation). Venous samples were centrifuged at 2000 g for 10 min and analyzed within 4 h of collection. Aiming for a minimum volume of 250 μL for each capillary sample, four samples were collected via Tasso-SST OnDemand sampling kit (Tasso, Inc., Seattle, WA). A topical anesthetic cream (Ametop gel 4%) was applied to two of four capillary collection sites (dominant-side arm, non-dominant arm, dominant-side back, and non-dominant back). To ensure sufficient volume for all analytes, sera from two capillary samples using the topical anesthetic were decanted into the same aliquot during sample processing and allocated to rapid processing (centrifuged at 2000 g for 10 min and tested within 4 h of collection). The remaining two capillary samples were assigned to one of the following conditions in a randomized factorial design; delayed processing (12, 24, or 36 h) and temperature (room temperature [RT] only vs. RT with freeze-thaw cycle (4 h at −20°C) (Figure 1), following which they were combined into one aliquot, centrifuged at 2000 g for 10 min and analyzed. All analytes were examined for quality control on the same day as the analysis.

Figure 1.

Schematic depiction of the study protocol.

The first 13 participants' samples were allocated to the freeze-thaw condition, but 5 of the 7 samples were not usable due to severe hemolysis and artifacts. To find a suitable protection during the freeze-thaw cycle, the factorial randomization was discontinued. Subsequently, patient samples were allocated in groups of 20 in the following order: 36 h at RT, then 24 h at RT, then 12 h at RT, then a 36-h delay with freeze-thaw cycle using adequate temperature protection (Phase 22, Cryopak, Anjou, Canada) and an insulated envelope (Insulated Products Corporation, Rancho Dominguez, United States).

All blood samples were analyzed at the Sainte-Justine UHC clinical laboratory following the manufacturer’s recommendations. The levels of 25(OH)D and SARS-CoV-2 S-protein IgG were measured by automated chemiluminescence analyzer using the DiaSorin LIAISON XL platform; CRP level was measured by quantitative immunoturbidimetric method using Abbott ARCHITECT platform; serum Ca, P, and ALP levels were measured using the same Abbott ARCHITECT platform. The mean laboratory temperature monitored in daily logbook was 21°C (minimum: 20°C; maximum: 22°C).

Statistical analysis

With the Shapiro–Wilk normality test serving to assess the distribution, the data are presented as mean ± standard deviation [SD] for normally distributed parameters, or median (25%, 75%), otherwise. Standard Bland–Altman plots, displaying the absolute mean within-subject difference by the mean values, were constructed to check for systematic and proportional biases. Importantly, modified Bland–Altman plots, that is, displaying the relative mean within-subject difference (y-axis) by the mean values (x-axis), were used for interpretation purposes. Indeed, for each analyte, the co-primary outcomes were the within-subject relative differences between: (i) capillary versus venous samples processed rapidly and (ii) capillary samples processed rapidly versus exposed to various processing delays (12, 24, 36-h) and temperature conditions (with/without with a freeze-thaw cycle). The desirable total error (TE) was selected as a decision limit of a clinically acceptable relative difference and was determined from local data based on the equation suggested by Biswas et al., ¹⁷ namely, 2.55% for Ca, 10.11% for Ph, 12.4% for ALP, 21.1% for 25(OH)D, and 56.6% for CRP. The laboratory intra-run precision for SARS-CoV-2 IgG was used as acceptable relative difference limit, which is 8% based on local data. Within-subject differences for 25(OH)D, Ca, Ph, and ALP were also evaluated with reference change values (RCV), which take into account within-subject biological variation and intermediate precision calculated from Sainte-Justine UHC biochemistry laboratory quality control data. RCVs are 6.53% for Ca, 22.8% for Ph, 18.6% for ALP, and 34.7% for 25(OH)D. RCV could not be calculated for SARS-CoV-2 IgG and CRP, as these two analytes were measured using semi-quantitative assays on an Abbott Architect analyzer and intra-individual biological variability could not be evaluated.

A two-step paired comparison was performed for capillary samples subjected to different delays and temperature conditions. First, we combined all samples subjected to any type of delays and conditions to compare them with their rapidly processed matched sample, and we calculated within-subject relative and absolute differences between the paired samples. The former was compared to desirable TE limits described above, while for the latter, we performed a one-sample t-test to ascertain the difference from zero. Second, we repeated the process comparing samples subjected to a given delay or temperature condition with the rapidly processed matched samples.

For all comparisons, two additional concordance analyses were performed between paired samples. We ascertained the proportion of pairs within the desirable TE (or within-subject biological CV for IgG). For IgG serology, we also compared the positive (IgG≥15) and negative (IgG<15) categories in IgG serology between rapidly processed venous versus capillary samples and between capillary samples rapidly processed and those subjected to various conditions. We also calculated the intraclass correlation coefficient (ICC) between matched samples; an ICC >0.90 was deemed indicative of excellent agreement; 0.75 < ICC <0.9, good reliability; 0.5 < ICC <0.75, fair-to-moderate reliability; and ICC <0.5, poor reliability. ¹⁸

Finally, we adapted the algorithm proposed by Nwankwo et al. ¹³ to determine the interchangeability of the paired sampling methods (venous vs capillary) and paired pre-analytical conditions based on three statistical tests, each scored one point, namely, (i) 95% of within-subject relative measurements within the desirable TEs (instead of within 10% relative difference) on the Bland–Altman analysis; (ii) no statistically and clinically significant mean difference, based on desirable variations (instead of relying only on the statistically significant difference on paired t-test); and (iii) ICC greater than 90% (instead of simple correlation coefficient (r) > 0.8). A value of three points indicated interchangeability between two methods or pre-analytic conditions; a value of two points was likely reliable and clinically useful for longitudinal monitoring but did not support interchangeability; a score of one indicated the need for further analysis to understand clinical utility; and 0 point signaled an unreliable method or condition.

All tests were 2-sided, and a p value of <0.05 was considered statistically significant. Statistical analyses were conducted using R version 4.2.1 (R Foundation for Statistical Computing, Vienna, Austria, 2022).

Results

A total of 90 participants (44.1% male) were enrolled in the study, with a median (25%, 75%) age of 32 (26, 40.2) years; of these, two were adolescents aged between 12 and 17 years.

Concordance between venous and capillary samples

Eighty-nine subjects were included in the analysis of the concordance between venous and capillary samples, with one subject excluded due to hemolysis of the venous sample. The distribution of each analyte based on the sampling method is presented in Figure S1.

Minimal variation was found between venous and capillary serum concentrations of anti-SARS-CoV-2 IgG antibody, CRP, ALP, Ca, and Ph based on absolute and relative within-patient mean differences (Table 1). Of note, serum 25(OH)D was significantly lower in the venous than capillary samples by a mean absolute within-subject difference (95% CI) of 1.59 nmol/L (1.01, −2.16); however, the magnitude of the relative difference (3.37%) and all values were well within the desirable TEs (21.1%), and thus, the observed difference was not considered clinically significant. Standard Bland–Altman plots have shown no visually apparent proportional bias for all parameters (Figure S2).

Table 1.

Concordance between rapidly processed capillary versus venous samples.

Test	No of paired samples	Within-subject MD		Modified Bland–Altman		ICC	Interchangeably in clinical setting based on			Score
		Absolute MD (95% CI)	Relative MD (95% CI)	Desirable TE	% values within desirable TE	ICC a (95% CI)	MD NS and not clinically meaningful a	95% values within desirable TE	ICC >0.9
SARS-CoV-2 IgG	82	−0.55 c (−1.30, 0.15)	3.77 (−56.76, 64.31)	N/A	N/A	1.00 (0.99, 1.00)	1	N/A	1	2
25(OH)D	79	1.59 (1.01, 2.16)	3.37 (2.25, 4.50)	±21.1	100	0.95 (0.93, 0.96)	1	1	1	3
CRP	89	−0.05 c (−0.15, 0.00)	−0.47 (−3.25, 2.31)	±56.61	98.9	0.99 (0.99, 0.99)	1	1	1	3
ALP	89	−0.5 c (−1.0, 0.5)	−0.48 (−1.86, 0.9)	±12.04	96.6	0.99 (0.99, 0.99)	1	1	1	3
Ca	89	−0.01 (−0.02, 0.003)	−0.41 (−5.92, 5.09)	±2.55	64.0	0.82 (0.67, 0.89)	1	0	0	1
Ph	88	−0.002 (−0.03, 0.02)	0.32 (−2.45, 3.10)	±10.11	84.1	0.88 (0.82, 0.92)	1	0	0	1

MD: mean difference; TE: total error; ICC: inter-class correlation; NS: not statistically significant; IgG: immunoglobulin G; 25(OH)D, 25-hydroxy vitamin D; CRP: C-reactive protein; ALP: Alkaline phosphatase; Ca: calcium; Ph: phosphate.

Reference category: Venous samples.

^aMeasured by paired t-test (parametric) or Wilcoxon Signed-Rank test (non-parametric data). To define whether the found variation is clinically acceptable, the desirable TE was selected as a decision limit.

^bICC<0.50: Poor reliability; ICC: 0.50–0.75: moderate reliability; ICC: 0.75–0.90: good reliability; ICC >0.90: excellent reliability. ¹⁸

^cResults represents median of absolute difference and not mean.

When relative within-subject differences were considered, 100%, 98.9%, and 96.6% of values for 25(OH)D, CRP, and ALP fell within desirable TEs, respectively; however, 64% and 84.1% of Ca and Ph values were within desirable TEs, respectively (Table 1, Figure 2). Our analysis also showed that over 95% of values for 25(OH)D, ALP, and Ca were within RCV range. As for anti-SARS-CoV-2 IgG antibody, 70.7% of within-subject differences were within laboratory intra-run precision CV (Figure 2). Moreover, the concordance of the interpretation of SARS-CoV-2 IgG antibody levels between capillary and venous samples was 95.4% for 22 subjects with positive results (≥15UA). One case showed a discrepancy, with the venous sample being positive (15.8UA) and the capillary sample negative (13.4UA). The intraclass correlation coefficient for all analytes, except Ca and Ph, showed excellent agreement (>90%) between capillary and venous serum levels. Ca and Ph showed good concordance (82% and 88%, respectively).

Figure 2.

Modified Bland–Altman comparison for venous and capillary samples. The relative difference between two within-subject measurements (capillary and venous) calculated as values in (capillary – venous)/venous multiple by 100 (Y-axis) is plotted against the mean of the two measurements (X-axis) for each analyte, namely a) SARS-CoV-2 Ig G b) 25-hudroxy vitamin D, c) C-reactive protein, d) alkaline phosphatase, e) calcium, and f) phosphorous. The mean relative difference is depicted by the solid line. The upper and lower limits of the desirable total error (TE) around the mean relative value are depicted by the dotted black lines with ± desirable TE indicated in the right upper corner of each graph.

Overall, (OH)D, CRP, and ALP met all three criteria set forth by the adapted Nwankwo algorithm, indicating interchangeability between capillary and venous samples. SARS-CoV-2 IgG antibody met at least two criteria, as determined by minimal absolute and relative mean difference and high ICC, but the lack of identified total error prevented the determination of interchangeability. On the other hand, Ca and Ph met only one criterion, as evidenced by minimal absolute and relative mean difference (Table 1).

Concordance between rapidly processed capillary samples versus exposed to pre-analytic conditions

The distribution of each analyte according to pre-analytic conditions is displayed in Figure S3. Although the absolute mean differences within the subject for anti-SARS-CoV-2 IgG antibody, 25(OH)D, ALP, Ca, and Ph were found to be statistically significant, the magnitude of these differences was small for all analytes, except Ph, when evaluated using the relative mean difference.

The comparison of rapidly processed capillary samples with those exposed to various pre-analytical conditions showed no statistically significant absolute and relative mean differences for all analytes, except for Ph (Table 2). Furthermore, the relative within-subject difference showed high concordance for 25(OH)D, CRP, and ALP, with 100%, 98.7%, and 98.7% of values falling within desirable TE ranges, respectively. On the other hand, as expected, levels of Ca and Ph showed poor agreement between rapidly processed capillary samples and those exposed to pre-analytic conditions, with only 57.0% and 10.8% of values falling within desirable TE ranges, respectively (Figure 3).

Table 2.

Concordance between rapidly processed capillary samples versus those subjected to any delay and temperature condition.

Test	No of paired samples	Within-subject MD		Modified Bland–Altman		ICC	Interchangeably in clinical setting based on			Score
		Absolute MD (95% CI)	Relative MD (95% CI)	Desirable TE	% values within desirable total error	ICC b (95% CI)	MD NS and not clinically meaningful a	95% values within desirable TE	ICC >0.9
SARS-CoV-2 IgG	65	−1.05 c (−2.20, −0.29)	−3.71 (−33.17, 25.73)	N/A	N/A	0.99 (0.998, 0.999)	1	N/A	1	2
25(OH)D	61	−1.07 (−1.71, −0.43)	−1.88 (−2.96, −0.81)	±21.1%	100.0	0.99 (0.993, 0.998)	1	1	1	3
CRP	78	−0.00 c (−0.09, 0.05)	−1.59 (−4.65, 1.46)	±56.61%	98.7	0.99 (0.999, 0.999)	1	1	1	3
ALP	75	1.5 c (0.50, 1.99)	2.20 (1.25, 3.06)	±12.04%	98.7	0.99 (0.997, 0.999)	1	1	1	3
Ca	79	0.03 c (0.01, 0.05)	1.32 (0.03, 2.61)	±2.55%	57.0	0.62 (0.40, 0.76)	1	0	0	1
Ph	74	0.98 c (0.57, 1.21)	89.59 (63.3, 115.9)	±10.11%	10.8	0.025 (−0.54, 0.38)	0	0	0	0

Pre-analytical conditions included 12-, 24-, and 36-hours delay with and without freeze-thaw cycle.

MD: mean difference; TE: total error; ICC: inter-class correlation; NS: not statistically significant; IgG: immunoglobulin G; 25(OH)D, 25-hydroxy vitamin D; CRP: C-reactive protein; ALP: Alkaline phosphatase; Ca: calcium; Ph: phosphate.

Reference category: samples analyzed immediately.

^aMeasured by paired t-test (parametric) or Wilcoxon Signed-Rank test (non-parametric data). To define whether the found variation is clinically acceptable, the desirable TE was selected as a decision limit.

^bICC<0.50: Poor reliability; ICC: 0.50–0.75: moderate reliability; ICC: 0.75–0.90: good reliability; ICC >0.90: excellent reliability. ¹⁸

^cResults represents median of absolute difference and not mean.

Figure 3.

Modified Bland–Altman comparison for capillary samples analyzed immediately and exposed to any pre-analytical conditions (12-, 24-, and 36-h delayed processing in room temperature and 36-h delayed processing with a freeze-thaw cycle (with phase change and insulated packaging)). The relative difference between two within-subject measurements (immediate vs. any delay) calculated as values in (any delay – immediate processing)/immediate processing multiple by 100 (Y-axis) is plotted against the mean of the two measurements (X-axis) for each analyte, namely a) SARS-CoV-2 Ig G b) 25-hudroxy vitamin D, c) C-reactive protein, d) alkaline phosphatase, e) calcium, and f) phosphorous. The mean relative difference is depicted by the solid line. The upper and lower limits of the desirable total error (TE) around the mean relative value are depicted by the dotted black lines with ± desirable TE indicated in the right upper corner of each graph.

Our results also showed that a high concordance was observed for 25(OH)D and ALP, with over 95% of within-subject differences falling within the RCV range. Conversely, only 55% and 11% of Ca and Ph values were within the RCV. Although only 75.4% of anti-SARS-CoV-2 IgG antibody CVs were within the local laboratory’s intra-run precision, there was no discordance noted in the interpretation of the antibody between rapidly processed and delayed processed samples. An excellent ICC (0.99) was observed between capillary samples that were rapidly processed and those subjected to pre-analytic conditions for anti-SARS-CoV-2 IgG antibody, 25(OH)D, CRP, and ALP (Table S1).

Next, we compared capillary samples analyzed immediately with those exposed to various pre-analytic conditions, including 12-, 24-, and 36-h delays at room temperature or a 36-h delay with a freeze/thaw cycle and insulated envelope. Our findings showed that serum concentrations of 25(OH)D, CRP, and ALP were comparable in both the immediately analyzed and pre-analytic conditioned samples based on within-subject relative differences (Table S1). No statistically and clinically significant differences were observed in Ca following a 12-h delay at room temperature or a 36-h delay with a freeze/thaw cycle and insulated envelope. Ph showed the highest discordance, with clinically and statistically significant within-subject absolute and relative differences observed between the immediately analyzed and delayed processed samples.

Furthermore, the relative within-subject differences showed high concordance in serum concentrations of 25(OH)D and CRP, with over 95% falling within desirable TEs. Ca and Ph values were within the desirable TEs in 22.2–81.0% and 0.0–23.8% of cases, respectively. 100% of within-subject differences between reference samples and those exposed to 2-h, 24-h, and 36-h conditions with freeze/thaw cycles were within desirable TEs for ALP, while 94.4% of those exposed to 36-h delay at room temperature were within desirable TEs.

Except for Ca and Ph, an excellent agreement (ICC >0.98) was seen between any of the paired samples for all analytes. The ICC for Ca varied from 0.21 to 0.84 across the four pre-analytic conditions, while it ranged from 0.44 (12-h delay at room temperature) to <0.10 (at 24 h and beyond) for Ph.

Overall, 25(OH)D, CRP, and ALP concentrations had a total score of three, indicating they remained unchanged despite up to 36 h of processing delays with or without freeze/thaw cycles. SARS-CoV-2 IgG antibody received a score of two, meeting the mean difference and ICC criteria, but the absence of total error prevented further assessment. Ca and Ph had a score of one, meeting only the minimal mean difference criteria and thus appeared to be most sensitive to pre-analytical conditions (Table 2).

Discussion

The present study demonstrated that 25(OH)D, CRP, and ALP met all a priori criteria indicating interchangeability between venous and capillary samples, whether the latter were processed immediately or exposed up to 36-h delay processing (with and without a protected freeze/thaw cycle). SARS-CoV-2 IgG antibody met two criteria, indicating longitudinal usefulness, and Ca and Ph met only one criterion, thus appearing most sensitive to type of sampling and pre-analytic conditions. This suggests the need for further analysis to determine clinical research usability.

The results of the present study confirmed that venous and capillary samples of anti-SARS-CoV-2 IgG antibody showed high agreement, as evidenced by the strong intraclass correlation and minimal mean difference. This is consistent with previous studies that reported capillary blood self-sampling to be a reliable alternative to venipuncture for SARS-CoV-2 serology.^9,19–21 In two previous studies, capillary blood was collected using finger-prick dried blood spots,^19,20 while the other two studies utilized the Tasso-SST self-collection device.^9,21 Hendelman and colleagues ⁹ specifically investigated the impact of delayed processing (up to 56 h) and varying temperatures (summer conditions: 8hrs at 40°C, 4hrs at 22°C, 2hrs at 40°C, 36hrs at 30°C, then 6hrs at 40°C; winter conditions: 8hrs at −10°C, 4hrs at 18°C, 2hrs at −10°C, 36hrs at 10°C, then 6hrs at −10°C) on IgG levels and found that they remained stable in all conditions as demonstrated by paired t-test and linear regression analysis. Our study confirmed these findings and also showed that SARS-CoV-2 IgG antibody concentrations in capillary samples remained stable for up to 36 h, whether processed immediately or exposed to room temperature or a freeze/thaw cycle in an insulated envelope.

Our findings showed that 25(OH)D levels in both capillary and venous samples were interchangeable, meeting all criteria for interchangeability, regardless of any pre-analytical conditions imposed. This is consistent with two previous studies, which showed that the 25(OH)D levels in capillary serum were nearly the same as in venous samples.^11,12 One study even reported stable 25(OH)D levels in venous samples after being kept at 32°C for 3 days and at 11°C for up to 7 days. ²² However, one study ¹⁰ found a significant correlation between the 25(OH)D levels in capillary and venous samples, with the average 25(OH)D level being 18 nmol/L higher in immediately analyzed capillary serum. Taken together, our results add to the growing evidence supporting the use of capillary samples for measuring 25(OH)D.

Our study found that capillary and venous testing was interchangeable for CRP and ALP. These results align with a previous study that found excellent agreement between capillary and venous CRP (based on paired t-test with r > 0.8 and 95% relative difference within 10%) and ALP (based on paired t-test with r > 0.8). ¹³ Another study found CRP levels to remain stable in capillary samples kept at room temperature for up to 36 h prior to processing. ²³ Similarly, in another study, CRP was reported to remain stable in capillary samples for 3 days in room temperature (demonstrated by minimal change in absolute values). ²⁴ Despite the paucity of studies on the topic, we identified one study that showed ALP levels were comparable between capillary samples collected by finger prick and stored at room temperature for 72 h and venous samples analyzed within 4 h, as shown by minimal absolute and relative difference. ¹⁵ Our findings are in line with these previous studies.

In our study, there was no statistically significant difference in the mean within-patient absolute and relative mean differences of Ca and Ph between paired venous and capillary samples. However, less than 90% of the values fell within the desirable TE range, and the ICC indicated good reliability for Ph and fair-to-moderate reliability for Ca. This is in line with a previous study that reported discrepancies between Ca levels in capillary and venous specimens. ²⁵ Our results showed that Ca and Ph were the most impacted by processing delays, with Ca exhibiting fair-to-moderate reliability as indicated by ICC, but with a suboptimal distribution of pairs within the desired total error limit up to 24 h. Ph showed poor concordance by all criteria with any processing delay. The limited available literature provides limited evidence on the effects of processing delays on Ca and Ph levels, which is not unexpected, particularly for Ph, due to its known poor stability during delayed separation of whole blood. ²⁶ While one study showed that rapidly processed capillary specimens had lower Ca levels than venous specimens (≤5%), ¹⁴ another study found that capillary blood collected via finger prick and stored at room temperature for 72 h had significantly higher Ca levels compared to rapidly processed venous samples, with a mean difference of 0.3 mmol/L. This clinically significant difference was attributed to the higher calcium in interstitial fluid than in capillary blood and to finger squeezing, which could have contributed to the increased component of interstitial fluid. Further studies are needed to determine the clinical significance of rapidly processed venous and capillary testing for Ca and Ph. However, as delayed processing has a major impact on the accuracy of serum concentrations of these two tests, home collection is not recommended for monitoring Ca and Ph.

The strengths and limitations of the study are as follows: First, all samples from a single individual were analyzed in a single run to minimize analytical variation. Second, a solid protection package was developed and tested for Canadian winter that effectively protected the samples, as evidenced by the lack of differences between immediately analyzed capillary samples and those exposed to freeze/thaw cycles in the majority of analytes. However, the small number of samples exposed to each pre-analytic condition may have limited the power to detect meaningful differences, and our interpretation considered the width of the confidence intervals. The target population was individuals over 12 years of age, but only two participants aged 12–17 years were recruited, so caution is advised when generalizing the findings to the pediatric population.

Conclusion

In summary, our study demonstrated that venous and capillary samples were interchangeable, as were capillary samples analyzed immediately and those subjected to 36-h delayed processing (with or without a protected freeze/thaw cycle) for 25(OH)D, CRP, and ALP. For SARS-CoV-2 IgG, although the absence of total error prevented us from confirming all criteria, the within-subject mean differences met two criteria, and our results were concordant with the interpretation of test results. Venous and capillary Ca and Ph values showed high concordance but exceeded acceptable limits, making Ca useful for up to 24 h of delayed processing, but Ph should be analyzed immediately due to its sensitivity to pre-analytic conditions. These results support using remote blood collection devices for measuring anti-SARS-CoV-2 IgG, 25(OH)D, CRP, and ALP.

ORCID iD

Banafshe Hosseini https://orcid.org/0000-0002-3962-2777