Comment regarding “Positive SARS-CoV-2 RT-qPCR of a nasal swab spot after 30 days of conservation on filter paper at room temperature”

Summary of Review

The Correspondence seeks to compare and discuss RNA stability on Whatman 3M filter paper (evaluated in Durand et al.) versus Whatman™ Flinders Technology Associates™ (FTA) cards (evaluated in this Correspondence and by the Author in previous work). This comparison and discussion is useful, as these two dry matrices may not exhibit equivalent performance for stabilizing RNA.

The Correspondence concludes that “FTA cards might ... be the better choice [compared to Whatman 3M filter paper] with respect to the stable preservation of SARS-CoV-2 RNA on a dry matrix.” However, issues in methodology and clarity of the evaluations in this Correspondence and the study evaluating RNA stability on Whatman 3M filter paper limit the ability to make a balanced and justified conclusion.

Qualitative result of detection of SARS-CoV-2 RNA could be used to compare the two dry matrices. Data from Table 1 in this Correspondence shows that 19 of 19 samples were detectable after 21 days on FTA cards at room temperature. In comparison, Durand et al. report that after 30 days on Whatman 3M filter paper at room temperature, 6 of 10 samples were detectable. While this might appear that FTA cards result in better preservation of RNA, it is important to note first that the time lengths (21 days versus 30 days) are not equivalent, and second that the starting concentrations of target RNA are not equivalent. The Ct value at T0 was below 25 for 74% (14 of 19) samples stored on FTA cards at room temperature for 21 days, but only 20% (2 of 10) samples (originating from 3 clinical specimens) stored at room temperature for 30 days. The lower rate of detection for samples stored on filter paper is likely impacted by lower starting RNA amounts than samples stored on FTA cards. Third, RNA stability on the dry matrix may also be sample dependent, and different samples were used between the evaluations.

Quantitative result of SARS-CoV-2 RNA amount after storage could be used to compare the two dry matrices. Ct values for a SARS-CoV-2 RNA target in 19 samples stored on FTA cards at room temperature for 21 days were a median of 2.56 higher than the Ct value for the same sample on day 0, which corresponds to approximately 5.9-fold less target RNA. Among 5 samples with detectable target RNA after 30 days on Whatman 3M filter paper at room temperature, Ct values were a median of 5 Ct higher than on day 0. In comparison, 0 of 19 samples (originating from 2 clinical specimens) stored on FTA cards for 21 days exhibited 5 Ct higher than on day 0. This trend would suggest that RNA is more stable on FTA cards than filter paper, however the filter paper condition is drastically underpowered, the time lengths are different, and different samples were evaluated in each condition.

FTA cards may be the better choice for stable preservation of SARS-CoV-2 RNA compared to Whatman 3M filter paper, but the data presented is insufficient to make this conclusion. Additional studies (ideally head-to-head studies, with equivalent time lengths, and equivalent samples that span a range of starting target RNA amounts) are needed, as well as appropriate quantitative analysis methods. 

Major Comments:
1. The Article would benefit from a brief description of how the composition/structure of Whatman 3M filter paper and the Whatman FTA cards differ.

2. “Of note, the filter paper seems to convey some virus inactivation properties because the authors failed to isolate virus obtained from a dried spot after 20 days of tissue culture.” 

The claim that the filter paper contributes to the inactivation of the virus is not well substantiated. Firstly, this conclusion relies on only a single sample. Second, inactivation of enveloped viruses after several days outside of a live host is expected, and believed to be due to the kinetics of drying/dehydration (see French et al. mBio 2023). While the process of drying can be affected by surface material (ie. filter paper), one would need to determine the independent contribution of the surface material to viral inactivation to substantiated this claim. The article referenced (from which the data on recovery of infectious virus from a single sample stored on filter paper originates) does not make the claim that filter paper contributed to viral inactivation, and does not compare sample storage on different materials. I encourage the Author of this Correspondence to provide substantiating evidence, or revise this statement. 

3. “After week 3, we observed differences between the C_t values obtained for storage temperatures at -20°C, 4-8°C, RT, and 37°C and a reference FTA card sample (i.e., RNA was extracted immediately after the filter paper had dried out) to range from -0.6 to 1.1%, -2.5 to 4.9%, -1.8 to 4.9%, and 0.6 to 6.4%, respectively.”

Relative differences in Ct values are challenging to interpret and compare, because the magnitude of the relative difference depends on the absolute Ct values. For example, imagine that half the RNA in a sample degrades after 10 days, and that this rate of degradation is not dependent on the amount of starting RNA. If the material started at a Ct of 10, we could expect a Ct of 11 after 10 days: this is an absolute difference of 1 Ct, and a relative difference of 10% (1/10). If the material started at a Ct of 15, we could expect a Ct of 16 after 10 days: this is an absolute difference of 1 Ct, but a relative difference of 7% (1/15). If the material started at a Ct of 25, we could expect a Ct of 26 after 10 days: absolute difference of 1 Ct, but a relative difference of 4% (1/25). If the material started at a Ct of 35, we could expect – theoretically - a Ct of 36 after 10 days: absolute difference of 1 Ct, but relative difference of approximately 3% (1/35). Equivalent fold-changes in material result in lower relative differences of Ct values at higher starting Ct values. Variance metrics for relative difference in Ct (e.g. range) do not reflect variation in degradation among samples. 

I encourage the Author to use a clearer and more interpretable summary metric to substantiate this claim. Absolute differences in Ct values (ΔCt) would be a more interpretable summary metric for readers than relative differences. Or, explain why relative difference in Ct values is superior and provide the equation used to calculate relative difference in Ct values.

4. “FTA cards ... were shown to be suitable for the stable preservation of SARS-CoV-2 RNA derived from nasopharyngeal swabs (NPS) even at elevated storage temperatures”

Please also clarify how “stable preservation” is defined in the context of this statement. Data in Table 1, 15% (3 of 19) of the samples presented in Table 1 exhibited an absolute difference of Ct values greater than 3.3 (approximately 10-fold loss of RNA) between T0 and T21. 

5. “In a first experiment, Durand et al.¹ were able to amplify filter-bound SARS-CoV-2 RNA obtained from a nasal sample after storage of 10, 20, and 30 days at RT resulting in C_t gaps of 7, 6 and 5 between the extemporaneous sample at T0 and samples taken at T10, T20 and T30, respectively.”

This statement re-reports a result from Durand et al., without offering additional perspective. I encourage the Author to add an original interpretation to put this result in context, or remove.

However, the Ct gaps reported in this statement, while correctly reproduced from Durand et al., do not make scientific sense. (Perhaps this was what the Author sought to call attention to?) I investigated the underlying statement, and see that Durand et al. report a Ct of 21 at T0, Ct of 28 at T10, Ct of 27 at T20, and a Ct of 26 at T30. It is extraordinary that more RNA (lower Ct) would be measured from the same sample over time, since RNA is expected to degrade over time, resulting in higher Ct values. Indeed, in the second experiment reported in Durand et al. Table 1, Ct values increase over time, for all dilution conditions, as expected.

I believe data from Durand et al.’s first experiment was erroneously reported, and reproducing the error in this Correspondence should be avoided. The values reported in the main text of Durand et al. do not match underlying data shown in Durand et al. Supplementary Figure 1. At T0 (Durand et al. Fig S1A), sample fluorescence appears to rise above the RFU threshold at a Ct of approximately 21.5. The main text reports a Ct of 21 for this timepoint. At T10 (Durand et al. Fig S1B) I would interpret the Ct as 26, while they report a Ct of 28. At T20 (Durand et al. Fig S1C) I would interpret the Ct as 26, while they report a Ct of 27. Finally at T30 (Durand et al. Fig S1D) I would interpret the Ct as 26, and they report a Ct of 26. It is reasonably expected that initial degradation of RNA would occur between T0 and T10 (from 21.5 to 26), followed by stable Ct values for the sample over time (26). I believe an error was made by Durand et al. when reporting these Ct values in the main text of their manuscript. 

It is critical to avoid propagating this error. I encourage the Author to review and independently interpret the underlying data from Durand et al. Supplementary Figure 1. If the Author believes the measured Ct values to truly be 21, 28, 27, and 26 for T0, T10, T20, and T30 as reported, the Author should explain how this could be physically possible (e.g. greater liberation of RNA from the filter than degradation over time, variability in sample loading or measurement between timepoints) and how these possible explanations impact the conclusions that can be made from the data available.

6. “In a subsequent experiment analyzing freeze-dried virus samples, the authors reported a median C_t gap between an extemporaneous sample isolated at day 0 and different virus dilutions spotted and stored for 30 days at RT of 3.7 and 3.5 for the Delta and the Omicron strain, respectively. Between T10 and T30, the C_t gap was 2.5 and 1.3, respectively. In contrast and given the presented C_t values for the Delta strain at dilutions representing a TCID₅₀ of 4 and 0.4, we calculate the median C_t gap to be 6.25, corresponding to the difference between the arithmetical mean C_t values of 36 at T30 and 29.75 at T0. ... In the absence of more information, it would be interesting to learn how the authors performed their calculations, especially considering the fact, that these calculations seem to overestimate the filter paper’s performance to stably preserve SARS-CoV-2 RNA.”

I agree that it is unclear in Durand et al. how the median values reported were calculated, and appreciate the Author of this Correspondence raising this point. I have reanalyzed the data in Durand et al. Table 1, and I was able to reproduce the value they report when following an erroneous analysis method: for the Delta strain, it appears that they subset the data to only dilution conditions where data from both T0 and T30 was available (TCID50 of 4 and TCID50 of 0.4). They then calculated the median Ct value of the two T0 measurements (27.5 and 32), and the median value of the two T30 measurements (35 and 37). They then took the difference between the median from T30 and the median from T0 to get a Ct gap of 3.75.

I agree with the Author of the Correspondence that this approach is inappropriate. Measurements of Ct value at T0 and Ct value at T30 for a sample are not independent. The method used here neglects the paired nature of the data, and overrepresents the observed stability of RNA on the filter paper.

It is more appropriate to calculate the Ct gap (or absolute difference in Ct values, or ΔCt) between T0 and T30 for each sample, and then take the median of the set of Ct gaps, which retains the paired nature of the measurements. The Author of the Correspondence applied this method to obtain a Ct gap of 6.25, which I believe more accurately reflects the effect of time on the amount of RNA measured from the filter paper based on the underlying data than the reported value of 3.75.

However, this is a median of only two observations. These two observations, which are dilutions of a single sample, are insufficient to make quantitative conclusions about the stability of SARS-CoV-2 RNA on filter paper more generally. I would encourage the Author to make the point that this analysis in the referenced study is underpowered.

7. “Furthermore, based on the C_t values for the Omicron strain at dilutions representing a TCID₅₀ of 4, 0.4, and 0.04, we calculate the median C_t gap to be 4.5 by subtracting the median C_t value of 28.5 at T0 from the median C_t value at of 33 at T30. With this respect, and by subtracting the median C_t value of 31 at T10 from the median C_t value of 33 at T30, the resulting C_t gap is 2.”

As discussed in my Comment 6, this method of calculating the median Ct gap is inappropriate, because it neglects the paired nature of T0 and T30 measurements of the same sample. The Author of this Correspondence correctly pointed out the inappropriateness of this mathematical approach in reference to Durand et al., but the critiqued approach was then used by the Author here.

If the correct approach (described in Comment 6 above) is applied, the median Ct gap for the Omicron variant observations TCID50 of 4, TCID50 of 0.4, TCID50 of 0.04 between T30 and T0 in Durand et al. Table 1 would be 5.

However, this is a median of only three observations, which are dilutions of a single sample. I would encourage the Author to make the point that this analysis in the referenced study is underpowered.

8. I am unable to reproduce summary metrics provided in the text of this Correspondence from data in Table 1

Table 1 provides the measured Ct values for samples stored at room temperature. These data are very helpful. However, I am unable to reproduce the summary metrics listed in the text using data from this table.

The range of relative differences in Ct values from T21 to T0 at room temperature is listed in the text as -1.8 to 4.9%. By my calculations from data in Table 1, the range should be 4% to 15%. The value 4% is from Sample 16 in Table 1, which has a T0 Ct of 25.61 and a T21 Ct of 26.75.    (26.75-25.61)/25.61 = 0.04451, or ~4%. The value 15% is from Sample 8 in Table 1, which has a T0 Ct of 24.39 and a T21 Ct of 28.04. (28.04-24.39)/24.39 = 0.14965 or ~15%. If the relative difference in Ct values (and/or other metrics reported in the text) were reported in Table 1 for each sample individually, readers could trace back how the reported values were obtained. 

However, as described in my Comment 3, I do not recommend reporting relative differences in Ct values. If the Author wishes to report relative differences in Ct values, clarity on how this calculation was performed to get the reported values is requested. Details for how these values were calculated is not provided in the Author's prior work cited. 

Additionally, the median C_t gap (absolute difference in Ct values, or ΔCt) to T0 is listed in the text as 2.80 for T10, 2.79 for T20, and 2.83 for T30. I calculated these values to be 2.37 for T10, 2.61 for T20, and 2.56 for T30. I cannot upload my re-analysis spreadsheet, but am happy to provide upon request. 

Minor comments:
1. In Table 1, NPS Sample 1’s Ct at T0 has a “,” while all other data in the Table contain decimal points (“.”).

2. “The median C_t gap between the extemporaneous NPS (T0) samples and the FTA spot-prep samples extracted at T7, T14, and T21 were 2.80, 2.79, and 2.83, respectively, indicating that FTA cards might not be the cheaper (i.e., approximately two euros per FTA card sample vs. 10 to 80 euro cents per filter paper sample, depending on quality grade and diameter) but the better choice with respect to the stable preservation of SARS-CoV-2 RNA on a dry matrix.”

Consider breaking this single sentence into multiple sentences for readability.