Paperonium

← Back to the plain-language case

Abstract reverses the direction of the Alpha versus B.1.177 viral-burden difference reported in Table 1

PLoS Pathog (2023) · doi:10.1371/journal.ppat.1011461 · PMC10449197

Summary

The Abstract states that, among unvaccinated people with no known prior exposure, viral burden was "44% lower among Alpha variant infections" than among B.1.177 infections, and one Results sentence says the same. The paper's own coefficient points the other way. In Table 1, B.1.177 has a Beta score of +1.00 Ct against the Alpha reference, so B.1.177 samples had higher Ct values. Under the paper's stated rule (3 Ct = 10-fold change in viral load), Alpha burden is 10^(1/3) = 2.15 times that of B.1.177: 115% higher, or equivalently B.1.177 is 54% lower. The Results subsection heading, the sensitivity analysis and the Discussion ("115% higher among Alpha") all agree with this. The Abstract's first quantitative finding therefore has the wrong sign and the wrong size, and the publisher correction issued in October 2023 does not address it. A second sentence, in the sensitivity analysis, reverses the Delta versus Alpha direction among vaccinated people, and several smaller conversions in the text also disagree with Table 1.

What the paper reports

Abstract:

By analysing the adjusted Ct values using partial least squares regression, we found that among unvaccinated individuals with no known prior exposure, viral burden was 44% lower among Alpha variant infections, compared to those with the predecessor strain, B.1.177.

Results, subsection headed "Among unvaccinated individuals with no known prior exposure, viral burden was higher among Alpha compared to B.1.177 infections":

For samples sequenced at Oxford, Ct values in this group were higher for B.1.177 samples (Beta score = 1.00, VIP = 1.05) compared to Alpha. This corresponds to a 44% reduction in viral loads with Alpha.

Table 1, block "Variant amongst unvaccinated individuals with no known prior exposure" (columns: Oxford N | Beta score (viral load factor change*) | VIP):

Ref = Alpha (Oxford), Delta (Northumbria) | 3526

B.1.177 | 658 | 1.00 (2.15) | 1.05

Table 1 caption footnote:

*based upon a Ct value decrease of 3 being equivalent to a 10-fold increase in viral load [34].

Results, sensitivity analysis (subsection "The association between novel variants and higher viral burden is robust to assumptions about the shape of the within-host viral trajectory"):

It is noteworthy that the VIP value remained greater than unity across plausible parameter combinations, providing support for the observation that viral burden is higher in infections with the Alpha-variant compared to B.1.177, among unvaccinated individuals with no known prior exposure.

The second paragraph of the same subsection, on Delta versus Alpha among vaccinated individuals:

This analysis therefore provides support for the finding that samples with the Delta variant had a lower viral burden compared to samples with the Alpha variant among vaccinated individuals. In summary, our sensitivity analysis supports the finding that both the shift from B.1.177 to Alpha and the shift from Alpha to Delta was associated with an increase in viral burden amongst the population (mostly unvaccinated and mostly vaccinated, respectively) that prevailed at each stage.

Discussion, third paragraph:

Among individuals with no known prior immunity, viral burden was, on average, 115% higher among Alpha-variant compared to B.1.177 samples.

The same paragraph and the Abstract also give the other conversions checked below:

We found that viral burden decreased over time and was higher among older individuals (13% and 17% higher for every 10 year age increase, in the samples sequenced in Oxford and Northumbria, respectively).

Vaccination reduced viral burden, with the average reduction being 75% and 30% amongst the samples sequenced at Oxford and Northumbria, respectively.

Abstract:

Vaccination reduced viral burden by 67%, and among vaccinated individuals, viral burden was 286% higher among Delta variant, compared to Alpha variant, infections.

Results, subsection "Viral burden is higher among older individuals":

Based upon a 3-point decrease in Ct value being equivalent to a 10-fold increase in viral load (see [34]), these age effects equate to an approximate 14% and 17% increase in viral load for every 10 year age increase, respectively.

What the paper's own numbers show

Method. A Beta score is the difference in (epidemiologically adjusted) Ct between a category and its reference. The paper converts it to a viral-load factor with the rule in the Table 1 footnote and the Results: a Ct decrease of 3 is a 10-fold increase, so factor (category / reference) = 10^(−Beta/3). I applied this rule to all 27 Beta scores in Table 1 and compared the result with the factor printed beside each one. I then converted every percentage in the text back to the Beta it implies. The PLS model itself cannot be re-run: the authors' public repository holds MATLAB code but no Ct data. Every number below is printed by recompute.py.txt.

Quantity Location Reported Recomputed from the paper's numbers
Ct(B.1.177) − Ct(Alpha), unvaccinated, Oxford Table 1; Results Beta = 1.00 ("Ct values ... were higher for B.1.177") B.1.177 has higher Ct, so lower viral burden
Viral load, Alpha / B.1.177 Table 1 factor 2.15 10^(1.00/3) = 2.154
Alpha relative to B.1.177 Discussion "115% higher among Alpha" +115.4% (matches)
Alpha relative to B.1.177 Abstract; Results sentence "44% lower among Alpha"; "44% reduction in viral loads with Alpha" +115.4%: opposite direction
B.1.177 relative to Alpha (the reverse reading) — −53.6%, not −44%
Beta implied by "44% lower among Alpha" derived — 3·log10(0.56) = −0.76, opposite sign to the printed +1.00
Table 1 factors that follow 10^(−Beta/3) Table 1, all rows 27 printed factors 26 of 27 match to two decimals; the exception is this B.1.177 row, where 0.46 is expected and 2.15 = 1/0.464 is printed

The other conversions in the text were checked the same way. These are smaller and do not change a conclusion:

Quantity Location Reported Recomputed
Vaccinated vs no known prior exposure, Oxford (Beta 1.44) Table 1; Abstract; Results; Discussion 0.33; "67%" reduction (Abstract, Results); "75%" (Discussion) factor 0.331, a 66.9% reduction; the Discussion's 75% does not match
Vaccinated vs no known prior exposure, Northumbria (Beta 0.34) Table 1; Discussion 0.77; "30%" reduction factor 0.770, a 23.0% reduction
Delta vs Alpha among vaccinated, Oxford (Beta −1.37) Table 1; Abstract; Results; Discussion 2.86; "286% higher" factor 2.862, which is 186% higher (286% of Alpha's level)
Age per 10 years, Oxford (Beta −0.016 per year) Table 1; Results; Discussion 1.13; "14%" (Results); "13%" (Discussion) factor 1.131, +13.1%
Age per 10 years, Northumbria (Beta −0.021 per year) Table 1; Abstract; Results; Discussion 1.17; "17%" factor 1.175, +17.5% (consistent)
Direction of Delta vs Alpha among vaccinated, Oxford (Beta −1.37) Results, sensitivity-analysis subsection, second paragraph "samples with the Delta variant had a lower viral burden compared to samples with the Alpha variant among vaccinated individuals" factor 2.862: Delta higher. Table 1, the Abstract ("286% higher among Delta"), the Results subsection on vaccinated individuals, the Discussion and the next sentence of the same paragraph ("the shift from Alpha to Delta was associated with an increase in viral burden") all say higher

Figure

Why it matters

The Alpha versus B.1.177 comparison is the first quantitative finding in the Abstract, which is the part of the paper most readers and indexing services see. As printed, it tells readers that the Alpha variant was associated with lower within-host viral burden than its predecessor among unvaccinated people. The paper's coefficient, its Table 1 factor, its section heading, its sensitivity analysis and its Discussion all report the opposite: Alpha was associated with about 2.15-fold (115%) higher burden. A reader who relies on the Abstract, or who cites the Results sentence, would carry forward the reversed direction and a magnitude (44%) that no conversion of the published coefficient under the paper's rule produces.

The paper's broader narrative, that each new variant brought higher viral burden in the population then prevailing (B.1.177 to Alpha among the mostly unvaccinated, Alpha to Delta among the mostly vaccinated), is supported by its own analysis and survives once the Abstract is corrected. The Abstract's closing summary ("is influenced by the interplay of vaccination status and viral variant") states no direction and is not affected. The secondary discrepancies matter less. The sensitivity-analysis sentence that Delta had "a lower viral burden" than Alpha among vaccinated people contradicts the summary sentence that immediately follows it and every other statement of that result, so a careful reader can resolve it, but a reader quoting that sentence alone would carry the reversed direction. In addition, "286% higher" for a 2.86-fold factor overstates the Delta effect by 100 percentage points in the Abstract, and the Discussion's vaccination reductions (75% and 30%) are larger than the Table 1 values (67% and 23%).

Explanations considered

Status

Suggested correction

  1. Abstract: replace "viral burden was 44% lower among Alpha variant infections, compared to those with the predecessor strain, B.1.177" with a statement consistent with Table 1, for example "viral burden was 115% higher (2.15-fold) among Alpha variant infections, compared to those with the predecessor strain, B.1.177".
  2. Results: change "This corresponds to a 44% reduction in viral loads with Alpha" to "This corresponds to a 115% increase in viral loads with Alpha (equivalently, a 54% reduction with B.1.177)".
  3. Table 1: print the B.1.177 factor in the same orientation as the other rows (0.46, B.1.177 relative to Alpha), or footnote that 2.15 is Alpha relative to B.1.177.
  4. Abstract, Results and Discussion: describe the Delta versus Alpha effect as "186% higher" or "2.86-fold", not "286% higher".
  5. Results, sensitivity-analysis subsection, second paragraph: change "samples with the Delta variant had a lower viral burden compared to samples with the Alpha variant among vaccinated individuals" to "... had a higher viral burden ...".
  6. Discussion: change the vaccination reductions from 75% and 30% to 67% and 23%, and align the Oxford age effect (14% in the Results, 13% in the Discussion; 13.1% from Table 1).