Paperonium

← Back to the plain-language case

Results text assigns each group's cAMP mean to the other group, contradicting its own "higher" claim

PLoS Pathog (2013) · doi:10.1371/journal.ppat.1003319 · PMC3635970

Summary

The Results sentence for Figure 6B says that CD4+ T cells from HIV+ patients have "higher" intracellular cAMP than controls. The means it gives point the other way: 548.3 ± 9.1 fmol/million cells for HIV+ and 761.6 ± 29.4 for controls, which puts HIV+ 28% lower. Reading the six points per group off the published Figure 6B gives 547.8 ± 9.1 for the HIV− squares and 760.5 ± 29.4 for the HIV+ circles. The printed numbers are the figure's numbers with the two group labels swapped, so the figure and the paper's direction of effect (HIV+ higher) stand, and the sentence as printed contradicts itself. Separately, the Figure 6B legend gives 8 controls where the text gives 6, and the figure and the printed SEM both fit 6. The higher-cAMP result is one step of the mechanism the abstract and conclusions put forward, so the sentence should be corrected.

What the paper reports

All quotations were checked verbatim against the PMC-archived text with src/show.py.

Results, section "In vitro activated CD4+ cells from HIV+ patients were not able to demethylate il-2 CpG site 1 ...", first paragraph:

Ex vivo CD4+ T cells from HIV+ patients (n = 6) exhibit also higher levels of intra-cytoplasmic cAMP (mean 548.3±9.1 fmol/million cells) as compared to controls (n = 6; mean 761.6±29.4 fmol/million cells) (P = 0.002; Figure 6B).

Same paragraph, last sentence:

All together these results demonstrate a constitutive high expression of A2AR and cAMP resulting in a clear inhibitory effect on CpG demethylation accompanied by the lack of IL-2 production in HIV+ART- upon anti-CD3/CD28 activation.

Figure 6 title and legend, panel B:

High levels of A2AR and endogenous cAMP in CD4+ T cells from HIV infected patients and lack of IL-2 production.

(B) CD4+ T cells were purified from the blood of ART-naive HIV-infected patients (n = 6) and healthy controls (n = 8). Intra-cellular cAMP is measured using the cAMP direct enzyme immunoassay from GE healthcare Biosciences.

Abstract:

We show here that CD4+ T-cells from HIV-1 infected individuals express high levels of A2AR and intracellular cAMP.

Author Summary:

We found that T cells from HIV patients have a higher expression on A2AR as well as intra-cellular cAMP and a lesser capacity to produce IL-2 upon stimulation than healthy subjects.

Introduction, third paragraph:

We also found that CD4+ T cells from HIV infected patients express high level of cytoplasmic cAMP

Discussion, third paragraph:

In accordance with this, we show significant increased intracellular cAMP levels in CD4+ T cells from HIV+ patients compared to healthy controls.

Discussion, concluding paragraph:

This defect is caused by increased intracellular cAMP, due in part to increased hydrolysis of inflammatory ATP by both expanded Treg/CD39+ and increased A2AR expression levels.

Materials and Methods, "Statistical analysis":

The non-parametric Mann-Whitney U, Fisher's exact and paired T tests were used for statistical analyses (GraphPad Prism 5.0 statistical software).

What the paper's own numbers show

Method. First, I compared the two means printed in the Results sentence. Second, I read the individual points of Figure 6B off the publisher's image (PLOS "large" figure g006, 2776 × 2063 px). I calibrated the y-axis on its five tick marks (500 to 900; one pixel = 0.67 fmol/million cells; tick residuals ≤ 0.13). I located each marker's centroid after a morphological opening that removes the thin mean and error-bar lines, and I located the drawn mean bars and error-bar caps from their pixel rows. Third, I recomputed the mean, SD and SEM of each group of points, and the Mann-Whitney P values that the possible sample sizes allow. The pixel positions and every calculation are in recompute.py.txt, which prints every number below. One marker (a square behind the HIV− mean bar) is partly hidden; its centroid is taken from its visible rows. Read-off values are accurate to about ±1 fmol/million cells.

Quantity Location Reported Recomputed from the paper's own numbers
Direction of the difference Results sentence "higher" in HIV+ 548.3 − 761.6 = −213.3; HIV+/control = 0.720 (HIV+ 28.0% lower)
HIV+ mean ± SEM Results sentence 548.3 ± 9.1 Figure 6B HIV+ circles: 760.5 ± 29.4 (n = 6; drawn bar 761.9 ± 30.1)
Control mean ± SEM Results sentence 761.6 ± 29.4 Figure 6B HIV− squares: 547.8 ± 9.1 (n = 6; drawn bar 548.0 ± 8.9)
Individual values, HIV− Figure 6B not stated 574.8, 572.8, 553.0, 532.8, 528.8, 524.8 (SD 22.3)
Individual values, HIV+ Figure 6B not stated 818.7, 808.9, 792.9, 768.9, 750.8, 622.8 (SD 72.0)
Control n Results text vs Figure 6B legend 6 vs 8 6 squares visible; SD/√6 of those six = 9.12, matching the printed ±9.1
SEM if two hidden controls lay at the mean (n = 8) hypothetical — 6.68 (would need SD 25.7 over 8 values to give 9.1)
P value, 6 vs 6, complete separation Results (P = 0.002) 0.002 exact Mann-Whitney 2/924 = 0.00216; normal approximation 0.0039
P value, 6 vs 8, complete separation if the legend's n were right — exact 2/3003 = 0.00067; normal approximation 0.0019
Exact P, 6 vs 8, U = 1 and U = 2 if not completely separated — 0.00133 and 0.00266
Size of the difference, corrected labels from printed means and SEMs — +213.3 (HIV+ 38.9% higher); Welch t = 6.93, P = 4.6 × 10-4

Taken at its word, the Results sentence reports HIV+ cAMP below the control level, the opposite of what the same sentence says. Each printed mean and SEM matches, to within read-off accuracy, the mean and SEM of the other group's points in Figure 6B. The figure's HIV+ circles all lie above its HIV− squares, which is the direction the text states. Because the twelve points are completely separated, the exact Mann-Whitney test gives P = 0.00216, which rounds to the reported 0.002.

Figure

The printed cAMP means match Figure 6B only with the group labels swapped

Figure 1. Left: the Results sentence and the means ± SEM it prints for each group (orange). Right: the mean ± SEM of the six points per group read off the published Figure 6B (blue), computed in recompute.py.txt. The connectors show that the value printed for HIV+ matches the figure's HIV− group and the value printed for controls matches the figure's HIV+ group. Bottom: the control sample size is 6 in the text and 8 in the legend. Six control squares are visible, and the printed ±9.1 equals the SD of those six divided by √6. The diagram reproduces no part of the paper's figure; it shows only summary numbers.

On a common axis, each printed mean sits on the other group's Figure 6B mean

Figure 2. The four means ± SEM on one axis from zero, in fmol/million cells. Orange circles are printed in the Results text; blue squares are computed from the six points per group read off Figure 6B. The printed HIV+ value (548.3) coincides with the figure's control value (547.8), and the printed control value (761.6) with the figure's HIV+ value (760.5). The interval widths match in the same crossed way (±9.1 and ±29.4).

Why it matters

Explanations considered

Status

Suggested correction

In the Results, replace

Ex vivo CD4+ T cells from HIV+ patients (n = 6) exhibit also higher levels of intra-cytoplasmic cAMP (mean 548.3±9.1 fmol/million cells) as compared to controls (n = 6; mean 761.6±29.4 fmol/million cells) (P = 0.002; Figure 6B).

with

Ex vivo CD4+ T cells from HIV+ patients (n = 6) exhibit also higher levels of intra-cytoplasmic cAMP (mean 761.6±29.4 fmol/million cells) as compared to controls (n = 6; mean 548.3±9.1 fmol/million cells) (P = 0.002; Figure 6B).

The authors should also confirm the number of controls in the cAMP assay from their records and make the Figure 6B legend agree with the text. If the six plotted controls are the full set, the legend should read "healthy controls (n = 6)". It would help to state that the ± values are SEMs.