Apixaban versus Warfarin in Patients with Atrial Fibrillation
2011 · paper · Clinical trials
A double-blind trial comparing apixaban with warfarin in 18,201 patients with atrial fibrillation. Apixaban was superior in preventing stroke or systemic embolism, caused less major bleeding, and was associated with lower death from any cause.
1 finding
Wording "Data on vital status at the end of the trial were missing for 380 patients (2.1%)", but the reasons then given (withdrawal of consent in 92 and 107 patients, loss to follow-up in 35 and 34) account for only 268 patients.
- p. 6 Results, Patients and Follow-up, second paragraph
missing for 380 patients (2.1%)
- p. 6 Results, Patients and Follow-up, second paragraph
consent in the case of 92 patients in the apixaban group (1.0%) and 107 patients in the warfarin group (1.2%)
- p. 6 Results, Patients and Follow-up, second paragraph
the case of 35 patients in the apixaban group (0.4%) and 34 in the warfarin group (0.4%)
- p. 6 Patients and Follow-up
Data on vital status at the end of the trial were missing for 380 patients (2.1%)
Only the authors can say what is right here.
Judge's note: Read on the p6 image. 92 + 107 + 35 + 34 = 268, leaving 112 of the 380 unaccounted for. Each printed percentage matches its own count (380/18,201 = 2.1%; 92/9120 = 1.0%; 107/9081 = 1.2%; 35/9120 and 34/9081 = 0.4%). The sentence "The absence of data on vital status was due to ... and was due to ..." reads as exhaustive. I searched the article text and found no third category or other explanation. Whether the 380 or the breakdown is off, or a reason is unstated, only the authors can say.
Second check: another model agreed
Report a problem with this finding
6 minor (typos and the like; most checked by the judge only)
Wording Upper 95% CI limit for death from any cause is printed as "0.80 to 0.99; P = 0.047" in the abstract and Results text but as "0.89 (0.80–0.998)" in Table 2.
- p. 1 Abstract
0.80 to 0.99; P = 0.047
- p. 5 Table 2, death from any cause
0.89 (0.80–0.998)
Judge's note: Checked the abstract on the p1 image, Table 2 on the p5 image and the Results text on the p7 image. The table prints three decimals (0.998) and the text two (0.99). 0.998 rounds to 1.00, so the text value is a truncation, presumably to keep the bound below 1 in line with P = 0.047. Both forms are defensible, so this is a minor consistency point and not an error in the data.
Second check: another model agreed
Report a problem with this finding
Wording Table 1 gives "Heart failure or reduced left ventricular ejection fraction" as "3235 (35.5)" and "3216 (35.4)" (total 6451), but Figure 2A gives the Heart failure "Yes" subgroup as "5,541", with the legend defining heart failure as "symptomatic heart failure or a left ventricular ejection fraction of 40% or less".
- p. 4 Table 1, Qualifying risk factors
3235 (35.5)
- p. 4 Table 1, Qualifying risk factors
3216 (35.4)
- p. 8 Figure 2A, Heart failure, Yes
5,541
- p. 8 Figure 2 legend
Heart failure was defined as symptomatic heart failure or a left ven-
Only the authors can say what is right here.
Judge's note: Checked Table 1 on the p4 image and Figure 2A and its legend on the p8 image. 3235 + 3216 = 6451 against 5,541 in the figure, a gap of 910. The figure is internally complete (5,541 + 12,660 = 18,201). Other Figure 2A subgroups I checked reproduce the Table 1 sums exactly: prior VKA 10,401, age 75 or over 5,678, diabetes 4,547. The paper gives no statement explaining a different definition; the Methods entry criterion adds "within the previous 3 months", but that qualifier appears in neither the table row nor the legend. A definitional difference between the qualifying risk factor and the subgroup variable is plausible, so only the authors can say which applies.
Second check: another model agreed
Report a problem with this finding
Wording The Methods name the region stratum "geographic region (North America, South America, Europe, or Asian Pacific)", while Table 1 and Figure 2A and 2B label the same four-level variable "Latin America".
- p. 4 Statistical Analysis, Cox model strata
geographic region (North America, South Amer-
- p. 4 Table 1, Region
Latin America
- p. 8 Figure 2A, Geographic region
Latin America
Judge's note: Checked the Statistical Analysis text and Table 1 on the p4 image and Figure 2 on the p8 and p9 images. The four strata correspond one to one, so the same region carries two names. "Latin America" is used three times and fits the Mexican site in the author affiliations (p12), so the evident fix is to the Methods wording. The meaning is recoverable; this is a labelling inconsistency.
Second check: another model agreed
Report a problem with this finding
Wording The Discussion states "apixaban is also associated with a reduction in the rate of gastrointestinal bleeding", while Table 3 reports gastrointestinal bleeding in 105 patients (0.76%/yr) versus 119 (0.86%/yr), hazard ratio "0.89 (0.70–1.15)", P = 0.37.
- p. 11 Discussion, left column
apixaban is also associated with a reduction in the rate of gastrointestinal bleeding
- p. 7 Table 3, row Gastrointestinal
0.89 (0.70–1.15)
- p. 11 Discussion, para on dabigatran
apixaban is also associated with a reduction in the rate of gastro-
- p. 11 Discussion, para 3
associated with a reduction in the rate of gastro-
Only the authors can say what is right here.
Judge's note: Checked the Table 3 row on the p7 image and the Discussion sentence on the p11 image. The rate is numerically lower with apixaban, but the table itself reports the difference as not significant (P = 0.37 as printed). For the analogous myocardial infarction result the paper adds "but the difference was not significant". The unqualified word "reduction" therefore overstates what Table 3 shows; the authors should confirm the intended wording.
Second check: another model agreed
Report a problem with this finding
Typo Misspelling in Dr. Goto's disclosure: "Otuska, Eisai, Sanofi-Aventis" (the same company is spelled "Otsuka" two lines later and in Dr. Granger's disclosure).
- p. 11 Disclosures, Dr. Goto
Otuska, Eisai, Sanofi-Aventis
Judge's note: Read on the p11 image. "Otuska" appears in the grants list and "Otsuka" in the lecture-fees list of the same disclosure, and "Otsuka Pharmaceutical" under Dr. Granger. The inconsistency is internal to the page, so this is a transposition typo.
Second check: none (judge only)
Report a problem with this finding
Typo Dr. Goto's disclosure lists "AstraZeneca, Asteras, Pfizer"; the name "Astellas" appears in Dr. Granger's disclosure on the same page.
- p. 11 Disclosures, Dr. Goto
AstraZeneca, Asteras, Pfizer
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: that "Asteras" is not a distinct company and is meant to be Astellas. The p11 image confirms "Asteras" is printed in Dr. Goto's lecture-fee list and "Astellas" in Dr. Granger's grants list. It is a likely misspelling, but the files alone cannot establish that the two names refer to the same company.
Second check: none (judge only)