The Effect of Cardiac Resynchronization on Morbidity and Mortality in Heart Failure
Congestive Heart Failure, 2005 · paper · Clinical trials
A randomized trial of 813 patients with severe heart failure and cardiac dyssynchrony, comparing medical therapy alone with added cardiac resynchronization. Over 29.4 months, death or major cardiovascular hospitalization occurred in 39% versus 55%, and deaths in 20% versus 30%.
4 findings
Substance Figure 2 splits systolic blood pressure at "<117 mm Hg" / "≥117 mm Hg", described in the caption as the study median, but Table 1 gives a median of 110 mm Hg in both groups.
- p. 9 Figure 2, Systolic blood pressure rows
<117 mm Hg
- p. 5 Table 1, Systolic blood pressure median
110
- p. 8 Figure 2 caption
divided according to the median value in the study popu-
Only the authors can say what is right here.
Judge's note: Page 8 caption: the systolic blood pressure subgroups are 'divided according to the median value in the study population'. Page 5, Table 1: median 110 (IQR 100–125) in both arms. If each arm's median is 110, the pooled median must also be 110, so a 401/402 split at 117 cannot be a median split of these data. The paper does not say why the two values differ; only the authors can say which is correct.
Second check: another model agreed
Report a problem with this finding
Substance In Table 2, the NYHA-class difference is printed "0.6 (0.4 to 0.7)", although footnote ‡ defines differences as CRT vs medical therapy and the means are 2.1 (CRT) and 2.7 (medical). The difference should be negative.
- p. 7 Table 2, NYHA class row
0.6 (0.4 to 0.7)
- p. 7 Table 2 footnote ‡
The difference shown is for the cardiac-resynchronization group as compared with the medical-therapy group.
Judge's note: Page 7 image: NYHA class is 2.7±0.9 with medical therapy and 2.1±1.0 with CRT. The difference is printed positive, while the Minnesota row (31 vs 40, printed −10) and the EQ-5D row (0.70 vs 0.63, printed +0.08) both follow footnote ‡'s CRT-minus-medical direction. The NYHA row's sign is inconsistent with that stated convention.
Second check: another model agreed
Report a problem with this finding
Reference or label First-page footnote reads "This article was published at www.nejm.org on March 7, 2004." The year conflicts with the article's own timeline and should evidently be 2005.
- p. 1 Front-page sidebar
org on March 7, 2004.
- p. 3 Follow-up section
as planned on September 30, 2004
- p. 3 Methods, Follow-up
On February 24, 2005, after this article had been submitted for publication
- p. 1 Footnote on first page
This article was published at www.nejm. org on March 7, 2004.
Judge's note: Page 1 image: the footnote shows 'March 7, 2004' just above 'N Engl J Med 2005;352:1539-49' and the 2005 copyright line. Page 3 image: the study was concluded 'as planned on September 30, 2004', and the DSMB statement of 'February 24, 2005, after this article had been submitted'. An online publication in March 2004 cannot fit that timeline, so the printed year is wrong. The likely fix is 2005, consistent with the April 14, 2005 issue.
Second check: another model agreed
Report a problem with this finding
Reference or label The Figure 2 NT-BNP cut-off is labelled "<214.5 pg/ml" / "≥214.5 pg/ml", but 214.5 matches the median in pmol/liter (about 1814 pg/ml). The unit label appears wrong.
- p. 9 Figure 2, NT-BNP subgroup rows
<214.5 pg/ml
- p. 5 Table 1, NT-proBNP medians
1806
- p. 8 Figure 2 caption
divide by 8.457
Judge's note: Page 9 image: the split is 122/366 vs 224/366, i.e. equal halves. Table 1 on page 5 gives median NT-proBNP of 1806 and 1920 pg/ml, with lower quartiles of 719 and 744, so a pooled median of 214.5 pg/ml is impossible. 214.5 × 8.457 = 1814 pg/ml, which lies between the two arm medians. The value is therefore in pmol/liter despite the pg/ml label.
Second check: another model agreed
Report a problem with this finding
3 minor (typos and the like; most checked by the judge only)
Wording The power statement, "statistical power of 80 percent to identify a 14 percent relative reduction ... given a conventional one-sided α value of 0.025 and a predicted number of 300 events", does not reproduce under the standard event-based calculation.
- p. 4 Statistical analysis
of 80 percent to iden-
- p. 4 Statistical analysis
tify a 14 percent relative reduction or a 5.7 percent-
- p. 4 Methods, Statistical analysis
of 80 percent to iden-tify a 14 percent relative reduction or a 5.7 percent-age point reduction
Only the authors can say what is right here.
Judge's note: Page 4 image: the sentence reads as quoted. Under Schoenfeld's formula with 1:1 allocation, a one-sided alpha of 0.025 and 80% power, detecting HR 0.86 (|ln 0.86| = 0.1508) needs about 1380 events, not 300. With 300 events, the detectable HR is about 0.72. The cited power method (ref 17, Shih) and the design paper (ref 11) are not in the files, so different assumptions may lie behind the figure. This is a query to the authors, not a confirmed error.
Second check: another model agreed
Report a problem with this finding
Wording The Discussion says "The benefits were similar among patients with ischemic heart disease and patients without ischemic heart disease", but Figure 2 shows a subgroup split by dilated cardiomyopathy, not by ischemic heart disease.
- p. 8 Discussion, first paragraph
The benefits were similar among patients with ischemic heart disease and pa-
- p. 9 Figure 2
Dilated cardiomyopathy
Only the authors can say what is right here.
Judge's note: Page 8 Discussion and page 9 Figure 2: the only aetiology subgroup is 'Dilated cardiomyopathy No 238/443 HR 0.68 / Yes 145/370 HR 0.51'. Table 1 (page 5) lists ischemic disease (309 patients) and other causes (134) separately; together these make the 443 non-dilated patients. No ischemic versus non-ischemic result is displayed, so the statement has no direct support among the results shown.
Second check: another model agreed
Report a problem with this finding
Typo In Table 2, the Minnesota score difference is printed "–10 (–8 to –12)", with the interval bounds in reverse order (lower bound −12 should come first).
- p. 7 Table 2, Minnesota Living with Heart Failure row
–10 (–8 to –12)
- p. 10 Table 3, interventricular delay row (for comparison)
–21 (–25 to –18)
Judge's note: Page 7 image shows '–10 (–8 to –12)'. Every other interval in Table 2, and in Table 3 on page 10 (e.g. '–21 (–25 to –18)'), runs from the lower to the upper bound. The values are plausible and only the order is reversed, so this is cosmetic.
Second check: none (judge only)