Effect of Dialysis Dose and Membrane Flux in Maintenance Hemodialysis
New England Journal of Medicine, 2003 · paper · Clinical trials
A randomized trial of 1846 hemodialysis patients testing standard versus high dialysis dose and low-flux versus high-flux membranes. Neither the higher dose nor the high-flux membrane significantly changed death from any cause, so no major benefit was found.
3 findings
Substance Planned design 'initial accrual period of 1.5 years, with 4 years of additional follow-up' does not match 'the planned follow-up ranged from 1 to 6.5 years'.
- p. 3 Statistical Analysis, sample-size paragraph
accrual period of 1.5 years, with 4 years of additional follow-up
- p. 3 Statistical Analysis, sample-size paragraph
the planned follow-up ranged from 1 to 6.5 years
- p. 3 Statistical Analysis
We planned for randomization of 900 patients during an initial accrual period of 1.5 years, with 4 years of additional follow-up
- p. 3 Statistical Analysis
initial accrual period of 1.5 years, with 4 years of additional follow-up
Judge's note: Read p.3 crops. The longest planned follow-up is accrual plus additional follow-up: 1.5 + 4 = 5.5 years, not 6.5. The minimum of 1 year fits replacement 'until the final year'. A 6.5-year maximum needs a 2.5-year accrual period, so the accrual figure is probably the error. Replacements can only shorten follow-up, so they do not explain the mismatch.
Second check: another model agreed
Report a problem with this finding
Substance Figure 1 numbers at risk at 60 months total 149 + 166 = 315 in Panel A (dose) but 149 + 160 = 309 in Panel B (flux).
- p. 6 Figure 1 Panel A, No. at risk, last column
149 166
- p. 6 Figure 1 Panel B, No. at risk, last column
149 160
Only the authors can say what is right here.
Judge's note: Read the p.6 image. Both panels split the same patients in the factorial design, so the combined number at risk at each time point should match. Totals match at 6, 18, 24, 42 and 48 months. They differ by 1 or 2 at 12 (1512 vs 1511), 30 (921 vs 919), 36 (781 vs 782) and 54 months (416 vs 415), and by 6 at 60 months. The paper gives no explanation, and only the authors can say which value is right.
Second check: another model agreed
Report a problem with this finding
Reference or label Table 1 footnote: 'to convert values for total cholesterol to micromoles per liter, multiply by 0.02586'; this factor converts mg/dl to millimoles per liter.
- p. 3 Table 1 footnote *
cholesterol to micromoles per liter, multiply by 0.02586
Judge's note: The p.3 image reads 'micromoles' for cholesterol. Cholesterol's molar mass is about 386.65 g/mol, so mg/dl × 0.02586 gives mmol/L; micromoles per liter would need a factor of 25.86. The creatinine factor in the same sentence (88.4 to µmol/L) is correct. The data are fine; only the unit label is wrong.
Second check: another model agreed
Report a problem with this finding
3 minor (typos and the like; most checked by the judge only)
Wording Relative risk reductions, (1 − RR) × 100, are called '4 percentage points lower' in the text but '4 percent lower' in the Figure 1 legend.
- p. 5 Results, Death from Any Cause
had a risk of death that was 4 percentage points lower
- p. 6 Figure 1 legend
mortality in the high-dose group was 4 percent lower
- p. 8 Figure 2 legend
subtraction of the relative risk from 1 followed by multiplication by 100
- p. 5 Results, Death from Any Cause
4 percentage points
Judge's note: The Figure 2 legend (p.8) defines the text's reductions as (1 − RR) × 100, so they are relative reductions. 'Percentage points' usually means an absolute difference. The Results and Discussion use 'percentage points', while the Figure 1 legend (p.6) uses 'percent' for the same numbers. The numbers themselves are consistent; only the wording is inconsistent.
Second check: another model agreed
Report a problem with this finding
Typo Appendix lists the Washington University investigator as 'J. Del-maz', while the byline and disclosures spell the name Delmez.
- p. 9 Appendix, Washington University
Washington University — J. Del-
- p. 1 Author byline
James A. Delmez
Judge's note: The p.9 image shows 'J. Del-' at the foot of the left column and 'maz, D. Coyne' at the top of the right column. The disclosures on the same page say 'Dr. Delmez', and the byline gives James A. Delmez. This is a misspelling.
Second check: none (judge only)
Report a problem with this finding
Typo Table 3 footnote ‡ reads 'Data includes 494 decreases…', while the other footnotes read 'Data include'.
- p. 5 Table 3 footnote ‡
Data includes 494 decreases of more than 15 percent in the albumin level
Judge's note: The p.5 image shows 'Data includes' in footnote ‡ and 'Data include' in footnotes *, †, ¶ and ||. This is a cosmetic grammatical inconsistency.
Second check: none (judge only)