A Multicenter, Randomized, Controlled Clinical Trial of Transfusion Requirements in Critical Care
New England Journal of Medicine, 1999 · paper · Clinical trials
A randomized trial of 838 critically ill patients comparing restrictive (hemoglobin below 7) and liberal (below 10) red-cell transfusion. Thirty-day mortality was similar overall; the authors conclude restrictive transfusion is at least as effective, with a possible exception for heart disease.
8 findings
Substance ICU mortality in the restrictive group is "13.9 percent vs. 16.2 percent, P=0.29" in the text but "56 (13.4)" in Table 2 (same outcome, same P).
- p. 5 Results, Outcome Measures, right column
unit (13.9 percent vs. 16.2 percent, P=0.29)
- p. 5 Table 2, Death, ICU row
56 (13.4)
- p. 5 Results, Outcome Measures
(13.9 percent vs. 16.2 percent, P=0.29)
- p. 5 Outcome Measures, text
(13.9 percent vs. 16.2 per-
Judge's note: Checked the p5 image. 56/418 = 13.4%; 13.9% would need 58/418. The CI midpoint of 2.8 fits 13.4. The table's separate 2.3 difference is handled under C006.
Second check: another model agreed
Report a problem with this finding
Substance Table 2, ICU death row: the absolute difference is printed "2.3", but the row's own values 56 (13.4) vs 68 (16.2) give 2.8, and the CI −2.0 to 7.6 is centred on 2.8.
- p. 5 Table 2, Death, ICU row, absolute difference
2.3
Judge's note: Checked the p5 image. The other death rows match a simple subtraction (30-day 23.3−18.7 = 4.6 ≈ 4.7 from the unrounded values; hospital 28.1−22.2 = 5.9 ≈ 5.8). The ICU row is off by 0.5, which rounding cannot explain.
Second check: another model agreed
Report a problem with this finding
Substance Table 2, adjusted multiple-organ-dysfunction score: difference "1.1" with 95% CI "0.8 to 2.2". The interval is centred on 1.5, not on 1.1, and is narrower than the reported SDs and Ns allow.
- p. 5 Table 2, Multiple-organ-dysfunction score, Adjusted score row
0.8 to 2.2
- p. 5 Table 2, Adjusted score
1.1 0.8 to 2.2
- p. 5 Table 2, Adjusted score
11.8±7.7 1.1 0.8 to 2.2
Only the authors can say what is right here.
Judge's note: Checked the p5 image. The Methods name an independent t-test. With SDs 7.5 and 7.7 and n = 418 and 420, SE ≈ 0.525, giving about 0.07 to 2.13 (consistent with P=0.03). The neighbouring rows (−0.1 to 1.1 for 0.5; 0.1 to 2.0 for 1.0) are symmetric. The lower bound '0.8' looks like a misprint (perhaps 0.1); only the authors can confirm.
Second check: another model agreed
Report a problem with this finding
Substance Table 2, organs failing = 2, liberal group: "108 (26.0)", but 108/420 = 25.7%.
- p. 5 Table 2, No. of organs failing, row 2, liberal column
108 (26.0)
- p. 5 Table 2, No. of organs failing, 2
109 (26.1) 108 (26.0)
Judge's note: Checked the p5 image. The liberal counts 82+149+108+63+18 sum to 420, so the denominator is 420. The other cells match n/N: 82/420 = 19.5, 149/420 = 35.5, 63/420 = 15.0, 18/420 = 4.3. The printed percentages sum to 100.3.
Second check: another model agreed
Report a problem with this finding
Substance Text gives ICU-stay mortality as "13.9 percent vs. 16.2 percent, P=0.29" but Table 2 prints ICU deaths "56 (13.4)" (and 56/418 = 13.4%); separately, the Table 4 footnote counts "134 patients who died in the intensive care unit: 61 in the restrictive-strategy group and 73", which exceeds the Table 2 ICU deaths of 56 and 68.
- p. 8 Table 4, Complications, Shock row
8 (15)
- p. 8 Table 4, footnote †
61 in the restrictive-strategy group and 73 in the liberal-strategy
- p. 8 Table 4 footnote
134 patients who died in the intensive care unit: 61 in the restrictive-strategy group and 73
- p. 5 Table 2
56 (13.4)
Only the authors can say what is right here.
Judge's note: Checked p.5 (Table 2 and the Results text) and p.8 (Table 4 and its footnote). The restrictive ICU death rate is 13.4% in Table 2 and 13.9% in the text. The liberal figure (68/420 = 16.2%) and the P value match in both places, so this is one quantity printed two ways. Separately, Table 4 covers the 176 patients who died within 30 days (78+98), and its footnote says 61 restrictive and 73 liberal of them died in the ICU. That is more than the 56 and 68 deaths that Table 2 gives for the whole ICU stay, even though ICU deaths within 30 days should be a subset of all ICU deaths. I found nothing in the paper that explains this. The shock-row percentage point (reviewer note 1) is adjudicated under C021.
Second check: another model agreed
Report a problem with this finding
Substance The liberal group trauma subgroup is stated as "100 had a traumatic injury", but its 30-day mortality is "8.8 percent", which cannot be obtained with a denominator of 100.
- p. 6 Subgroup Analyses, right column
175 had cardiac disease, 100 had a traumatic injury
- p. 7 Results continuation, left column
(10.0 percent and 8.8 percent, respectively; P=0.81)
Only the authors can say what is right here.
Judge's note: Checked p6 and p7. With n=100 only whole percentages are possible. 8.8% fits, for example, 7/80, and 80 is the liberal primary-trauma count in Table 1. The restrictive 10.0% fits 10/100. Either the count or the rate is misprinted.
Second check: another model agreed
Report a problem with this finding
Substance Table 4, primary diagnosis at death, liberal group (N=98): 15 + 44 + 9 + 6 + "25 (26)" = 99, not 98.
- p. 8 Table 4, Primary diagnosis at death, liberal column
25 (26)
Only the authors can say what is right here.
Judge's note: Checked the p8 image. The restrictive column sums to 78 (15+28+11+9+15), and the liberal organs-failing block sums to 98 (13+31+30+24+0). The categories are mutually exclusive, so one liberal cell is off by one; which cell cannot be determined from the files.
Second check: another model agreed
Report a problem with this finding
Substance Table 4, liberal-strategy column (N=98): primary diagnosis at death counts "15 (15)", "44 (45)", "9 (9)", "6 (6)", "25 (26)" sum to 99, not 98.
- p. 8 Table 4, Primary diagnosis at death
15 (15)
Only the authors can say what is right here.
Judge's note: Checked the zoomed crop of the p.8 Table 4. The restrictive column sums to 78, which matches its N. The liberal column sums to 99 against N=98, and its percentages also sum to 101. Primary diagnosis is a single category per patient, and the paper says nothing about double counting. Only the authors can say which count is off by one.
Second check: another model agreed
Report a problem with this finding
11 minor (typos and the like; most checked by the judge only)
Wording Hemoglobin entry criterion is strict in the Abstract ("concentrations of less than 9.0 g per deciliter") but inclusive in the Methods ("a hemoglobin concentration of 9.0 g per deciliter or less").
- p. 1 Abstract, Methods
concentrations of less than 9.0 g per deciliter within
- p. 2 Methods, Study Population
had a hemoglobin concentration of 9.0 g per deciliter or less within 72 hours
Only the authors can say what is right here.
Judge's note: Checked both crops. The Abstract says '<9.0' and the Methods say '≤9.0'. No other text resolves which one applies. Only the boundary value of exactly 9.0 is affected.
Second check: another model agreed
Report a problem with this finding
Wording The primary outcome is defined as "death from all causes in the 30 days after randomization" (Methods) but as "all causes in the 30 days after admission to the intensive care unit" (Results), and the Figure 2 caption also says "after Admission to the Intensive Care Unit".
- p. 2 Methods, Outcome Measures
death from all causes in the 30 days after randomization
- p. 5 Results, Outcome Measures
all causes in the 30 days after admission to the in- tensive care unit
- p. 5 Results, Outcome Measures
30 days after admission to the in-
Only the authors can say what is right here.
Judge's note: Checked p2, p5 and p6. Randomization could take place up to 72 h after ICU admission (Methods), so the two time origins can differ. The paper does not say which origin was used.
Second check: another model agreed
Report a problem with this finding
Wording Liberal group infection subgroup is "104 had a severe infection or septic shock" (p6), versus Table 1 "Infection — no. (%) ... 108 (26)", while the restrictive count is 114 in both places.
- p. 6 Subgroup Analyses
104 had a severe infection or septic shock
- p. 4 Table 1, Infection
108 (26)
Only the authors can say what is right here.
Judge's note: The labels differ ('severe infection or septic shock' vs 'Infection', 222 = 114+108 as primary or secondary diagnosis on p4), so different definitions are possible. However, the identical restrictive count suggests a common definition. The liberal subgroup mortality of 29.7% fits neither 104 (31/104 = 29.8) nor 108 (32/108 = 29.6). Only the authors can say which count is right.
Second check: another model agreed
Report a problem with this finding
Wording 60-day death: "111 (26.5)" for the liberal group fits 111/419 (excluding the patient lost to follow-up), while the restrictive "95 (22.7)" fits 95/418 (not excluding its two losses), and the difference "3.7" fits 111/420.
- p. 5 Table 2, Death, 60-day row
111 (26.5)
- p. 5 Results text, right column
(22.7 percent vs. 26.5 per- cent, P=0.23)
Judge's note: Footnote † (p5) says 3 patients were lost at 60 days (2 restrictive, 1 liberal), which explains 26.5 = 111/419. But the denominators are handled inconsistently across the two columns and the difference column. This is a minor rounding-level inconsistency.
Second check: another model agreed
Report a problem with this finding
Wording Restrictive group under 55 years: "173 patients" with 30-day mortality "5.7 percent"; no whole count over 173 rounds to 5.7%.
- p. 6 Subgroup Analyses, right column
group, 173 patients were younger than 55 years
- p. 6 Subgroup Analyses, right column
(5.7 per- cent vs. 13.0 percent
Only the authors can say what is right here.
Judge's note: Checked p6. 10/173 = 5.78 (rounds to 5.8) and 9/173 = 5.20. The other subgroup rates match their stated Ns (18/207 = 8.7, 35/217 = 16.1, 21/161 = 13.0, 31/151 = 20.5, 40/175 = 22.9). Truncation would give 5.7, so this may be a 0.1-point discrepancy only.
Second check: another model agreed
Report a problem with this finding
Wording Discussion: "found that administering transfusions liberally or expectantly resulted not only in increased overall use of transfusions" is ambiguous about which arm is being compared with which.
- p. 8 Discussion, col 1, last paragraph
administering transfusions liberally or expectantly resulted not only in increased overall use of transfusions
Judge's note: Read on p.8, left column. Ref 23 compares early transfusion with a control arm. 'Liberally or expectantly' can be read as two synonyms (transfusing in anticipation of need). But 'expectant' in medicine usually means watchful waiting, which would be the comparator arm. The sentence never names the comparator, so the wording is ambiguous rather than clearly wrong.
Second check: another model agreed
Report a problem with this finding
Wording Table 4 shock row "8 (15)" and "12 (18)" does not match the footnote denominators of 61 and 73 (8/61 = 13%, 12/73 = 16%).
- p. 8 Table 4, Complications, Shock
8 (15) 12 (18)
Only the authors can say what is right here.
Judge's note: Checked the zoomed crop of Table 4 (p.8). Most complication and intervention rows fit 61/73 (e.g., cardiac 25/61 = 41%, any type 40/61 = 66%, dialysis 10/73 = 14%), but the shock row does not. The percentages would fit denominators of about 53 and 67, which suggests missing data for this item, but the table does not state this. It is a query, not a clear error.
Second check: another model agreed
Report a problem with this finding
Wording Table 4 antibiotics row "52 (83)" and "64 (85)" does not match the footnote denominators of 61 and 73 (52/61 = 85%, 64/73 = 88%); the "Failure of ≥3 organs" row "23 (37)" and "24 (32)" is also off (38%, 33%).
- p. 8 Table 4, Interventions
52 (83) 64 (85)
Only the authors can say what is right here.
Judge's note: Checked the zoomed crop of Table 4 (p.8). The footnote states that 61 and 73 patients had complication and intervention data. Several rows match these denominators, but the antibiotics row (and, by my own check, the ≥3 organ failures row) does not. This may reflect per-item missing data, which the table does not state.
Second check: another model agreed
Report a problem with this finding
Typo Table 4 column header reads "RESTRICTED-TRANSFUSION STRATEGY"; everywhere else the paper says "restrictive-transfusion strategy".
- p. 8 Table 4, column header
RESTRICTED-
- p. 8 Table 4 header
RESTRICTED- LIBERAL-
Judge's note: The Table 4 header on p.8 (crop C020-a) clearly reads 'Restricted-'. Table 2 and the text use 'Restrictive-'. This is an inconsistency in terminology only.
Second check: none (judge only)
Report a problem with this finding
Typo Reference 1 title reads "RCB transfusion in the ICU: is there a reason?"; the red-blood-cell abbreviation is RBC.
- p. 9 References, ref 1
RCB transfusion in the ICU
- p. 9 References, ref 1
RCB transfusion in the ICU: is there a reason?
Judge's note: Confirmed in the p.9 reference list crop. Ref 10 in the same list uses 'RBC'. 'RCB' is a transposition typo.
Second check: none (judge only)
Report a problem with this finding
Typo Appendix lists "Wellesley Hospital, Wellesley"; the city given is the hospital's own name rather than its location.
- p. 9 Appendix
Wellesley Hospital, Wellesley
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: that the Wellesley Hospital was in Toronto. The appendix on p.9 does read 'Wellesley Hospital, Wellesley'. Other entries give the city where the hospital is (e.g., 'St. Michael's Hospital, Toronto'). The files cannot confirm where the hospital was, so only the authors can confirm this.
Second check: none (judge only)