Thrombectomy 6 to 24 Hours after Stroke with a Mismatch between Deficit and Infarct
2018 · paper · Clinical trials
A randomized trial of thrombectomy plus standard care versus standard care alone in 206 stroke patients last known well 6 to 24 hours earlier, with a mismatch between deficit and infarct. Functional independence at 90 days was 49% versus 13%.
4 findings
Substance Abstract gives the functional-independence adjusted difference as "33 percentage points; 95% credible interval, 24 to 44", but Table 2 prints "33 (21–44)" and the Results text prints "95% credible interval, 21 to 44" for the same adjusted Bayesian estimate.
- p. 1 Abstract, Results
credible interval, 24 to 44
- p. 7 Results, Efficacy Outcomes, first paragraph
terval, 21 to 44
- p. 6 Table 2, Functional independence row, Adjusted Difference column
33 (21–44)
- p. 1 Abstract, Results
adjusted difference, 33 percentage points; 95% credible interval, 24 to 44
Judge's note: Checked the abstract crop on p1, the Results crop on p7 and Table 2 on p6. All three report the same adjusted Bayesian difference of 33 points for the same coprimary end point and population. The lower bound is 24 in the abstract and 21 in both the table and the text. The lower bound of 24 matches the unadjusted absolute difference CI in Table 2, 36 (24–47), so the abstract has probably mixed up the two intervals. Nothing in the paper explains the difference.
Second check: another model agreed
Report a problem with this finding
Substance Discussion says "the lower rate of treatment with intravenous alteplase (14% in our trial vs. 88% in the pooled analysis)", but Table 1 gives 13 (13) in the control group, 5 (5) in the thrombectomy group and 18/206 (9%) overall.
- p. 9 Discussion, first paragraph
(14% in our trial vs. 88% in the pooled analysis)
- p. 5 Table 1, Treatment with intravenous alteplase
5 (5) 13 (13)
- p. 5 Table 1, Treatment with intravenous alteplase
13 (13)
- p. 5 Table 1
5 (5)
Judge's note: Checked Table 1 on p5 and the Discussion on p9. The sentence compares the control groups, so the matching Table 1 value is 13/99 = 13%. No population in the paper gives 14%.
Second check: another model agreed
Report a problem with this finding
Substance Control-group 24-hour recanalization is "in 36% of the patients in the control group" in the Results text, but Table 2 prints "39 (39)". The table's own absolute difference "40 (27–52)" fits 36% better than 39%.
- p. 7 Results, Efficacy Outcomes, second paragraph
in 36% of the patients in the control group
- p. 6 Table 2, Recanalization at 24 hr row
82 (77) 39 (39) 40 (27–52) 2 (2–4)
- p. 6 Table 2, Recanalization at 24 hr
39 (39)
- p. 6 Table 2, Recanalization at 24 hr, absolute difference
40 (27–52)
Only the authors can say what is right here.
Judge's note: Checked the p7 text crop and the Table 2 row on p6. They report the same end point, group and time point, and the table gives no denominator other than N=99. 39/99 is 39.4%, giving a difference of 37.2 points against the printed 40. 36/99 would give 40.2. It is unclear whether the 36% in the text or the 39 (39) in the table is wrong, so only the authors can say.
Second check: another model agreed
Report a problem with this finding
Reference or label Table 3 row label is "Neurologic deterioration at 24 hr‡", but footnote ‡ and the Methods define it as "an increase in the NIHSS score of 4 or more points within 5 days after stroke".
- p. 9 Table 3, row label
Neurologic deterioration at 24 hr‡
- p. 9 Table 3, footnote ‡
an increase in the NIHSS score of 4 or more points within 5 days after stroke
- p. 4 Methods, End Points (safety)
in the NIHSS score of ≥4 points within 5 days
- p. 4 Methods, End Points
within 5 days after stroke
Judge's note: Seen on the p9 image. The row label's time window conflicts with the row's own footnote and with the Methods on p4. The 24-hour window belongs to the symptomatic ICH row directly above.
Second check: another model agreed
Report a problem with this finding
4 minor (typos and the like; most checked by the judge only)
Wording Table 2 early response row prints "51 (48) 19 (19) 29 (16–41) 3 (2–4)"; the crude risk ratio from these counts is about 2.5, which would round to 2 at the table's integer precision.
- p. 6 Table 2, Early response row, Risk Ratio column
29 (16–41) 3 (2–4)
Only the authors can say what is right here.
Judge's note: Read the row on the p6 image. The table does not say how the risk ratios were estimated; footnote ** covers only the Fisher P value. A model-based estimate could differ from the crude one, so this is a query for the authors, not a correction. The recanalization row's crude ratio (1.94) agrees with its printed value of 2.
Second check: another model agreed
Report a problem with this finding
Wording Text says "Among the patients who underwent thrombectomy, immediate reperfusion was achieved in 84%", but 84% is 90/107 (the whole thrombectomy group, as in Table 2), whereas 105 patients underwent thrombectomy (90/105 = 86%).
- p. 7 Results, Efficacy Outcomes paragraph 2
Among the patients who underwent throm-bectomy, immediate reperfusion was achieved in 84%
- p. 6 Table 2, mTICI 2b or 3
90 (84)
- p. 6 Results text
Thrombectomy was performed in 105 of the 107 patients
Judge's note: Table 2 on p6 prints 90 (84) with N=107. The Results on p6 say thrombectomy was performed in 105 of the 107 patients. The text's description of the denominator therefore does not match the percentage given.
Second check: another model agreed
Report a problem with this finding
Typo Table 2 row label reads "Infarct volume at 24 hour — ml††", whereas the adjacent rows use "24 hr".
- p. 6 Table 2, row label
Infarct volume at 24 hour — ml††
- p. 6 Table 2, row label
Infarct volume at 24 hour — ml
Judge's note: Seen on the p6 image. This is an inconsistent unit abbreviation and is cosmetic only.
Second check: none (judge only)
Report a problem with this finding
Typo Dr. Levy's disclosure reads "consulting fees from Pul-sar Vasulcar" (misspelling of "Vascular").
- p. 10 Disclosures, Dr. Levy
Pul-sar Vasulcar
- p. 10 Disclosures, Dr. Levy
consulting fees from Pul-sar Vasulcar
Judge's note: The p10 image shows "Pulsar Vasulcar"; the same paragraph spells "Abbott Vascular" correctly. This is a letter transposition, not an extraction artefact.
Second check: none (judge only)