Base Editing of HBG1 and HBG2 Promoters for Sickle Cell Disease
New England Journal of Medicine, 2026 · paper · Clinical trials
A phase 1–2 study of base-edited stem-cell therapy risto-cel in 31 patients with sickle cell disease. It reports rapid engraftment, fetal hemoglobin above 60% of total hemoglobin, reduced sickle hemoglobin, and no severe vaso-occlusive crises after the transfusion window.
3 findings
Substance The Methods tie plerixafor dosing to body weight — "a weight-based approach (in patients with a body weight of >83 kg) or a fixed-dose approach according to weight tiers (in patients with a body weight of ≤83 kg)" — but the Results state that of the 34 patients weighing 83 kg or less, "7 received weight-based plerixafor and 27 received fixed-dose plerixafor".
- p. 4 Methods, Study Design
a weight-based approach (in patients with a body weight of >83 kg) or a fixed-dose approach according to weight tiers (in patients with a body weight of ≤83 kg)
- p. 5 Results, Treatment
34 had a body weight of 83 kg or less; among these 34 patients, 7 received weight-based plerixafor and 27 received fixed-dose plerixafor
Only the authors can say what is right here.
Judge's note: Read the Methods paragraph on the p004 image (crop C005-a) and the Treatment paragraph on the p005 image (crop C005-b / C003-b right column). The two statements cannot both hold as written: the Methods rule assigns all ≤83 kg patients to fixed dosing, yet 7 of the 34 such patients received weight-based dosing (7+27=34). Reference 34 compares fixed-dose with weight-based plerixafor in this same study, suggesting a change in dosing strategy during the trial, but no such amendment is described in the Methods, so the article contradicts itself as published. Only the authors can say which description is right.
Second check: another model agreed
Report a problem with this finding
Reference or label Table 2 states the dose unit with a negative exponent — "Median risto-cel dose infused (range) — ×10−6 viable CD34+ cells/kg", value 6.2 (3.2–23.4) — whereas the text gives "a median risto-cel dose of 6.2×106 viable CD34+ cells (range, 3.2×106 to 23.4×106) per kilogram" and the Methods a "minimum cell dose of 3.0×106 viable CD34+ cells per kilogram".
- p. 7 Table 2, row "Median risto-cel dose infused (range)"
Median risto-cel dose infused (range) — ×10-6 viable CD34+ cells/kg
- p. 6 Results, Treatment paragraph
median risto-cel dose of 6.2×106 viable CD34+ cells (range, 3.2×106 to 23.4×106) per kilo-gram
- p. 4 Methods, Study Design
with a minimum cell dose of 3.0×106 viable CD34+ cells per kilogram of body weight
- p. 7 Table 2
×10−6 viable
Judge's note: Read the Table 2 row on the p007 page images (crops C007-a, C007-d and C003-c); the superscript minus in "×10-6" is plainly visible on the image, so this is not an extraction artefact (extraction drops minus signs rather than inventing them). Compared with the Methods on p004 (crop C007-c) and the Results on p006 (crop C007-b), both of which use ×10^6, and with the abstract ("≥3.0×106 viable CD34+ cells per kilogram"). The tabulated values are right; the exponent sign in the unit label is wrong.
Second check: another model agreed
Report a problem with this finding
Reference or label Reference 35 reads "Coisman B, Lucini C, Ye L, et al. Robust autologous CD34+ HSPC manufacturing ... Presented at the European Hematology Association (EHA) Congress, Milan, June 12–15, 2025" but its own URL is "library.ehaweb.org/eha/2024/eha2024-congress/419566/paul.kopesky.robust.autologous.cd342B.hspc.manufacturing.with.a.closed.and.html", giving the 2024 congress and a different presenting author.
- p. 12 Reference list, reference 35
Coisman B, Lucini C, Ye L, et al. Robust autologous CD34+ HSPC manufacturing with a closed and automated process optimized for patients with sickle cell disease. Presented at the European Hematology Association (EHA) Congress, Milan, June 12–15, 2025.
- p. 12 Reference list, reference 35, URL
library.ehaweb.org/eha/2024/eha2024-congress/419566/paul.kopesky.robust.autologous.cd342B.hspc.manufacturing.with.a.closed.and.html
- p. 12 Reference list, entry 35, URL
library.ehaweb.org/eha/2024/eha2024-congress/419566/paul.kopesky.robust
- p. 4 Methods, Study Design
up to 4 days were allowed per mobilization cycle.35
Only the authors can say what is right here.
Judge's note: Read reference 35 on the p012 page image (crop C015-a) and compared it with reference 36 on the same page, which cites the same meeting ("EHA Congress, Milan, June 12–15, 2025") with the matching path "library.ehaweb.org/eha/2025/eha2025-congress/4159577/paul.kopesky.integrated..." (visible in crop C001-b). Within entry 35 the stated year (2025) and the URL year (2024) disagree, and the printed first author (Coisman B) differs from the author string in the link (paul.kopesky); the two fields of the same entry cannot both be right, and only the authors can say which. The third reviewer point in this cluster — that the Methods sentence "To alleviate the mobilization-related treatment burden, up to 4 days were allowed per mobilization cycle.35" (crop C015-d, p004) should cite the mobilization abstract (ref 34) rather than the manufacturing abstract — is not established: collection days feed the manufacturing process described in ref 35, and the entry's content cannot be checked from the files.
Second check: another model agreed
Report a problem with this finding
4 minor (typos and the like; most checked by the judge only)
Wording The Introduction states without qualification that the risto-cel editing "process generates precise and uniform edits, avoids insertions or gene rearrangements, and prevents p53 activation", while the Methods state the nickase "can create two nicks on the same DNA strand with potential generation of a 4.92-kb deletion that includes HBG2" and the Results report that deletion in patients (mean 1.4±1.5%, range undetected to 3.6, at month 3).
- p. 2 Introduction, final paragraph
cess generates precise and uniform edits, avoids insertions or gene rearrangements, and prevents p53 activation
- p. 5 Methods, End Points
nickase can create two nicks on the same DNA strand with potential generation of a 4.92-kb deletion that includes HBG2
- p. 7 Results, Editing Characteristics
HBG2 deletion (assessed as the percentage of alleles containing the deletion, as measured by ddPCR) was evaluated in blood samples obtained from 22 patients at month 3, with a mean value of 1.4±1.5% (range, undetected to 3.6)
Only the authors can say what is right here.
Judge's note: Read the Introduction sentence on p002 (crops C001-a/C003-a region), the End Points paragraph on p005 (crop C003-b) and the Results paragraph on p007 (crop C003-c). The paper itself attributes the 4.92-kb HBG2 deletion to the same two-nick mechanism and reports it in blood (mean 1.4%) and bone marrow (mean 1.8±0.5%), and the Discussion notes "the deletion was observed in less than 4% of HBG2 alleles". The blanket introductory claim about avoiding gene rearrangements is therefore not qualified anywhere in the article; wording issue for the authors, as the intended sense may be "no double-strand-break-mediated insertions or rearrangements".
Second check: another model agreed
Report a problem with this finding
Wording The Figure 1 Panel C annotation beside the normal red cell reads "HbF per cell stays above threshold for polymerization", whereas the only definition of this threshold in the paper (Fig. 2 legend) is "the protective threshold of HbF per F cell, defined as the level of HbF per F cell that inhibits deoxygenated HbS polymerization (10 pg)".
- p. 3 Figure 1, Panel C, label beside the normal red cell
HbF per cell stays above threshold for polymerization
- p. 9 Figure 2 legend
the protective threshold of HbF per F cell, defined as the level of HbF per F cell that inhibits deoxygenated HbS polymerization (10 pg)
- p. 2 Introduction
Protection against red-cell sickling is achieved under normal conditions at a concentration of 10 pg of HbF per F cell
- p. 9 Fig 2 Panel C and legend
inhibits deoxygenated HbS polymerization (10 pg)
Only the authors can say what is right here.
Judge's note: Viewed the Figure 1 Panel C crop (C004-a, p003), the Introduction sentence on p002 (crop C004-c: "Protection against red-cell sickling is achieved under normal conditions at a concentration of 10 pg of HbF per F cell") and the Figure 2 legend on the upright page image of p009. Read literally the Fig. 1 label says staying above the threshold gives polymerization, which is the opposite of the protective threshold defined in Fig. 2 and the Introduction, and the label annotates a non-sickled normal red cell. A word such as "protective" or "inhibition of" appears to be missing; wording defect only. The second reviewer point in this cluster (position of the dashed line in Fig. 2C) is adjudicated under C013.
Second check: another model agreed
Report a problem with this finding
Wording The Discussion states "The resolution of anemia, sickling, and hemolysis that was seen in all patients is consistent with improved red-cell health and function", while the Results report only that hemolysis markers "normalized or decreased after the risto-cel infusion (Fig. S5)" and present group means for shrinking subsets (13 patients at month 6, 5 at month 12).
- p. 11 Discussion, penultimate paragraph
The resolution of anemia, sickling, and hemolysis that was seen in all patients is consistent with improved red-cell health and function.
- p. 8 Results, Hemoglobin and Other Blood Analyses
Markers of hemolysis (lactate dehydrogenase, indirect bilirubin, and haptoglobin levels and reticulocyte counts) normalized or decreased after the risto-cel infusion
- p. 6 Results, Adverse Events (patient who died)
successful engraftment had occurred by month 2 after the risto-cel infusion (total hemoglobin level, 11 g per deciliter
- p. 11 Discussion, paragraph 4
seen in all patients is consistent with improved
Only the authors can say what is right here.
Judge's note: Read the Discussion sentence on the p011 image (crop C014-d and the full page), the Results paragraph on p008 (crop C014-b) and the account of the patient who died on p006 (crop C014-c: "total hemoglobin level, 11 g per deciliter" at month 2). The article presents no per-patient analysis showing resolution in every treated patient; hemoglobin is given as means with ranges (month 1 mean 12.2, range 9.0 to 14.7) and the sickling and hemolysis data are deferred to Figs. S4 and S5, which are not among the files supplied. The unqualified "in all patients" is stronger than the Results wording; wording/overstatement query for the authors.
Second check: another model agreed
Report a problem with this finding
Wording In Figure 2 Panel A the month-1 bar is labelled HbF 23.8% and HbS 9.2%, but the "HbF/(HbF+HbS) — %" row prints 69 at month 1, whereas the panel's own figures give 72; the nine other timepoints agree with the printed row after rounding.
- p. 9 Figure 2, Panel A, month-1 bar labels and the "HbF/(HbF+HbS) — %" row
HbF/(HbF+HbS ) — % 9 69 70 67 66 65 65 64 64 64 66
- p. 9 Figure 2, Panel A, month-1 stacked bar
23.8%
- p. 9 Figure 2, Panel A, month-1 stacked bar
9.2%
- p. 9 Figure 2A, month 1 bar and HbF/(HbF+HbS) row
23.8% 9.2% ... HbF/(HbF+HbS) — % 9 69 70
Only the authors can say what is right here.
Judge's note: Worked from the upright p009 page image. Baseline 6.8/(6.8+65.5)=9.4 (printed 9); month 2 46.3/66.3=69.8 (70); month 3 57.8/86.2=67.1 (67); month 4 62.3/93.7=66.5 (66); month 5 64.8/99.4=65.2 (65); month 6 64.7/99.5=65.0 (65); month 9 63.4/99.3=63.8 (64); month 12 63.9/99.3=64.3 (64); month 15 63.5/99.4=63.9 (64); month 18 65.4/99.4=65.8 (66). Only month 1 (23.8/33.0=72.1 vs printed 69) does not reconcile. The caption defines the row as HbF ÷ [HbF + HbS] but does not say whether it is a mean of per-patient ratios (which can differ from the ratio of means at month 1, when residual transfused HbA varies widely) or restricted to nontransfused samples; the Results text does refer to "nontransfused blood by month 1". Because no such restriction is stated in the figure, the panel is internally unexplained at that one timepoint; raised as a query, not a plain arithmetic error.
Second check: another model agreed