Sinusoidal obstruction syndrome after infant conditioning - an age signal, a regulatory exposure threshold, and a prophylaxis that failed its randomised trial
also called SOS · VOD · veno-occlusive disease · hepatic veno-occlusive disease · HVOD · hepatic sinusoidal obstruction syndrome
Busulfan and the other conditioning drugs can block the small veins draining the liver, which makes the liver swell, the belly fill with fluid and the child turn yellow. It kills about one in five children who develop it. Being under a year old looks like a risk factor, and the drug that was supposed to prevent it failed when it was properly tested.
The same thing, for a scientist
Sinusoidal obstruction syndrome is endothelial injury to the hepatic sinusoids caused by conditioning, presenting in the first weeks after transplant with weight gain, painful hepatomegaly, ascites and hyperbilirubinaemia. In 105 paediatric allogeneic transplants at one centre, 15 (14.3%) developed SOS at a median of 12 days with 20.0% mortality; patient age under 1 year carried a univariate odds ratio of 7.25 (p=0.037) that did NOT survive multivariable adjustment, where gemtuzumab ozogamicin (OR 9.24), ferritin above 2400 ng/mL (OR 5.74) and INR at or above 1.3 (OR 8.02) did [E1]. The mechanistic link to busulfan is regulatory as well as clinical: the US intravenous busulfan label warns of increased hepatic veno-occlusive disease risk above a per-dose AUC of 1,500 uM*min [E2], which converts to 98.5 mg*h/L cumulative over 16 doses - within half a percent of the FORUM overexposure boundary of 98.0 mg*h/L [E3]. The obvious countermeasure does not work: prophylactic defibrotide in 372 high-risk patients gave day-30 SOS-free survival of 67% against 73% on best supportive care alone and was stopped early for futility [E4].
The age signal, stated honestly
Age under 1 year: OR 7.25, p=0.037, univariate 1. It did not survive multivariable adjustment.
That is the whole of what this beat found. The number of infants in that 105-patient cohort is not given, the confidence interval is not reported in the abstract, and three other factors - gemtuzumab exposure, ferritin above 2400 ng/mL and INR at or above 1.3 - displaced age in the multivariable model. An odds ratio of 7.25 that does not survive adjustment is a signal in a small dataset, not an established independent risk factor, and this entry records it as such.
It is nonetheless the most specific infant SOS finding available, and it is consistent with what is mechanistically expected: SOS is endothelial injury from alkylator exposure, and infants receive three alkylators in eight days through a hepatic conjugating system that is still maturing.
What is not established, and is worth saying plainly: no study retrieved by this beat reports SOS incidence in infants conditioned with the Interfant-06 busulfan/cyclophosphamide/melphalan regimen. The largest relevant comparison, FORUM's busulfan-versus-treosulfan analysis, reports exactly one fatal veno-occlusive death in each arm of 308 patients - a number too small to compare 3.
The exposure threshold, and the unit trap in it
The US intravenous busulfan label states: 'Hepatic Veno-Occlusive Disease (HVOD): Increased risk of developing HVOD at AUC greater than 1,500 uMmin. Monitor serum transaminases, alkaline phosphatase and bilirubin daily'* 2.
That threshold is per dose. The modern trial literature reports cumulative exposure in mg*h/L. Converting requires busulfan's molecular weight of 246.3 and the number of doses:
1,500 uM*min x 0.2463 mg/L per uM / 60 min per h = 6.158 mg*h/L per dose
6.158 x 16 doses = 98.5 mg*h/L cumulative
The FORUM overexposure boundary, derived independently from event-free and GVHD-free relapse-free survival in 145 children, is 98.0 mg*h/L 3. A regulatory hepatotoxicity limit set decades ago and a 2026 trial-derived efficacy ceiling agree to 0.5%.
That convergence is worth two things. It is a genuine cross-validation of the upper bound - two entirely different methods, one toxicological and one outcome-based, landing on the same number. And it is a worked example of a trap: the same quantity published in incommensurable units, one per-dose and one cumulative, comparable only through a molecular weight and a dose count. A reader who has seen one of these numbers and not the other has a self-consistent and incomplete picture, and nothing in either document signals that the other exists.
The prophylaxis that failed
HARMONY was an open-label randomised phase 3 trial in 104 centres across 14 countries, enrolling 372 patients at high or very high risk of SOS, including paediatric patients from one month of age, randomised 1:1 to prophylactic intravenous defibrotide 25 mg/kg/day plus best supportive care versus best supportive care alone 4.
| defibrotide (n=190) | best supportive care (n=182) | |
|---|---|---|
| day-30 SOS-free survival | 67% (58-74) | 73% (62-80) |
| hazard ratio | 1.27 (0.84-1.93), p=0.85 |
Enrolment was stopped early for presumed futility on the recommendation of the independent data monitoring committee.
The practical consequence for this beat is sharp. Identifying an infant as high-risk for SOS currently changes monitoring, not treatment: daily transaminases, alkaline phosphatase and bilirubin per the busulfan label 2, fluid and weight surveillance, and readiness to treat established SOS. There is no evidence-based prophylactic intervention to offer. The one modifiable determinant that remains is conditioning exposure itself - keeping busulfan below the 98 mg*h/L ceiling that both the label and FORUM point at, which requires the therapeutic drug monitoring Interfant-06 made optional.
| kind | toxicity |
|---|---|
| severity | 14.3% incidence and 20.0% mortality in one paediatric series; median onset 12 days after transplant |
| mitigation | no effective prophylaxis - defibrotide prophylaxis failed a randomised phase 3 trial for futility. Available measures are daily transaminase, alkaline phosphatase and bilirubin monitoring per the busulfan label, weight and fluid surveillance, early defibrotide THERAPY on recognition, and keeping busulfan exposure below the ~98 mg*h/L ceiling |
| infant signal | age under 1 year, univariate OR 7.25 (p=0.037), NOT confirmed on multivariable analysis; a second paediatric cohort of the same size found no age effect on SOS at all |
| not measured | No study retrieved by this beat reports SOS incidence in infants conditioned with the Interfant-06 busulfan/cyclophosphamide/melphalan regimen specifically. |
affects
[ "drug-busulfan", "drug-cyclophosphamide", "drug-melphalan", "drug-treosulfan", "drug-thiotepa" ]
busulfan exposure threshold
{
"label_per_dose": "AUC > 1,500 uM*min",
"converted_cumulative_16_doses": "98.5 mg*h/L",
"forum_overexposure_boundary": "98.0 mg*h/L",
"agreement": "within 0.5%",
"conversion_factor": "1 uM*min = 0.004105 mg*h/L using busulfan MW 246.3",
"caution": "the conversion assumes the 16-dose q6h schedule; once-daily busulfan gives a different per-dose AUC for the same cumulative exposure"
}harmony prophylaxis result
{
"n": 372,
"defibrotide_day30_sos_free_survival": "67% (58-74)",
"bsc_day30_sos_free_survival": "73% (62-80)",
"hazard_ratio": "1.27 (0.84-1.93), p=0.85",
"outcome": "stopped early for presumed futility"
}- There is no prophylaxis. Identifying an infant as high-risk for sinusoidal obstruction syndrome currently changes how closely they are watched and not what they receive, because the one candidate drug failed a properly powered randomised trial for futility.
- The only modifiable determinant is the exposure the protocol does not require you to measure. Both the regulatory threshold and the trial-derived ceiling point at roughly 98 mg*h/L of cumulative busulfan, and Interfant-06 makes measuring it optional.
- The two published thresholds are in incommensurable units and nothing warns you. A per-dose uM*min limit on a label and a cumulative mg*h/L window in a trial paper describe the same hazard, and comparing them needs a molecular weight and a dose count that neither document supplies.
- Three alkylators in eight days through a maturing hepatic conjugating system is the mechanistic setup for endothelial injury, and the protocol front-loads the most hepatotoxic agent as an ORAL preparation whose absorption in a vomiting infant is uncontrolled.
- A univariate odds ratio of 7.25 that does not survive adjustment is exactly the kind of number that gets quoted without its second half. This graph records it with the adjustment result attached, because the same finding is cited elsewhere as though it were established.
- Is age under 1 year an independent risk factor for sinusoidal obstruction syndrome? Two paediatric cohorts of about 105 patients each disagree - one finds a univariate OR of 7.25 that does not survive adjustment, the other finds no age effect at all - and this is answerable from any large paediatric transplant registry.
- What is the SOS incidence after Interfant-06 busulfan/cyclophosphamide/melphalan conditioning in infants? The regimen has never been reported as a named analysis, so the hepatotoxicity of the exact combination given to these babies is unpublished.
- Does keeping busulfan below 98 mg*h/L reduce SOS, as opposed to reducing the composite of 'treatment-related toxicities'? The convergence of the label threshold and the FORUM ceiling is striking, and neither source isolates SOS as the endpoint.
- Would prophylactic defibrotide have worked in infants specifically? HARMONY enrolled from one month of age but had a median age of 14 years, so the age group with the strongest risk signal is the one least represented in the trial that closed the question.
- E1moderate confidence 3 of 5: One strong study, or several smaller consistent ones.paper
In 105 children, adolescents and young adults undergoing allogeneic HSCT at a single paediatric centre over 12 years, 15 (14.3%) developed hepatic sinusoidal obstruction syndrome by EBMT paediatric criteria at a median of 12 days after transplant, with 20.0% mortality; age under 1 year carried a univariate odds ratio of 7.25 (p=0.037) but was not confirmed on multivariable analysis, where gemtuzumab ozogamicin (OR 9.24, p=0.048), ferritin above 2400 ng/mL (OR 5.74, p=0.023) and INR at or above 1.3 (OR 8.02, p=0.007) were confirmed.
Analysis of risk factors for hepatic sinusoidal obstruction syndrome following allogeneic hematopoietic stem cell transplantation in pediatric patients. · Kloehn J, Brodt G, Ernst J, Gruhn B · Journal of Cancer Research and Clinical Oncology 2022What this evidence cannot show
Single centre, 105 patients over 12 years, 15 SOS events - a multivariable model fitted on 15 outcomes. The number of patients under 1 year is not stated in the abstract, so the odds ratio of 7.25 rests on an unknown and probably small numerator and its confidence interval is not reported. It is a univariate finding that did NOT survive adjustment. Conditioning regimens are not broken out, so the contribution of busulfan specifically cannot be recovered, and no Interfant-06-style busulfan/cyclophosphamide/melphalan subgroup exists.
Evidence that argues against this (1)
- A larger and better-controlled paediatric dataset found no age effect on SOS at all: in 105 children conditioned with myeloablative busulfan for non-malignant disease, there was no impact of age or busulfan exposure on sinusoidal obstruction syndrome incidence, despite a very large age effect on donor chimerism in the same cohort. Two similarly-sized paediatric cohorts, one finding an age signal and one finding none, is the honest state of this question. Dvorak CC et al. 2025
- E2established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.database
The US label for intravenous busulfan warns of increased risk of hepatic veno-occlusive disease at an AUC greater than 1,500 uM*min and directs daily monitoring of serum transaminases, alkaline phosphatase and bilirubin; it also carries a boxed warning for myelosuppression occurring in 100% of patients and notes cardiac tamponade in paediatric thalassaemia patients receiving high-dose oral busulfan with cyclophosphamide.
BUSULFAN injection, for intravenous use - US prescribing information (Structured Product Label, ANDA215102) · Eugia US LLC · DailyMed / FDA Structured Product Labeling, openFDA drug label API 2024What this evidence cannot show
Retrieved from api.fda.gov on 2026-09-20; SPL set id 06883ac0-4b2a-4f26-89bf-512d37b9ba7f, effective 2024-06-27. A regulatory document states what is approved, not current best practice, and the 1,500 uM*min threshold is inherited from decades-old data. The conversion to 98.5 mg*h/L cumulative was performed by this beat using busulfan MW 246.3 and a 16-dose schedule; it is arithmetic on the label's own numbers and is not a figure the label prints. The label describes intravenous busulfan, while Interfant-06 specifies ORAL busulfan, whose bioavailability is variable - so the exposure a given oral dose produces is less predictable than the label's numbers imply.
Evidence that argues against this (1)
- A per-dose toxicity threshold cannot be transferred to a cumulative exposure target without assuming the dosing schedule, and the assumption is load-bearing: at 16 doses the conversion gives 98.5 mg*h/L, but once-daily busulfan schedules with four doses give a completely different per-dose AUC for the same cumulative exposure. The convergence with FORUM's 98.0 mg*h/L ceiling is real only for the 16-dose schedule and should not be quoted as a general equivalence. Ben Hassine K et al. 2026
- E3strong confidence 4 of 5: Replicated in peer-reviewed work by more than one group.paper
In FORUM, busulfan exposure above a cumulative AUC of 98.0 mg*h/L was associated with an increased risk of treatment-related toxicities and graft-versus-host disease in 145 paediatric ALL patients; in the trial's prespecified busulfan-versus-treosulfan comparison across 308 children, exactly one case of fatal veno-occlusive disease occurred in each group.
Refining busulfan exposure enhances pediatric ALL HSCT outcomes: insights from the International FORUM study. · Ben Hassine K, Gloor Y, Dupanloup I, Uppugunduri CRS, Mlakar V, Gonzales F, Guengoer T, Ifversen M, et al. · Blood Advances 2026What this evidence cannot show
The 98.0 mg*h/L ceiling comes from a post hoc analysis of 145 patients whose primary endpoints were event-free and GVHD-free relapse-free survival, not SOS - 'treatment-related toxicities' is a composite and the SOS-specific contribution is not isolated. The one-fatal-VOD-per-arm figure comes from the separate Kalwak busulfan-versus-treosulfan report and is two events in 308 patients, which is uninformative as a comparison. All FORUM patients were over 4 years old at HSCT.
Evidence that argues against this (1)
- Two fatal veno-occlusive events in 308 children on modern chemoconditioning is a very low rate against the 14.3% SOS incidence and 20% SOS mortality reported in a single-centre paediatric series using varied regimens. Either contemporary fludarabine-based conditioning is substantially less hepatotoxic than older busulfan-cyclophosphamide regimens, or FORUM's reporting counts only fatal cases while the single-centre study counted all EBMT-criteria SOS. The two figures are not measuring the same thing. Kalwak K et al. 2025
- E4established confidence 5 of 5: Multiple independent lines of evidence; treated as settled in the field.paper
In HARMONY, an open-label randomised phase 3 trial across 104 centres in 14 countries enrolling 372 patients at high or very high risk of sinusoidal obstruction syndrome (including paediatric patients from one month of age), prophylactic intravenous defibrotide 25 mg/kg/day plus best supportive care gave day-30 SOS-free survival of 67% (95% CI 58-74) versus 73% (62-80) with best supportive care alone (HR 1.27, 95% CI 0.84-1.93; p=0.85), and enrolment was stopped early for presumed futility following the planned interim analysis.
Defibrotide plus best standard of care compared with best standard of care alone for the prevention of sinusoidal obstruction syndrome (HARMONY): a randomised, multicentre, phase 3 trial. · Grupp SA, Corbacioglu S, Kang HJ, Teshima T, Khaw SL, Locatelli F, Maertens J, Stelljes M, et al. · The Lancet Haematology 2023What this evidence cannot show
Median age 14.0 years (IQR 4.0-41.0), so the trial is dominated by adolescents and adults and the infant subgroup is small and not separately reported; 'paediatric patients from one month of age' describes eligibility, not enrolment. Open-label, and stopped early at interim analysis, so the final estimates rest on fewer events than planned. A trial that fails to show prophylactic benefit in a mixed high-risk population does not establish that no subgroup benefits - and defibrotide remains established as TREATMENT for established SOS, which this result does not challenge.
Evidence that argues against this (1)
- Prophylaxis failing does not mean the risk factor is not worth identifying. Early defibrotide THERAPY on recognition of SOS remains standard, and a paediatric study of early defibrotide therapy and post-transplant SOS risk factors exists in the same period, so an infant identified as high-risk still warrants the intensified surveillance that permits early treatment. The negative here is specific to giving the drug prophylactically to everyone deemed high-risk. Goto H et al. 2024
label warns of hepatic veno-occlusive disease above a per-dose AUC of 1,500 uM*min = 98.5 mg*h/L over 16 doses
alkylator component of the Interfant-06 conditioning block contributing to hepatic endothelial injury
14.3% incidence and 20% mortality in paediatric HSCT; no effective prophylaxis after the HARMONY futility result
three alkylators in eight days through a maturing hepatic conjugating system