Exposure and outcomes definitions
While debate exists among clinicians and statisticians about the best representation of GV, CV has begun to emerge as the preferred metric due to its advantage of being relative to mean glucose and thus more descriptive of hypoglycemic excursions than SD, which is largely only reflective of hyperglycemic excursions.4-7 Therefore, CV was used to measure GV for each subject using laboratory-measured serum glucose values over pre-transplant (day -14 to 0) and post-HSCT (days 0-30) intervals to evaluate for HSCT-related stress GV and its effects. Because glucose CV is meant to reflect nonfasting glucose over an extended period, all glucose values were included, exceptglucose values within one hour of a previous glucose value were censored to reduce oversampling. Evaluation of potential outliers did not significantly impact results. Whole blood glucose measurements were typically performed daily in the early morning (between 12 and 3 A.M.) during hospital admission, and typically one to three daytime measurements per week after discharge. Because natural glucose CV categories are not clearly defined, we analyzed glucose CV continuously for all models. However, for descriptive purposes, glucose CV was categorized into four groups approximating published nonfasting CV ranges in various populations. The groups are: healthy adults (Grade 1, <20%), adults who are obese or receive aggressive oral therapy for type 2 diabetes (Grade 2, ≥20 and <27%), adults with type 1 or 2 diabetes that is well-controlled with insulin (Grade 3, ≥27 and <36%) and adults with poorly controlled type 1 diabetes (Grade 4, ≥36%).5,8
Post-HSCT outcomes included time to: death from any cause, transplant-related mortality (TRM), diagnosis of severe GVHD (defined as acute grades III-IV or chronic) 9,10, clinically significant infection in the first 100 days, and ICU admission in the first 100 days post-HSCT. Clinically significant infections were defined by hospitalization for a positive microbiology result or diagnostic/problem list code. Infections were further categorized by serious bacterial infection (SBI), viremia/viruria, or invasive fungal infection. Serious bacterial infections included infections such as bacteremia, bacterial peritonitis, meningitis, and pneumonia with hypoxemia. To be classified as clinically significant for viremia/viruria, we required both positive high viral load (of Ebstein Barr virus, Human herpes virus 6, adenovirus, cytomegalovirus, and BK virus) and symptoms requiring hospitalization and treatment, when available. To reduce the risk of misclassification bias, subjective outcomes (infection, GVHD) were reviewed by at least one study physician; any unclear outcomes were classified by 1-3 additional study physicians.
A list of exposure variables/covariates and their definitions is provided in the appendix (Table S1). Data were censored at the time of any recurrence/progression of primary disease, graft failure, or second HSCT. Because of the nature of HSCT care, there was not a notable amount of missing data or loss to follow-up.