Change of Treatment for Control of Chronic Graft-Versus-Host Disease (cGVHD) as a Time-Dependent Covariate for Nonrelapse Mortality and Survival.

Time to first exacerbation of cGVHD was reported recently as a prognostic risk factor for survival in a study of 96 patients (pts) by Kim et al. (doi:10.1038/sj.bmt.1705806). Pts with exacerbation of cGVHD experienced worse survival and higher nonrelapse mortality (NRM) compared to pts with no exacerbation. This association was particularly notable among pts with earlier onset of exacerbation. We conducted a retrospective study to validate these findings in 738 consecutive pts with cGVHD diagnosed after 5/1/01 without a prior history of recurrent or persistent malignancy after allogeneic hematopoietic cell transplantation. Pts had cGVHD according to NIH criteria, and the dataset did not include pts with persistent, recurrent or delayed onset of acute GVHD. Dates and the reasons for subsequent change of treatment after the initial treatment for cGVHD have been prospectively recorded since 05/01/01. Change of treatment is defined as

1. addition of a new treatment for control of cGVHD,

2. increased dose of corticosteroids to at least 1 mg/kg every other day for control of cGVHD, or

3. addition of a topical immunosuppressive medication for treatment of a previously unaffected site.

Change of treatment because of toxicity alone was not included, since the focus of the study was to use change of treatment as a surrogate marker for inadequate control of cGVHD. Fifty-six percent (397/738) of pts had a change of treatment for one or more of the following reasons: progressive signs or symptoms of cGVHD (n = 214), new site (n = 190), unimproved or persistent cGVHD (n = 118) or physician preference (n = 5). The median interval time to change of treatment was 3.8 months (range, 0.1–54), and 75% of the changes occurred within 9.1 months after the initial treatment. In time-dependent Cox proportional hazards regression models, change of treatment was associated with an increase in risk of NRM compared to pts without change of treatment [hazard ratio (HR), 2.14; 95% C.I, 1.5–3.1; p < 0.0001]. The HR for change of treatment within 4 months was 2.40 (95% C.I., 1.7–3.4), and the HR for a change of treatment beyond 4 months was 1.51 (95% C.I., 0.9–2.5). Thus, relative to a change of treatment within 4 months of cGVHD diagnosis, the increase in NRM associated with change in treatment after 4 months was somewhat attenuated (HR, 0.63; 95% C.I., 0.5–1.0; p = 0.05). Risk factors that were significant for changing treatment were skin involvement at diagnosis (HR, 1.31; 95% C.I, 1.0 – 1.7; p = 0.02), use of growth-factor mobilized blood cells for transplantation (HR, 1.41; 95% C.I, 1.0 – 1.9; p = 0.03), male recipient with female donor (HR, 1.26; 95% C.I, 1.0 – 1.6; p = 0.04), and HLA-mismatched recipient (HR, 1.32; 95% C.I, 1.0 – 1.7; p = 0.05). Changing treatment was strongly associated with decreased survival (HR, 3.82; 95% C.I. 2.8–5.2; p < 0.0001), but not with decreased risk of recurrent malignancy (HR = 0.80; 95% C.I., 0.5–1.3; p = 0.38). Our results show that a change of treatment is associated with increased risk of NRM, and that this increase may not be greatly affected by the time at which the change of treatment is made. The association of a change in treatment with increased NRM suggests that a first change of treatment at any time could be used as an indicator of unfavorable outcome in double-blind clinical trials of treatment for cGVHD.