Impact of Prior Acute GVHD on Chronic GVHD Outcomes: a Chronic Graft versus Host Disease Consortium Study

Chronic graft-versus-host disease (CGVHD) is one of the most debilitating complications post hematopoietic cell transplant (HCT). Several studies have evaluated factors affecting survival in CGVHD,^1–8 most utilizing the older definition of CGVHD where patients with any symptoms of GVHD (acute or chronic) may be classified as CGVHD if symptomatic beyond day 100 post HCT. However, the new NIH consensus guidelines have re-defined criteria for CGVHD diagnosis and staging⁹ based on actual symptoms at presentation and excludes patients with persistent, recurrent or late AGVHD who were previously considered to have CGVHD. The CGVHD consortium follows a large multi-center prospectively enrolled cohort of CGVHD patients utilizing NIH consensus guidelines for assessments.¹⁰ Several reports from this consortium have identified factors affecting survival including global severity, thrombocytopenia, KPS and the overlap syndrome.^11,12

Prior AGVHD has been associated with a higher mortality in patients who develop CGVHD,^1,3 indicating that these patients may be more debilitated and hence suffer worse outcome. However, the clinical and prognostic importance of prior AGVHD on subsequent CGVHD in the current era using NIH diagnostic and staging criteria is unknown. We evaluated the significance of prior AGVHD on CGVHD presentation, overall response and survival in a contemporary cohort of CGVHD patients prospectively enrolled by the CGVHD consortium.

Cases enrolled in the consortium could be incident (enrollment < 3 months after diagnosis) or prevalent (enrollment ≥3 months after diagnosis).¹⁰ Detailed CGVHD assessments were performed reflecting the recommendations of the NIH Consensus Conference,^9,13 and are published in the cohort study rationale and design summary.¹⁰ Prior AGVHD was considered present if maximum grade I-IV was reported. Comparisons were performed between two groups: Prior AGVHD and no prior AGVHD. Statistical comparisons between groups were made with the two-sample t-test for continuous variables, and chi-squared test for categorical variables. Spearman correlation coefficients (ρ, rho) and simple kappa statistics were used to describe rank correlation and agreement between acute and CGVHD overall severity scores and skin, GI, and liver severity. Survival was estimated by the Kaplan-Meier method.¹⁴ Overall survival was calculated from enrollment, and non-relapse mortality (NRM) was defined as death without prior relapse. Failure-free survival (FFS) was defined as survival without prior relapse or adding new systemic medications to treat CGVHD. Cox regression¹⁵ was used for multivariate analysis of survival relative to presence or absence of prior acute GVHD and other risk factors.

The total cohort included 496 patients: 329 with prior AGVHD and 167 with no prior AGVHD (de-novo CGVHD) (table 1). The prior AGVHD and de-novo CGVHD groups did not differ with respect to time between HCT and CGVHD diagnosis, or time between HCT and enrollment. The proportion of patients with classic and overlap CGVHD and the distribution of CGVHD global severity (mild, moderate and severe CGVHD), using clinician or patient reported and NIH scores, was similar. These observations support the hypothesis that CGVHD in patients with prior AGVHD does not differ fundamentally from de-novo CGVHD. In particular, prior AGVHD was not associated with AGVHD manifestations at the time of CGVHD diagnosis (overlap syndrome). There were some differences between prior AGVHD and de-novo CGVHD groups: Myeloablative conditioning, a mismatched or unrelated donor, and thrombocytopenia were more frequent in those with prior AGVHD. No further differences were observed between the two groups, including frequency of organ involvement, functional assessments or patient reported outcomes.

Table 1.

Patient and CGVHD characteristics

Patient characteristics
Characteristic	All patients (N = 496)		Prior acute GVHD (N = 329)		No prior acute GVHD (N = 167)		P-value*
	n	Count (%)	n	Count (%)	n	Count (%)
Site	496		329		167		<.001
Fred Hutchinson Cancer Research Center		227 (46%)		171 (52%)		56 (33%)
Others		269 (54%)		158 (48%)		111 (67%)
Case Type	496		329		167		0.91
Incident		281 (57%)		187 (57%)		94 (56%)
Prevalent		215 (43%)		142 (43%)		73 (44%)
Adults (≥ 18 years)	496	482 (97%)	329	321 (98%)	167	161(96%)	0.46
Patient age at registration (years), median (range)	496	51 (2–79)	329	51 (2–79)	167	52 (8–74)	0.82
Patient gender: male	496	290 (58%)	329	201 (61%)	167	89 (53%)	0.10
Race	496		329		167
White		444 (90%)		291 (88%)		153(92%)	0.28
Other		52 (10%)		38 (12%)		14 (8%)
Months from HCT to onset of CGVHD, median (range)	496	7.1 (1.2–291)	329	7.4 (1.3–57.6)	167	6.7 (1.2–291)	0.52
Months from onset of CGVHD to enrollment, median (range)	496	2.1 (0–32.5)	329	2.3 (0–32.5)	167	1.9 (0–29.8)	0.96
Months from HCT to enrollment, median (range)	496	12 (2.9–294)	329	12.1 (2.9–60.7)	167	11.6 (3.9–294.2)	0.53
Diagnosis	496		329		167		0.84
AML		164 (33%)		104 (32%)		60 (36%)
ALL		63 (13%)		44 (13%)		19 (11%)
MDS		73 (15%)		48 (15%)		25 (15%)
NHL		70 (14%)		46 (14%)		24 (14%)
Others		126 (25%)		87 (26%)		39 (24%)
Disease status	495		328		167		0.23
Early		164 (33%)		110 (33%)		54 (32%)
Intermediate		214 (43%)		134 (41%)		80 (48%)
Advanced		117 (24%)		84 (26%)		33 (20%)
Transplant source	496		329		167		0.14
Bone marrow		35 (7%)		28 (9%)		7 (4%)
Cord blood		23 (5%)		17 (5%)		6 (4%)
Peripheral blood		438 (88%)		284 (86%)		154 (92%)
Transplant type	495		329		166		0.02
Myeloablative		283 (57%)		200 (61%)		83 (50%)
Not myeloablative		212 (43%)		129 (39%)		83 (50%)
Donor-patient gender combination	491		326		165		0.77
Female to male		144 (29%)		97 (30%)		47 (28%)
Others		347 (71%)		229 (70%)		118 (72%)
Donor type	495		328		167		0.01
Matched related		216 (44%)		130 (40%)		86 (52%)
Matched unrelated		200 (40%)		136 (41%)		64 (38%)
Mismatched		79 (16%)		62 (19%)		17 (10%)
Grade of prior acute GVHD	492		325		167		NA
Grade 0		167 (34%)		0 (0%)		167 (100%)
Grade I		54 (11%)		54 (17%)		0 (0%)
Grade II		223 (45%)		223 (69%)		0 (0%)
Grade III		41 (8%)		41 (13%)		0 (0%)
Grade IV		7 (1%)		7 (2%)		0 (0%)
Clinician and patient-reported chronic GVHD characteristics at enrollment
CGVHD type	496		329		167		0.39
Classic		85 (17%)		53 (16%)		32 (19%)
Overlap		411 (83%)		276 (84%)		135 (81%)
NIH Consensus severity score	496		329		167		0.92
Less than mild		3 (1%)		2 (1%)		1 (1%)
Mild		48 (10%)		30 (9%)		18 (11%)
Moderate		283 (57%)		187 (57%)		96 (58%)
Severe		162 (32%)		110 (33%)		52 (30%)
Platelet count at enrollment of < 100 x109/L	493	79 (16%)	327	61 (19%)	166	18 (11%)	0.03
Bilirubin at enrollment > 2mg/dl	491	28 (6%)	325	16 (5%)	166	12 (7%)	0.30
KPS at enrollment	496		329		167		0.98
≥ 80		252 (51%)		167 (51%)		85 (51%)
< 80		154 (31%)		103 (31%)		51 (31%)
Missing		90 (18%)		59 (18%)		31 (18%)
NIH skin erythema score, median (range)	495	0 (0–100)	328	0 (0–100)	167	0 (0–100)	0.59
NIH skin movable sclerosis score, median (range)	495	0 (0–60.3)	328	0 (0–53.1)	167	0 (0–60.3)	0.38
NIH skin non-movable sclerosis score, median (range)	495	0 (0–49.5)	328	0 (0–49.5)	167	0 (0–31.5)	0.25
VSS skin score, median (range)	495	0.8 (0–19.1)	328	0.8 (0–16.2)	167	0.6 (0–19.1)	0.92
Skin sclerosis score	496		329		167
0		395 (80%)		266 (81%)		129 (77%)	0.67
1		74 (15%)		46 (14%)		28 (17%)
2		9 (2%)		5 (2%)		4 (2%)
3		6 (1%)		5 (2%)		1 (1%)
4		12 (2%)		7 (2%)		5 (3%)
Clinician 0–3 skin score	496		329		167		0.13
None		191 (39%)		120 (36%)		71 (42%)
Mild		111 (22%)		83 (25%)		28 (17%)
Moderate		120 (24%)		75 (23%)		45 (27%)
Severe		74 (15%)		51 (16%)		23 (14%)
Clinician fascia score	496		329		167		0.21
None		408 (82%)		277 (84%)		131 (78%)
Mild		65 (13%)		39 (12%)		26 (16%)
Moderate		15 (3%)		7 (2%)		8 (5%)
Severe		8 (2%)		6 (2%)		2 (1%)
Clinician 0–3 GI tract score	496		329		167		0.13
None		345 (70%)		220 (67%)		125 (75%)
Mild		117 (24%)		87 (26%)		30 (18%)
Moderate		32 (6%)		20 (6%)		12 (7%)
Severe		2 (<1%)		2 (1%)		0 (0%)
Clinician GI esophagus score	496		329		167		0.21
None		419 (84%)		270 (82%)		149 (89%)
Mild		57 (11%)		44 (13%)		13 (8%)
Moderate		10 (2%)		8 (2%)		2 (1%)
Severe		10 (2%)		7 (2%)		3 (2%)
Clinician upper GI score	496		329		167		0.62
None		400 (81%)		264 (80%)		136 (81%)
Mild		64 (13%)		46 (14%)		18 (11%)
Moderate		21 (4%)		13 (4%)		8 (5%)
Severe		11 (2%)		6 (2%)		5 (3%)
Clinician lower GI score	496		329		167		0.06
None		429 (86%)		276 (84%)		153 (92%)
Mild		49 (10%)		37 (11%)		12 (7%)
Moderate		15 (3%)		14 (4%)		1 (1%)
Severe		3 (1%)		2 (1%)		1 (1%)
Malabsorption involved	496	7 (1%)	329	5 (2%)	167	2 (1%)	0.77
Esophageal stricture or web involved	496	3 (1%)	329	1 (<1%)	167	2 (1%)	0.23
Weight loss since onset	477	164 (34%)	318	114 (36%)	159	50 (31%)	0.34
Clinician 0–3 liver score	494		328		166		0.10
None		239 (48%)		163 (50%)		76 (46%)
Mild		188 (38%)		128 (39%)		60 (36%)
Moderate		65 (13%)		35 (11%)		30 (18%)
Severe		2 (<1%)		2 (1%)		0 (0%)
Functional characteristics and patient-reported outcomes
Walk test (feet)	426	500 (170–1150)	285	500 (170–1150)	141	482 (211–800)	0.11
Grip strength	474	60.0 (2–167)	315	61.4 (4.9–167)	159	57.7 (2–138.8)	0.43
FACT physical well-being score	409	22.0 (1–28)	271	22.0 (1–28)	138	22.0 (4–28)	0.30
FACT social/family well-being score	407	23.0 (0–28)	269	23.0 (0–28)	138	23.3 (7–28)	0.29
FACT emotional well-being score	408	19.1 (4–24)	270	20.0 (6–24)	138	19.0 (4–24)	0.38
FACT functional well-being score	407	16.0 (2–28)	271	16.0 (2–28)	136	16.7 (4–28)	0.75
FACT-BMT subscale	407	27.0(11–40)	271	27.0 (11–39)	136	27.0 (13–40)	0.28
FACT-BMT trial outcome index (TOI)	407	64.7(22–95)	271	64.0 (22–93.8)	136	64.8 (25–95)	0.37
FACT-G score	404	79.4 (23–108)	268	79.5 (33–106)	136	79.4 (23–108)	0.51
FACT-BMT total score	404	106 (36–146)	268	105.8 (48.1–140)	136	106.2 (36–146)	0.39
SF36 physical functioning score	405	42.3 (14.9–57)	269	42.3 (14.9–57)	136	44.4 (14.9–57)	0.12
SF36 role physical score	408	37.3 (17.7–56.9)	271	34.8 (17.7–56.9)	137	37.3 (17.7–56.9)	0.11
SF36 bodily pain score	409	46.1 (19.9–62.1)	271	46.1 (19.9–62.1)	138	46.1 (24.1–62.1)	0.93
SF36 general health score	406	40.1 (16.2–63.9)	268	40.1 (18.6–63.9)	138	41.0 (16.2–63.9)	0.54
SF36 vitality score	410	45.8 (20.9–70.8)	272	45.8 (20.9–70.8)	138	45.8 (20.9–70.8)	0.11
SF36 social functioning score	410	40.5 (13.2–56.8)	272	40.5 (13.2–56.8)	138	40.5 (13.2–56.8)	0.60
SF36 role emotional score	407	48.1 (9.2–55.9)	271	48.1 (9.2–55.9)	136	48.1 (9.2–55.9)	0.86
SF36 mental health score	410	51.4 (13.4–64.1)	272	50.0 (16.2–64.1)	138	52.8 (13.4–64.1)	0.55
SF36 physical component scale (PCS)	399	39.3 (15.5–60.7)	264	38.7 (15.9–59.3)	135	40.4 (15.5–60.7)	0.15
SF36 mental component scale (MCS)	399	50.0 (15.3–68.4)	264	49.8 (15.3–67.4)	135	50.7 (17.1–68.4)	0.95
HAP maximum activity score	408	73.0 (36–94)	271	73.0 (36–94)	137	73.0 (36–94)	0.79
HAP adjusted activity score	408	63.0 (14–94)	271	63.0 (16–94)	137	63.0 (14–94)	0.79
Modified HAP maximum activity score	408	54.0 (19–74)	271	55.0 (19–74)	137	53.0 (24–74)	0.74
Modified HAP adjusted activity score	408	48.0 (11–74)	271	48.0 (11–74)	137	48.0 (11–74)	0.94
Lee symptom nutrition score	422	5.0 (0–70.0)	279	5.0 (0–70.0)	143	5.0 (0–50.0)	0.97
Patient 0–10 skin itching score	414	1.0 (0–10.0)	274	1.0 (0–10.0)	140	1.0 (0–10.0)	0.83
Lee symptom skin score	424	15.0 (0–100.0)	280	15.0 (0–100.0)	144	15.0 (0–100.0)	0.74
Patient 0–10 cGVHD overall score	411	4.0 (0–10.0)	270	4.0 (0–10.0)	141	4.0 (0–10.0)	0.72
Lee symptom overall score	424	19.8(0–65.3)	280	20.4 (0–64.0)	144	18.8(0–65.3)	0.41

^*P-value for two-sided two-sample t-test (continuous variables) or Chi-square test of independence (categorical variables)

Abbreviations: AML, acute myeloid leukemia; ALL, acute lymphoblastic leukemia; MDS, myelodysplastic syndrome; NHL, non Hodgkin’s lymphoma; GI, gastro-intestinal; CGVHD, chronic graft-versus-host disease; KPS, Karnofsky performance score

Acute GVHD Organ-Chronic GVHD Organ Correlation

Overall severity of prior AGVHD (including non-involvement) did not correlate well with overall NIH calculated severity of CGVHD at study enrollment (rho = 0.04). Similarly, AGVHD symptom severity (including non-involvement) of skin (rho = 0.01), gut (rho = 0.14), and liver (rho = 0.06) showed little correlation with CGVHD organ-specific involvement. Dichotomizing severity scores as ≤1 versus >1, the agreement between AGVHD and CGVHD severity was poor. The simple kappa statistics were 0.03 for overall global severity, −0.05 for skin, 0.02 for GI and 0.01 for liver comparisons, all considered poor agreement.

Response to treatment at 6 and 12 months from cohort enrollment

No difference was observed in NIH calculated response rates between those with and without prior AGVHD for both 6 months (n=302, p=0.88) and 12 month (n=228, p=0.77) assessments. Thirty six percent of patients with no prior AGVHD versus 37% with prior AGVHD had CR or PR at 6 months, and 33% versus 35% at 12 months.

Survival from enrollment

Of 496 patients in this cohort, 84 (17%) have died, with median follow-up time for survivors of 20 months (range 0–48). Two-year survival probability was similar at 0.83 (95%CI: 0.77–0.87) for those with prior AGVHD and 0.79 (95%CI: 0.71–0.85) for those without prior AGVHD. The multivariate analysis was performed restricting the study sample to incident cases only (n=281) to address potential survivor bias as patients who die soon after developing CGVHD do not enter the cohort as prevalent cases. Prior AGVHD was not a predictor of overall survival in either univariate (HR: 0.74, 95%CI: 0.41–1.31, p=0.30) or multivariate (Table 2, HR: 0.68, 95%CI: 0.35–1.30, p=0.24) analysis. We tested for and did not find a significant interaction between prior AGVHD and CGVHD type (classic vs. overlap). Overall survival was predicted by several manifestations at enrollment, including presence of thrombocytopenia, a lower karnofsky performance status, a higher total bilirubin, and a higher NIH severity score. There was no evidence of association between prior AGVHD and NRM (HR: 0.73, 95%CI: 0.30–1.74, p=0.47) or failure-free survival (HR:1.15, 95%CI: 0.79–1.66, p=0.47) in multivariate models with the same covariates as for overall survival.

Table 2.

Multivariate Analysis: Overall Survival from time of enrollment including incident cases only

Parameter	Category	Hazard Ratio	95% Confidence Limits		p-value
Prior acute GVHD	Yes	0.68	0.35	1.30	0.24
	No	1.0
Site	FHCRC	0.91	0.46	1.79	0.77
	Others	1.0
Time from HCT to enrollment	< 12 months	0.89	0.43	1.85	0.75
	≥ 12 months	1.0
Platelet count at enrollment	< 100 x109/L	2.44	1.26	4.73	0.008
	≥ 100 x109/L	1.0
Total bilirubin at enrollment	> 2 mg/dL	4.83	1.95	11.93	0.001
	≤ 2 mg/dL	1.0
KPS at enrollment	< 80	2.02	1.04	3.91	0.04
	Missing	1.77	0.78	4.05	0.17
	≥ 80	1.0
Age at transplant	≥ 50	1.19	0.63	2.28	0.59
	< 50	1.0
Donor match	Matched unrelated	0.97	0.48	1.95	0.93
	Mismatched	1.45	0.67	3.14	0.35
	Matched related	1.0
Donor-patient gender	Female donor male patients	1.46	0.75	2.83	0.27
	Others	1.0
Transplant type	Myeloablative	0.72	0.38	1.34	0.29
	Not myeloablative	1.0
NIH severity score	Severe	4.60	1.03	20.51	0.05
	Moderate	2.18	0.48	9.92	0.31
	Less than mild/Mild	1.0
CGVHD type	Overlap	3.41	0.45	25.96	0.24
	Classic	1.0

Abbreviations: FHCRC, Fred Hutchinson Cancer Research Center; CGVHD, chronic graft-versus-host disease; KPS, Karnofsky performance score

We next repeated the analysis in the whole group including both incident and prevalent cases (supplemental table). No evidence of association was found between prior AGVHD and overall survival (HR: 0.96, 95% CI: 0.60–1.55, p= 0.87). The analysis was also repeated after including only incident patients with more severe AGVHD (grade II-IV) in the prior AGVHD group. Results were similar to the current analyses (HR: 0.67, 95% CI: 0.35–1.27, p=0.21).

We undertook this analysis to evaluate the impact of prior AGVHD in a prospectively enrolled cohort of patients with CGVHD using NIH criteria.⁹ In our analysis, prior AGVHD was not associated with CGVHD characteristics, presence of overlap or classic symptoms, global CGVHD severity, responsiveness to treatment, quality of life, functional status, NRM or survival. There could be several reasons for our findings. Under the new CGVHD diagnostic criteria, late, persistent and recurrent AGVHD are not automatically considered CGVHD. These patients are not included in our cohort and may account for the poor prognosis associated with prior AGVHD seen in other studies that used older diagnostic criteria. Another possibility is that patients with more severe CGVHD may have higher early mortality and may not survive to be enrolled in our study. To address this concern, we restricted our multivariate analysis to include only the incident cases who were enrolled within three months of CGVHD diagnosis and demonstrated no effect of prior AGVHD. We then repeated the analysis in the whole cohort (incident and prevalent) and our results were similar (supplemental table). In the whole cohort, similar to a prior study,¹² we also noted an adverse impact of overlap syndrome on overall survival. In the current analysis, in the incident patients only, the frequency of overlap patients was much higher (n=250, 89%) with only 11% patients with classic chronic GVHD, which could account for the non-significant result. It is unlikely that including the overlap variable in the analysis is masking the importance of prior AGVHD, since there was no correlation between prior AGVHD and overlap syndrome at the time of enrollment.

Our results suggest that once a patient develops CGVHD according to the NIH diagnostic criteria, presence of prior AGVHD is not an adverse risk factor. This is in contradiction to a recently published large analysis from the CIBMTR, in which prior grade II-IV AGVHD was an adverse prognostic factor for survival and TRM.³ However, the CIBMTR study defined CGVHD according to the older criteria based on time since transplantation and involved mostly younger patients (median age 36 years) undergoing myeloablative conditioning (89%) and mostly calcineurin inhibitor-based GVHD prophylaxis. Median follow up in our cohort was 20 months compared to 73 months in the CIBMTR study. Our cohort was accrued from nine centers with an interest and focus on CGVHD, and is also more recent so differences in patient populations may also explain the differing results for AGVHD as a risk factor.

Progressive onset of CGVHD has been found to be an adverse factor for survival and TRM.⁴ We tried to investigate this issue, but patients do not enter our cohort until they are diagnosed with CGVHD by NIH diagnostic criteria. Retrospective review of medical records before patients were enrolled into the cohort showed that most documentation was inadequate to distinguish between progressive onset and interrupted onset.

In conclusion, our results suggest that presence of prior AGVHD does not affect the clinical presentation of either classic or overlap CGVHD and is not associated with lower survival or less responsive CGVHD. Instead, what seems to matter for predicting prognosis is the current status of CGVHD.