Refined National Institutes of Health response algorithm for joint- and fascia-associated chronic graft-versus-host disease

Establishment of joint-/fascia-associated graft-versus-host disease (GvHD) occurs in 3–24% of patients with chronic GvHD (cGvHD).¹ The consequences of joint/fascia involvement can be hugely debilitating, with symptoms including joint stiffness and restricted movement, limb tightness, edema, and subcutaneous sclerosis/fasciitis. In 2005, the National Institutes of Health (NIH) applied response criteria for joint/fascia GvHD, with the aim to characterize GvHD organ involvement.² Subsequently, in 2014, a response algorithm with updated recommendations was implemented.¹ Following establishment of the aforementioned response criteria, ibrutinib became the first approved treatment for the indication in 2017.³

The 2014 criteria suggested that a decrease in NIH joint/fascia score or an increase in photographic range of motion (P-ROM) score by ≥ one point at any site be classed as an improvement. On the other hand, progression is classified by an increase in NIH score by ≥ one point (including from 0–1) or a decrease in P-ROM score by ≥ one point. Although the 2014 algorithm has been imperative for advances in treatment of cGvHD, its real-world use has been challenged. Firstly, opposing changes in joints are considered overall progression, but the appropriateness of this has not been evaluated. Furthermore, a change from 0–1 on the NIH organ scoring algorithm is not considered progression in most other GvHD involvement sites, and there is no evidence supporting the exception of joint/fascia.  Finally, there are inconsistencies between NIH and P-ROM scores, with occasions of one improving and the other exacerbating.¹

Yoshihiro Inamoto, National Cancer Center Hospital, Tokyo, JP, aimed to address these limitations and uncertainties, and we hereby present a summary of the multicenter evaluation of the performance of the 2014 NIH response algorithm for joint/fascia GvHD.¹

Study Design

• Prospective, multicenter, longitudinal, observational study of two independent study cohorts by the Chronic GvHD Consortium
• Clinically meaningful changes were defined by the distribution method (half a standard deviation [SD]) and anchor-based methods (changes in the measures that correlated with patient- or clinician-reported changes in joint/fascia involvement)

Patients

• Adult patients (≥ 18 years) with systemically treated cGvHD were recruited from 2007–2012 (training cohort; n = 488) and 2013–2017 (replication cohort; n = 357)
• Diagnosis and assessment of cGvHD in the training and replication cohorts were made according to the 2005 and 2014 NIH consensus criteria, respectively
• Training cohort:
  • 209 patients presenting with joint/fascia involvement in ≥ one visit
  • Total visits: 1,578
• Replication cohort:
  • 191 patients presenting with joint/fascia involvement in ≥ one visit
  • Total visits: 1,195
  • Eight patients (2%) and 93 visits (7%) had joint/fascia abnormalities not caused by GvHD and were excluded
• cGvHD organ involvement and manifestations were reported by patients and clinicians at enrollment and every six months thereafter

Patient characteristics

• Table 1 illustrates patient characteristics at the time of enrollment

Table 1. Patient characteristics of the training and replication cohorts at the time of enrollment²

 Results

• One half of a SD in the total P-ROM scores for the training and replication cohorts were 1.17 and 1.49, respectively, suggesting that a two-point change in total P-ROM score is clinically relevant

Divergent* response in individual P-ROM scores

• In the training cohort, 455 paired visits revealed joint/fascia manifestations, be it in the previous or most current visit
• Worse individual P-ROM scores occurred in 15–21% of paired visits
• Where individual P-ROM scores were used to calculate overall response, 26% elicited improvement, 32% stability, and 43% worsening†
• In contrast, when assessed using a total P-ROM score, 88% of cases were classified as stable
• Divergent responses were observed in 12% of the visits

*Divergent responses were classified as improvement in one joint but worsening in another with reference to individual P-ROM scores

†Worsening in any joint was considered overall worsening, even when other joints are improved

Changes in NIH joint/fascia scores from 0–1

• Sixty-three of the 455 paired visits in the training cohort elicited a change in NIH joint/fascia score from 0–1, while their total P-ROM score did not worsen
• Of these 63 visits, very few patients or clinicians perceived the joint condition as worse (Table 2), suggesting that a change from 0–1 NIH score should not be considered as worsening

Table 2. Overall assessment for 63 paired visits with change in NIH joint/fascia score from 0–1 without worsening in total P-ROM score²

*Patient perception not available in ten paired visits

Contrasting NIH and total P-ROM scores

• Thirteen of the 455 paired visits showed divergence between NIH joint/fascia score and total P-ROM score
• Clinicians perceived these visits to be mainly stable (54%) or improved (38%), whereas patients perceived their condition as mostly stable (44%) or worse (33%)
• Although divergence between NIH and overall P-ROM scores is rare, the results from these 13 visits suggest that overall response cannot be determined in these cases

Refined response algorithm for joint/fascia GvHD

• A refined algorithm for the assessment of joint/fascia GvHD involvement was based on data from the training cohort and is summarized in Table 3

Table 3. Refined response algorithm for cGvHD²     

NIH, National Institutes of Health; P-ROM, photographic range of motion

• In both the training and replication cohorts, the proportion of visits with worsening joint/facia GvHD involvement decreased from around 50% to < 20% when using the 2014 NIH and refined algorithms, respectively
• Joint/fascia GvHD involvement was classed as uninterpretable in the case of divergent NIH and total P-ROM scores in the training (3%) and replication (2%) cohorts
• Reclassification following assessment using the refined algorithm occurred in 40% of the training cohort and 35% of the replication cohort
  • Many of the cases previously interpreted as worsened were reclassified as improved or stable

Conclusion

• The study provides an improved algorithm for assessing joint/fascia cGvHD involvement that can be used in clinical trials
  • By assessing ½ SDs, it was advised that a two-point change in total P-ROM score is clinically relevant
  • In line with the 2014 NIH criteria scoring for other sites, a change from 0–1 in NIH joint/fascia score should not be considered as worsening
  • Overall response should be defined as uninterpretable when rare divergent responses remain between NIH joint/fascia score and total P-ROM score
• When following the refined algorithm, the level of visits reporting worsening of joint/fascia GvHD involvement was around 30% lower in both cohorts compared to the 2014 criteria