Background Systemic lupus erythematosus (SLE) is a heterogenous autoimmune disease. Treatment trajectories following high disease activity state (HDAS), as defined by SLEDAI score 10, have not been well described.
Methods Longitudinal trajectories of patients from the Australian Lupus Registry were studied. HDAS periods were defined as the time from which HDAS begins, until the patient fulfils criteria for Low Lupus Disease Activity (LLDAS), or up to 365 days. Treatment escalation is defined as either an addition of hydroxychloroquine (HCQ), prednisolone (PNL) and immunosuppressant (IS), or any change in IS drug. De-escalation is either dose reduction or cessation of HCQ or IS without meeting treatment escalation criteria. Treatment trajectories were examined as the rolling sum (over time) of escalations and de-escalations and were clustered using k-means clustering methods. Different clustering partitions were tested. The R package kml was used for cluster determination and quality criterion calculations. The differences in time to resolution of HDAS between clusters were tested using likelihood ratio test. The relationships between continuous covariates (cumulative PNL, number of escalations, number of de-escalations, time to first escalation, number of mild-to-moderate flares, number of severe flares and change in damage index) and cluster were examined using analysis of variance (ANOVA) and Tukeys HSD.
Results A total of 210 HDAS periods (104 patients) were identified. Of the HDAS periods, patients were classified as treatment naïve (10%), HCQ inadequate response (20%), IS inadequate response (68%), and combination IS inadequate response (2%). The most commonly used IS was mycophenolate (23% of all HDAS periods). The trajectories were categorized into 3 final clusters: Cluster A (42/210) had more escalations than Cluster B (132/210) and Cluster C (36/210), see figure 1. There was no difference between clusters in the duration of time spent in HDAS, but a trend for higher cumulative PNL in Cluster A and they had significantly more and earlier escalations than Cluster B and C. Damage accrual appeared to be highest in Cluster C (the de-escalators) although not statistically significant. There was no difference between the distribution of the baseline treatment groups in each cluster.
Conclusions Treatment trajectories can be described using clustering that examines treatment escalations and de-escalations. This pilot study showed that treatment trajectories appear to have an effect on clinical outcomes. Further studies are planned to explore the relationship of patient characteristics or physician treatment decisions have on these clusters.
Funding Source(s): nil
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.