To the Editor:

We read with great interest the recent article by Kim and colleagues in The Journal of Rheumatology on disease flares in patients with systemic lupus erythematosus (SLE) with endstage renal disease (ESRD) on dialysis.1 The authors reached the conclusion that continued follow-up and appropriate therapy, including immunosuppressants, should be considered for patients with SLE receiving dialysis. We support and appreciate the authors’ work and agree with their conclusions, but we have some concerns about some of the details in the article.

First, the only constitutional symptom in Table 1 is fever.1 Constitutional symptoms of SLE usually include fever, fatigue, poor mental performance, loss of appetite, weight loss, and myalgia. We suggest the authors add more descriptions of constitutional symptoms to make the study more objective and realistic. In addition, it is unclear whether patients with co-infection with HIV or hepatitis virus, organ transplantation, other autoimmune diseases, tumors, or pregnancy should be excluded, as these usually interfere with the study results. Further, the patients enrolled in this study were those who started dialysis. Given the relatively high disease activity in patients who have just started dialysis, this should be taken into account in the interpretation of the SLE flare results. The enrollment of patients several months after starting dialysis may be more significant for the study since, at that point, the level of disease activity and inflammation in patients stabilize.2

Second, studies have shown that a subset of patients may experience higher rates of disease activity after initiation of renal replacement therapy (RRT), but in general, patients with SLE on dialysis therapy experience a decrease in clinical and serological activity, a phenomenon known as the burnout period.3,4 As in the case of patients in the peritoneal dialysis group, the cumulative use of glucocorticoids (GCs) should theoretically decrease. Why did the cumulative dose of GCs increase significantly over 1 year in the hemodialysis group? The authors do not explain the reasons for this in the article. Based on the difference before and after nonrenal SLE Disease Activity Index (SLEDAI), could this indicate that patients in the hemodialysis group had poorly controlled SLE disease activity and required higher doses of hormones for maintenance and treatment during the course of therapy?

Third, although patients in both the peritoneal dialysis and hemodialysis groups were investigated within the same time period, limitations in terms of differences across time periods and length of follow-up should be considered. Subgroup analysis across time periods could help reduce study bias. Although SLE flares are defined as a measurable increase in disease activity in ≥ 1 organ systems involving new or more severe clinical symptoms and/or laboratory measurements, a combination of elements such as change or addition of therapy, adjustment of immunosuppressive therapy, patient response, and nonrenal SLEDAI scores were considered so that flares were described more accurately and measured more objectively. However, the diagnosis of SLE flares is ultimately considered based on the clinical judgment of the rheumatologist and may be biased due to the common belief that SLE remains quiescent after ESRD and the fact that not all patients are closely followed by a rheumatologist.

Finally, and most importantly, the authors performed a Cox proportional regression analysis of the trial variables in the text to identify risk factors for disease flares in patients with lupus nephritis ESRD during RRT. If the authors used Cox proportional regression analysis, they should have described clearly the specific details associated with the Cox proportional regression analysis, including the endpoint events of this study, the specific variables included, and the critical values set. To say the least, it would be inappropriate to use Cox proportional regression analysis to identify risk factors in this study. Instead, logistic regression analysis would be more appropriate. The process includes variables that showed significance in the univariate analysis that were included in the multivariate analysis. Multivariate analyses were performed using a binary logistic regression model and expressed as odds ratios, with P values < 0.05 considered statistically significant.

We would like to thank Kim and colleagues again for their contributions to this study and look forward to hearing from them.

We would like to thank the members and staff of the Department of Rheumatology and Immunology of the Zhuzhou Central Hospital who contributed to this manuscript.