Usefulness of monitoring of B cell depletion in rituximab-treated rheumatoid arthritis patients in order to predict clinical relapse: a prospective observational study

Abstract

Our objective was to evaluate the contribution of monitoring B cell subset depletion after rituximab in patients with rheumatoid arthritis (RA) in order to guide reintroduction to forestall relapse. This prospective, monocentre study included all RA patients receiving two 1-g rituximab infusions at a 15-day interval. The patients were followed clinically and biologically every 2 months until rituximab reintroduction. The physician was blinded to lymphocyte-typing results to diagnose relapse and, hence, retreatment. Among the 39 patients included between March 2010 and December 2011 and followed until April 2013, seven received two rituximab cycles, yielding a total of 46 cycles for analysis. After the two rituximab cycles, the total number of CD19⁺ B cells decreased significantly (0·155 versus 0·0002 G/l, P < 0·0001), with complete depletions in all patients of CD19⁺CD38⁺⁺CD24⁺⁺ (transitional) (P < 0·0001) and CD19⁺CD27⁺ (memory) B lymphocytes. A significant majority of patients relapsed within the 4 months following repopulation of total B (P = 0·036), B transitional (P = 0·007) and B memory (P = 0·01) lymphocytes. CD19⁺ B lymphocyte repopulation preceded clinical RA relapse and enabled its prediction 4 months in advance. Hence, monitoring of CD19⁺ B lymphocytes could serve as a tool to predict those relapses.

Introduction

Multiple factors are implicated in the pathophysiology of rheumatoid arthritis (RA). The demonstrated major role of B lymphocytes renders rituximab use of interest in that context 1. B lymphocytes are derived from pluripotent haematopoietic precursors 2. They begin their maturation in the bone marrow, which they leave as transitional B cells expressing the CD19⁺CD38⁺⁺CD24⁺⁺ phenotype. The latter are seen mainly after complete lymphocyte depletion and are the first to appear after bone-marrow engraftment. Thereafter, the percentage of B transitional cells declines progressively, as peripheral lymphocyte maturation continues 3. After encountering an antigen, B cells are activated and pursue their differentiation into antigen-specific antibody-secreting plasma cells and CD19⁺CD27⁺ B memory cells.

Rituximab, a monoclonal chimeric anti-CD20 antibody, recognizes that determinant expressed on intramedullary pre-B to B memory-stage lymphocytes. It has been used to treat B lymphomas since 1997, and the first controlled trial on RA was in 2004 4. It is administered as a two-infusion cycle, each of 1 g, and separated by 15 days. A new cycle is initiated only after disease activity recurs.

Rituximab causes notable biological modifications, starting with the expected depletion of circulating B cells and lowering of inflammatory parameter levels, e.g. erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) 5–7, associated with treatment efficacy. The rheumatoid factor (RF) concentration also decreases significantly 6–11, and is associated with response to treatment 6. Apparently, rituximab also diminishes the anti-citrillunated protein antibody (ACPA) titre, albeit to varying degrees among studies 6,8,11, but seems to be associated with therapeutic response. Two constant observations have been made: rituximab induces an overall depletion of CD19⁺ B cells 12 and no clinical RA relapse has been documented in the absence of circulating B cells 6,7,13.

The kinetics of B cell-subpopulation reappearance showed the re-emergence of B transitional CD19⁺CD38⁺⁺CD24⁺⁺ 7,11,13, while B memory CD19⁺CD27⁺ lymphocytes appeared later 7.

The currently used rituximab administration scheme relies upon the recurrence of clinical RA activity before reinfusion, which contributes to further structural deterioration for patients subjected to numerous inflammatory flares between two cycles, or systematic infusion at fixed intervals (e.g. biannually), but potentially exposing the patient to overtreatment and the enhanced risk of adverse events 14,15. Moreover, the time to relapse is specific to each patient and remains difficult to predict prior to treatment. Because rituximab specifically targets CD20-expressing B cells and, based on reported B lymphocyte depletion and regeneration kinetics after its administration, monitoring the depletions of different B cell subsets to anticipate clinical RA relapse and initiate reinfusions as early as possible to ensure tight disease control warrants investigation 16.

To address that need, we undertook this study to follow clinical RA progression and the evolution of B lymphocyte populations at regular intervals in rituximab-treated patients. Our primary objective was to determine whether lymphocyte typing would enable us to anticipate a clinical RA relapse in rituximab-treated patients and serve as a tool for therapeutic follow-up. The secondary aim was to identify factors predictive of a good response to rituximab.

Patients and methods

Patients

This single-centre, prospective, observational study was conducted from March 2010 to April 2013. All rituximab-treated patients whose RA satisfied American College of Rheumatology/European League Against Rheumatism (ACR/EULAR) criteria 17 were included. A standard rituximab protocol was applied: two 1-g infusions at a 15-day interval. Patients initially taking methotrexate or oral corticosteroids continued those drugs, the doses of which were lowered during follow-up as a function of the response to rituximab.

After the two rituximab infusions, patients' clinical and biological follow-up every 2 months consisted of physical examination, with calculation of the disease activity score 28 (DAS28-ESR) and monitoring of inflammatory parameters (ESR and CRP), autoantibody levels (RF with semi-quantitative latex and Waaler–Rose tests and ACPA with a second-generation anti-cyclic citrullinated protein test), immunoglobulin levels (g/l), and B lymphocyte typing (total CD19⁺, transitional CD19⁺CD38⁺⁺CD24⁺⁺ and memory CD19⁺CD27⁺). Adverse events observed and/or reported by the patients were recorded systematically.

Patients were followed by the same rheumatologists, who were blinded to the B cell-depletion and -typing results. The decision to start another rituximab cycle was based solely upon clinical findings, inflammatory parameters and, sometimes, imaging results, according to routine follow-up procedures. Relapse was defined as DAS28 >3·2 and/or a modification of the EULAR therapeutic response (from a good to moderate or null response, or moderate to null). All patients gave consent to take part in this study. Given that this is an observational study, with no supplementary biological samples obtained and no experimental research, according to current French law no ethical approval was required.

Lymphocyte typing

B cells and B cell subsets were measured by flow cytometry. Samples of 100 μl of peripheral blood were stained with the following monoclonal antibodies: CD19 [B4, immunoglobulin (Ig)G1]-fluorescein isothiocyanate (FITC)-labelled (Beckman Coulter, Miami, FL, USA); CD19 (J4, 119, IgG1) ethyl cysteinate dimer (ECD)-labelled, CD38 (LS 198-4-3, IgG1)-)-phycoerythrin-cyanin 5 (PE-Cy5)-labelled and CD24 (ALB 9, IgG1)-)-PE-CY5-labelled (all from Beckman Coulter); CD19 (J4119, IgG1)-PE-Cy5-labelled (Beckman Coulter) and CD27 (M-T 271, IgG1)-FITC-labelled (Becton Dickinson, Franklin Lakes, New Jersey, USA). Red blood cell lysis was performed with the TQ prep cell preparation system (Beckman Coulter). Labelled cells were analysed on an Epic XL flow cytometer (Beckman Coulter) with System II and Expo 32 software (Fig. 1). Absolute B cell counts were calculated from the total lymphocyte count provided by the white blood cell count and differential. The normal CD19⁺ B cell count, according to the Immunology Laboratory norms, was 1·0–0·4 × 10⁹ cells/l.

Figure 1.

Identification of CD19⁺ B cells, CD24⁺⁺CD38⁺⁺ transitional B cells and CD19⁺CD27⁺ B cells before the first rituximab infusion in one representative patient. Population sizes are measured as percentages of total lymphocytes.

Although CD19⁺ B cells were recorded as present, CD19⁺CD38⁺⁺CD24⁺⁺ and CD19⁺CD27⁺ B cell subpopulations were considered absent (0 cells/l) when <100 CD19⁺ lymphocytes were analysed during 5 min of acquisition. Lymphocyte repopulation was defined as a >0·05% increase of the CD19⁺ lymphocyte count compared to the patient's lowest previous value.

Statistical analyses

All the variables were subjected to descriptive analysis, with number (%) of each. Student's t-test was used to compare quantitative variables between two paired groups or Pearson's χ² test for comparisons of qualitative variables. All tests were performed considering bilateral hypotheses. All analyses were performed using r version 3·1·1 and the SARP package or spss version 10·0 software (SPSS, Inc., Chicago, IL, USA).

For relapse prediction, quality of B lymphocyte apparition as a predictor of relapse was performed on all patients who responded to the treatment and is expressed as sensitivity and specificity and their 95% exact confidence interval (CI) (using binomial distribution). Offset between B lymphocyte apparition and relapse was analysed for all patients presenting both events. Results are given as minimum, maximum, median and its exact 95% CI and interquartile range (IQR).

For response prediction factors, association between clinical variables and binary response (quantified as EULAR ‘good’ therapeutic response or EULAR ‘moderate or none’ response) was studied using either Fisher's exact test for contingency tables [for nominal predictors: RF-positive, ACPA-positive, immunoglobulin G (IgG) >12·7 g/l and presence of a combined disease-modifying anti-rheumatic drug (DMARD)] or logistic regression (for quantitative predictors: DAS28). Results are expressed as odds ratios (ORs) and their 95% CIs. Association was considered significant if P < 0·05. Due to the small sample size, no multivariate analyses could be conducted.

Results

Patient characteristics

This cohort study followed 39 patients [29 women and 10 men; mean age ± standard deviation (s.d.): 57 ± 11·7, range 26·5–78 years]. All had RA satisfying ACR/EULAR 2010 criteria. Their main characteristics are reported in Table 1.

Table 1.

Characteristics of the 39 rituximab-treated rheumatoid arthritis (RA) patients

Characteristic	Value
Sex, female/male	29 (74·4%)/10 (25·6%)
Mean ± s.d. age (years)	57 ± 11·7
RF-positive	25 (64·1%)
ACPA-positive	29 (74·4%)
RF+/ACPA+	22 (56·4%)
RF−/ACPA−	7 (17·9%)
Previous biotherapies
1 anti-TNF	13 (33·3%)
2 anti-TNF	8 (20·5%)
3 anti-TNF	4 (10·3%)
Abatacept	10 (25·6%)
Anakinra	5 (12·8%)
Tocilizumab	1 (2·6%)
None	5 (12·8%)
Previous rituximab cycles
0	17 (43·6%)
1	9 (23·1%)
2	6 (15·4%)
3	5 (12·8%)
4	1 (2·6%)
6	1 (2·6%)
DMARD
0	12 (30·8%)
Methotrexate	19 (48·7%)
Leflunomide	7 (17·9%)
Tiopronin	1 (2·6%)
Initial combined corticosteroid dose
None	17 (43·6%)
1–5 mg/day	9 (23·1%)
6–10 mg/day	10 (25·6%)
>10 mg/day	3 (7·7%)

ACPA = anti-citrullinated protein antibodies; DMARD = disease-modifying anti-rheumatic drugs; RF = rheumatoid factor; TNF = tumour necrosis factor; s.d. = standard deviation.

Rituximab was first-line therapy for five patients, three because anti-tumour necrosis factor (TNF) was contraindicated and the two others because of RA-associated lupus and anti-phospholipid syndrome or autoimmune hepatitis and sclerosing cholangitis.

Among the 39 patients, 22 (56·4%) were taking corticosteroids (median dose 5, range 0–20) mg/day.

Because seven patients received their second rituximab cycle during follow-up, 46 treatment cycles could be analysed. The mean initial DAS28 was 5·44 ± 0·93.

Rituximab-induced B lymphocyte depletion

Just before each first perfusion of the 46 cycles, the mean number of CD19⁺ B lymphocytes was 0·145 × 10⁹ cells/l (range 0·0053–0·9765), with means of 0·0269 × 10⁹ CD19⁺CD38⁺⁺CD24⁺⁺ transitional cells/l (range 0·0002–0·3027) and 0·0195 × 10⁹ CD19⁺CD27⁺ memory cells/l (range 0·0004–0·1137). Two months after the first rituximab infusion of each cycle, CD19⁺ B cells were depleted significantly, with a mean circulating number of 0·0002 × 10⁹ cells/l (range 0–0·0159) (P = 0·006), and complete depletion for 26 of the 46 cycles. For all cycles analysed, CD19⁺CD38⁺⁺CD24⁺⁺ (P < 0·001) and CD19⁺CD27⁺ B lymphocytes (P < 0·001) were depleted completely.

Evolution of B lymphocyte depletion

During follow-up the 46 cycles obtained 44 good or moderate EULAR therapeutic responses, with two non-responders; among these, at the end of the observation period, no relapses occurred in two patients whose B lymphocytes did not reappear. One patient relapsed without reappearance; two relapsed with re-emergence of their total CD19⁺ B cells, without detectable subpopulations (as explained, fewer than 100 CD19⁺ B cells were detected, therefore B cells were considered as present but B cell subsets could not be measured reliably).

The sensitivities of B lymphocyte detection to predict clinical RA relapse were 0·9762 (95% CI = 0·8743, 0·9994) for CD19⁺, 0·881 (95% CI = 0·7437, 0·9602) for CD19⁺CD38⁺⁺CD24⁺⁺ and 0·881 (95% CI = 0·7437, 0·9602) for CD19⁺CD27⁺. Specificities were 1 (95% CI = 0·1581, 1) for CD19⁺, for CD19⁺CD38⁺⁺CD24⁺⁺ and for CD19⁺CD27⁺.

Analysis of the curves of clinical RA relapse and B lymphocyte and subset regeneration as a function of time after the first rituximab infusion of each cycle showed that re-emergence preceded clinical relapse and occurred after approximately 4 months (Fig. 2). This was confirmed by the analysis of delay between B lymphocyte and subset regeneration and RA relapse with a median of 4 months for CD19⁺, for transitional CD19⁺CD38⁺⁺CD24⁺⁺ and for memory CD19⁺CD27⁺ (Table 2, Fig. 3).

Figure 2.

Curves of the cumulative numbers of patients with clinical rheumatoid arthritis (RA) relapses (–) or CD19⁺ (– – –), CD19⁺CD38⁺⁺CD24⁺⁺ (– – –) or CD19⁺CD27⁺ (···) B lymphocyte subpopulation regeneration as a function of the number of months after the first rituximab infusion of each cycle.

Table 2.

Delay in months between clinical rheumatoid arthritis (RA) relapse and Blymphocytes and subset regeneration

B lymphocyte typing	Minimum (months)	Median (months)	Maximum (months)	Median (95% CI)	IQR
CD19+	2	4	16	(4, 6)	2
CD19+CD38++CD24++	0	4	14	(4, 4)	1
CD19+CD27+	0	4	14	(4, 4)	0

CI = confidence interval; IQR = interquartile range.

Figure 3.

Box-plot representing the delay for each patient in months between clinical rheumatoid arthritis (RA) relapse and B lymphocyte regeneration.

Based on these observations, we examined whether the repopulation–relapse interval was significant for all cycles with B cell repopulation and/or clinical relapse (n = 42). A significantly higher percentage of patients' relapses occurred within the 4 months after B cell regeneration: 23 of 42 cycles for CD19⁺ (P = 0·036), 30 for transitional CD19⁺CD38⁺⁺CD24⁺⁺ (P = 0·007) and 29 for memory CD19⁺CD27⁺ (P = 0·01) (Fig. 4).

Figure 4.

Distributions of the numbers of patients with clinical rheumatoid arthritis (RA) relapses as a function of the time of their occurrence or context: ≤4 months (), >4 months after () or without re-emergence (▪) of the different B lymphocyte subsets.

Monitoring of clinical responses and factors predictive of a good response

The mean efficacy duration of a treatment cycle was 13 ± 4 months. At the end of the study, only four patients had not relapsed and, among them, two were still completely devoid of B lymphocytes.

During follow-up, only one severe adverse event was reported: the spontaneous rupture of a woman's spleen. After notification to the pharmacovigilance authorities and analysis of case imputability, rituximab could not be incriminated formally.

After 6 months of direct observations, among the 46 cycles analysed, 22 (47·8%) yielded good and 22 moderate EULAR therapeutic responses; concerning the two (4·4%) non-responders, one each was RF- or ACPA-positive.

In our search for potential prognostic factors based on this population, the univariate analysis retained low DAS28 as a good predictor of a good therapeutic response to rituximab (Table 3).

Table 3.

Factors predicting EULAR therapeutic responses to rituximab based on univariate analysis

	EULAR therapeutic response
Factor	Good (n = 22)	Moderate or none (n = 24)	Odds ratio (95% CI)	P
DAS28, mean ± s.d.	5·13 ± 0·75	5·73 ± 0·98	0·45 (0·21, 0·90)	0·023
RF-positive	17 (77·3%)	13 (54·2%)	2·88 (0·83, 11·11)	0·100
ACPA-positive	17 (77·3%)	16 (66·7%)	1·7 (0·47, 6·67)	0·425
IgG >12·7 g/l	7 (31·8%)	9 (37·5%)	0·83 (0·24, 2·84)	0·771
Combined DMARD	15 (68·2%)	16 (66·7%)	1·07 (0·31, 3·76)	0·913

ACPA = anti-citrullinated protein antibodies; CI = confidence interval; DAS28 = disease-activity score-28; DMARD = disease-modifying anti-rheumatic drugs (methotrexate or leflunomide); IgG = immunoglobulin G; RF = rheumatoid factor; EULAR = European League Against Rheumatism; s.d. = standard deviation.

Discussion

The results of this study showed that monitoring B lymphocyte depletion was a reliable tool, able to predict clinical RA relapse in rituximab-treated patients. Indeed, observing B lymphocyte and subsets regeneration has a high sensitivity to predict clinical relapse. Moreover, a significant majority of our cohort patients relapsed within 4 months after B cell repopulation, particularly transitional and memory subsets. We did not find a published study whose primary objective was to determine whether lymphocyte repopulation could anticipate clinical RA relapse. Half the 24 patients investigated in 2006 by Leandro et al. relapsed ≤2 months following B cell repopulation and the other half 5–12 months thereafter 13. In another study, half the relapses occurred ≤4 months after regeneration but only 12 patients were evaluated 11.

Notably, two B cell subsets (transitional CD19⁺CD38⁺⁺CD24⁺⁺ and memory CD19⁺CD27⁺) re-emerged simultaneously herein, whereas earlier reports observed transitional B cell reappearance before that of memory lymphocytes 3,7,18. Only one of our patients experienced a clinical RA relapse without prior B lymphocyte regeneration, in contrast to previous findings of their reappearance always preceding clinical relapse 6,7,13. After re-evaluating that patient's file, it was concluded that her DAS28, calculated during the consultation, was probably over-estimated because the declared ‘mixed pains’, most probably of mechanical (long-standing structural involvement) rather than RA inflammatory origin, were attributed erroneously to a disease flare. Both non-responders had only partial B lymphocyte depletions, in agreement with Vital et al., who described such therapeutic failures in that context 19.

Our response rate to rituximab, 95·6%, appears to be much higher than seen previously in other studies. As shown previously, rituximab may be beneficial both in seronegative and seropositive RA; conversely, RF and/or ACPA positivity consistently enhanced the clinical response 20–24. Moreover, a majority of our patients are double-positive for RF and ACPA and double-positive patients are known to achieve significantly larger improvement of DAS28 compared to other patients 20.

Discrepant information on factors predicting the response to rituximab has been published 25. Few studies focused on factors predictive of a good EULAR response, as we have done; most studies generally sought those prognostic of a good or moderate response. However, given our small number of patients, we could use only univariate analysis. Therefore, these results need to be put into perspective.

In agreement with Chatzidionysiou et al.'s meta-analysis of 10 European registries 20, our univariate analysis retained an initial low DAS28 as being associated significantly positively with a good EULAR therapeutic response. None the less, the fact that low DAS28 predicts a good response can seem obvious, as a smaller improvement is required to achieve this therapeutic response. However, available information on DAS28 has varied. Indeed, in another registry analysis, a subgroup analysis identified high initial DAS28 as predictive of a better therapeutic response in patients who failed on anti-TNF 21, and in a French cohort a multivariate analysis found high DAS28 to be predictive of a good EULAR therapeutic response 26.

Based on our cohort, RF was not a factor predictive of response, unlike Soliman et al.'s registry analysis of 646 patients and Couderc et al.'s study of 64 patients 21,27. The recent meta-analysis by Isaacs et al. 23 found better DAS28 improvement in RF-positive patients, but that finding was not used to examine EULAR therapeutic responses, and the authors of several investigations reported that RF-positivity remained as predictive of a good to moderate EULAR therapeutic response in multivariate analyses 22,24,28,29. However, only two studies exclusively included good responders to attempt to identify prognostic factors, as did we, and their results were also discordant. Chatzidionysiou et al.'s meta-analysis of 2019 patients did not retain RF-positivity as a predictive factor 20, whereas Solau-Gervais et al.'s analysis based on 95 patients found it to be prognostic for a good EULAR therapeutic response 26.

Combining a DMARD with rituximab was not identified herein as being predictive of EULAR therapeutic response, in agreement with previous publications 4,21,26, which would seem to support the not, as yet, formally established hypothesis that this biological could be used alone.

In one study 28, an elevated intial IgG level was found to be prognostic of a good-to-moderate therapeutic response. That observation was not confirmed by our cohort, in agreement with Couderc et al. 27

The main strength of our study is its prospective and observational design, which included sequential clinical, biological and B lymphocyte subpopulation typing during long-term follow-up of rituximab-treated RA patients in ‘real-life’ routine practice (notably with various prior treatments and various RA clinical presentations). Despite this clinical diversity, our analysis identified the reappearance of CD19⁺ B lymphocytes and subsets as a marker of disease progression at least 4 months before a clinical diagnosis of RA relapse could be confirmed. That finding demonstrated the contribution of monitoring B lymphocyte depletion as a predictive biological parameter during the follow-up of rituximab-treated patients to guide the decision to retreat precociously to achieve tight control of RA.

However, our relatively small population did not allow confirmation of parameters described previously as predictive of a therapeutic EULAR response, even though the latter was mostly evaluated as good-to-moderate, and we restricted this assessment to only good EULAR responses. Moreover, the 2-month interval between follow-up consultations for B lymphocyte typing prevented the determination of the precise reappearance kinetics of different subsets, especially CD19⁺CD38⁺⁺CD24⁺⁺ transitional and CD19⁺CD27⁺ memory B cells, which re-emerged synchronously herein, and leads us to recommend their assessment at a shorter interval or with an ultrasensitive technique 30.

Conclusion

Monitoring the depletion of different B lymphocyte subsets after rituximab administration anticipates clinical RA relapse ≤4 months after their reappearance. This strategy provides physicians with a pertinent and reproducible follow-up parameter for rituximab-treated patients, enabling them to envisage rituximab retreatment as closely as possible to the relapse to tightly control the disease.

The results of this study justify the initiation of a study on a larger population. Once this predictive marker could be validated, the next step would be an evaluation of the therapeutic contribution, over the short and long terms, of anticipated rituximab readministration, perhaps at a lower dose 31 compared to current practice, which recommends retreatment only after clinical RA relapse and at least 6 months after the preceding rituximab cycle.

Author contributions

A. P. T. enrolled patients, conducted clinical evaluations and wrote the manuscript. V. G. conceived of the study, participated in its design and coordination, enrolled patients, helped to draft the manuscript and gave final approval of the manuscript. S. J. conducted the lymphocyte typing, data analysis and manuscript revision and gave final approval of the manuscript. S. G. and E. C. performed the statistical analysis. I. D. and A. R. carried out lymphocyte typing. H. B. enrolled patients and conducted biological sampling and data collection. O. V., X. L. L. and O. B. participated in the design of the study and manuscript revision and gave final approval of the manuscript. All authors read the final manuscript.

Disclosure

O. V. reports personal fees from Roche outside the submitted work; V. G. reports advisory board participation for Roche/Chugai, outside the submitted work. The other authors declare no conflicts of interest.

Funding

None.