Hypogammaglobulinaemia during rituximab treatment in multiple sclerosis: A Swedish cohort study

Abstract Background and purpose Mechanisms behind hypogammaglobulinaemia during rituximab treatment are poorly understood. Methods In this register‐based multi‐centre retrospective cohort study of multiple sclerosis (MS) patients in Sweden, 2745 patients from six participating Swedish MS centres were identified via the Swedish MS registry and included between 14 March 2008 and 25 January 2021. The exposure was treatment with at least one dose of rituximab for MS or clinically isolated syndrome, including data on treatment duration and doses. The degree of yearly decrease in immunoglobulin G (IgG) and immunoglobulin M (IgM) levels was evaluated. Results The mean decrease in IgG was 0.27 (95% confidence interval 0.17–0.36) g/L per year on rituximab treatment, slightly less in older patients, and without significant difference between sexes. IgG or IgM below the lower limit of normal (<6.7 or <0.27 g/L) was observed in 8.8% and 8.3% of patients, respectively, as nadir measurements. Six out of 2745 patients (0.2%) developed severe hypogammaglobulinaemia (IgG below 4.0 g/L) during the study period. Time on rituximab and accumulated dose were the main predictors for IgG decrease. Previous treatment with fingolimod and natalizumab, but not teriflunomide, dimethyl fumarate, interferons or glatiramer acetate, were significantly associated with lower baseline IgG levels by 0.80–1.03 g/L, compared with treatment‐naïve patients. Switching from dimethyl fumarate or interferons was associated with an additional IgG decline of 0.14–0.19 g/L per year, compared to untreated. Conclusions Accumulated dose and time on rituximab treatment are associated with a modest but significant decline in immunoglobulin levels. Previous MS therapies may influence additional IgG decline.

Hypogammaglobulinaemia has been suggested to contribute to MS fatigue [25].
During pregnancy, immunoglobulin levels decrease physiologically by about 50% from the second trimester, returning to prepregnancy levels after delivery [34].
There is a need for more knowledge regarding the effect of long-term rituximab treatment on immunoglobulin levels in patients with MS and whether specific patient or treatment-related factors contribute to the risk of developing decreasing levels of immunoglobulins.

ME THODS
This multi-centre retrospective cohort study collected data from MS patients in Sweden with rituximab treatment.Patients were eligible for inclusion if diagnosed with MS or clinically isolated syndrome, had received at least one dose of rituximab at a participating centre and were monitored in the Swedish MS Registry (SMSreg).

Three centres in Stockholm (Danderyd Hospital, Karolinska
University Hospital, Academic Specialist Centre), one centre in Gothenburg (Sahlgrenska University Hospital) and two centres in Northern Sweden (Umeå University Hospital and Östersund Hospital) participated in this study.Inclusion started from the first registered dose of rituximab, on 14 March 2008, with follow-up until 25 January 2021.Patients who received plasmapheresis or intravenous immunoglobulin (IVIG) less than 100 days (circa five half-lives of immunoglobulins) before a rituximab dose were excluded.There were no available data on pre-existing immunodeficiencies, for example common variable immunodeficiency.The awareness of immunoglobulin measurements was lower during the first study years when pre-infusion measurements of immunoglobulins did not occur in clinical routine.However, it was possible to collect several immunoglobulin measurements from the diagnostic lumbar puncture occurring before the treatment started.When pre-treatment assessments of immunoglobulin levels started, immunoglobulin measurements below the normal range were uncommon in the cohort (Figure S1).

Pregnancy
Pregnancy data coverage was not complete in the SMSreg.There were 44 registered pregnancies during the study period.Fertile female patients with transient lowering of IgG levels were therefore additionally assessed through journals for unregistered pregnancies.

Data management and storage
Data were pseudonymized and stored in Excel with a secure login.

Main outcomes and measurements
The primary outcome was the mean reduction of IgG and IgM lev- The risk of developing hypogammaglobulinaemia below the lower limit of normal (LLN) (IgG <6.7 g/L) or low IgM (<0.27 g/L), categorized as mild (IgG 4-6.6 g/L) or severe (<4.0 g/L) hypogammaglobulinaemia, was evaluated.The laboratory reference interval at the 2.5th centile for IgG and IgM was used as a cut-off measurement, reflecting the clinical threshold for detecting patients at risk of imminent immunodeficiency.

Statistical methods
Linear regression was used to analyse the relationship between the outcomes IgG and IgM levels regarding exposure variables age, sex, centre and previous DMT at baseline.

Associations between exposures time on rituximab (in years)
or accumulated dose on rituximab (in grams) and outcome vari- In two separate models for IgG/IgM, the main exposure was changed to the total accumulated dose of rituximab.Additionally, for the IgM models, quadratic terms were included for both the time from the rituximab treatment start and total rituximab dosage to investigate the nonlinear effects (geeglm function in R geepack package).As a sensitivity analysis, the model was fitted with a truncated dataset for different follow-up times.Censoring at 3 and 5 years after the first rituximab dose, no significant impact in length of follow-up time on the estimated mean change over time was found.
Data were checked for outliers (above 1.5 times the upper IQR for IgG and IgM) and apparent mistyped values were corrected (IgG and IgM values mixed up; B-lymphocyte count data input as 10 6 instead of 10 [9]).No outliers were excluded.

RE SULTS
In all, 2752 patients were identified at the participating centres in Sweden treated with rituximab between 2008 and 2021.After removing five duplicates (transferring between participating centres) and two patients lost to follow-up (non-retrievable data after moving to another centre), the cohort consisted of 2745 patients (2134 women, 611 men): Stockholm n = 1869, Gothenburg n = 417 and Northern region n = 458.In total, 13,155 IgG and 8180 IgM measurements were paired with a subsequent infusion.The mean number of IgG measurements per patient was 5.0 (SD 2.6) and the mean number of IgM measurements was 3.9 (SD 2.3) (Figure S2).

Baseline characteristics and previous DMT
Demographics at the rituximab treatment start and rituximab exposure variables are summarized in Table 1.The mean age at the first rituximab dose was 42.1 (SD = 11.6) years with 77.7%

Treatment regimens
The mean time on rituximab treatment was 3.1 (1.0-5.2) years, and the mean accumulated dose of rituximab was 3830 mg (SD = 2340), displayed in Table 1.The median number of rituximab infusions was 6 (range 1-28) (Figure S3).Patients received 500-1000 mg intravenously every 6-12 months.Higher initial treatment doses were used initially, commencing in the Northern region.The dosing regimen has gradually decreased over time (Figure S4).The median start dose was 1000 g, and the mean start dose was 804 g (SD = 350).The median interval between doses was 183 to 245 days (Table S3).
TA B L E 1 Baseline characteristics and exposure variables for the hypogamma-MS cohort.

Effect on immunoglobulin levels
The mean IgG decrease for the entire cohort was 0.27 (95% confidence interval 0.17-0.36)g/L per year on rituximab therapy (Table 2) when controlling for known factors, for example pregnancy-induced immunoglobulin decrease, plasmapheresis, IVIG therapy, age and sex.The mean IgG decrease was linear over time (Figure 1a), but IgM was nonlinear (Figure 1b, Table 3).The mean IgG decrease was also analysed in relation to the accumulated dose of rituximab, with a decrease of 0.18 (95% confidence interval 0.16-0.20)g/L per 1000 mg of administered rituximab (Table 4).Inter-individual differences and IgG decrease patterns are displayed in a spaghetti plot (Figure S1).
There was a small, but significant, decrease of baseline IgG with age at inclusion (Figure 2a), and similar but nonlinear for IgM (Figure 2b).The rate of IgG decrease diminished slightly over time by 0.005 g/L per year of age (data not shown) which could be due to lower dosage protocols for older patients or patient selection.There were no statistically significant differences in IgG measurements between sexes at baseline (Tables 2 and 3).
Analysing immunoglobulin kinetics, those with the lowest baseline IgG and IgM measurements decreased slightly less than those in higher baseline groups (Figure 2a,b; Table S1A

Previous treatment and risk of low baseline IgG levels
The impact of previous DMTs for MS on baseline IgG levels before commencing rituximab treatment compared to treatment-naïve patients was adjusted for age.For those switching from fingolimod, the mean baseline IgG was 1.03 g/L lower, for natalizumab 0.80 g/L lower (Table 2) and for interferonβ-treated patients 0.40 g/L higher than for the treatment-naïve patients.For teriflunomide, glatiramer acetate or dimethyl fumarate, there was no significant difference.
For patients previously treated with dimethyl fumarate, the IgG decrease rate was 0.18 g/L/year higher, compared to untreated patients.Similarly, patients treated with interferon β experienced an additional decrease in IgG levels, of 0.14 g/L/year compared to untreated patients.

DISCUSS ION
In this large MS cohort compiling retrospective data from about half of all rituximab-treated patients in Sweden, mean IgG levels decreased linearly by 0.27 g/L per treatment year.This decrease is interpreted as relatively modest, but its linear nature suggests that it may not be a direct consequence of rituximab treatment.Instead, it could be attributed to the attrition of plasma cells, which are not adequately repopulated by novel B-cells [35,36], or by a secondary or downstream effect.Identifying the factors that expedite this plasma cell attrition could provide critical insights for mitigating the risk of diminished IgG levels in patients treated with rituximab.Even

TA B L E 1 (Continued)
with treatment interruption, immunoglobulins remained low or even continued to decrease, and from our clinical experience it may take years for IgG to begin increasing.
There were no differences in immunoglobulin decrease between sexes, despite known differences in drug pharmacokinetics [37].
Furthermore, a small but significant association between age and lower IgG levels at baseline was detected, driven by the patients over 50 years, contrasting with healthy individuals, where older individuals exhibit higher levels of IgG [38].
The IgG decrease during rituximab therapy appeared linear in our population.This agrees with other studies in which starting with a lower value increased the risk of reaching LLN [36,39].Surprisingly,  In the present study, natalizumab and fingolimod were identified as predictors for reduced baseline IgG levels in patients with MS.However, it is worth noting that the mechanisms underlying these reductions appear to vary between treatments.Natalizumab has no known effect on B-cell function besides migration properties.However, other MS studies have reported immunoglobulin decrease during treatment with natalizumab for MS [11,[40][41][42].Possible mechanisms presented have been a natalizumab-induced alteration of B-cell gene expression, blocking of VLA4-mediated interaction of plasma cells in the bone marrow [43], epigenetic changes [42] or interference with B-cell homing to peripheral lymphoid tissue [41].Also, significantly lower mean IgG and IgM are noted for fingolimod compared to healthy individuals [11].Despite these findings, current guidelines do not recommend routine monitoring of IgG levels during treatment with natalizumab and fingolimod.We argue that monitoring immunoglobulin levels in MS therapies may be extended beyond anti-CD20 therapies.
Hypogammaglobulinaemia secondary to immunosuppressive therapies is increasingly reported in MS and other autoimmune  The comparison group is previously untreated pwMS.
glycoprotein antibody disorder or NMOSD, 17% of patients experienced hypogammaglobulinaemia (LLN <6.0 g/L), with a median of 5.4 years to hypogammaglobulinaemia onset [20].In RA, hypogammaglobulinaemia and risk of infections are recognized complications of long-term B-cell depleting therapies and are reported in up to 24% of patients [6,8,39].
The data from our cohort indicate a lower rate of IgG decline compared with the NMOSD and RA studies, which possibly in part could be explained by dose reductions and extended infusion intervals at Swedish centres [2,31,32].In the rituximab studies for RA, concomitant medication with methotrexate and oral steroids can contribute to this risk [8].Higher accumulated doses of rituximab and not allowing B-cell repopulation may account for a higher risk in NMOSD treatment [11,16,41,45].In NMOSD, there were no available data on other concomitant medications, for example steroids or other immunosuppressants, possibly contributing to IgG decrease [7].
It is essential to manage the increased infection risk in hypogammaglobulinaemia during rituximab therapy [7,16,44].The Swedish real-world prospective cohort by Luna et al. [17]  In a French observational study [16] assessing the risk of hypogammaglobulinaemia and infections in 188 MS patients with rituximab treatment, IgG levels <6 g/L were significantly associated with a risk of developing infections.Compared to our study, a higher proportion developed IgG <6 g/L (23.4%) and <4 g/L (4.2%).The number of doses was comparable; however, the median EDSS was higher than in our cohort [16].
Ocrelizumab and ofatumumab [26] were not associated with hypogammaglobulinaemia in the MS registration studies OPERA [46] and ASCLEPIOS [47].The OPERA ocrelizumab 5-year ex- levels remained relatively stable [45].A study comparing rituximab and ocrelizumab during the first year found lower IgG levels in the ocrelizumab group compared to the rituximab group [30].Notably, LLN was >1 g/L lower in the ocrelizumab and ofatumumab trials than in our study [46][47][48].
The presented lower risk for IgG decrease for ofatumumab in the ALITHIOS [48] and ASCLEPIOS [47] studies compared with our data may be an effect of a lower anti-CD20 antibody dosage.However, it may in part also be explained by differences in follow-up time and patient selection.Moreover, steroids are not used when administering ofatumumab, which could influence the risk of reducing IgG values [47].

Strengths and weaknesses
A strength of this real-world retrospective study is the large number Future studies need to establish a dosing schedule balancing the risks of negative immunological consequences and treatment efficiency, establishing the optimal B-cell repopulation before redosing and making pre-treatment assessments to establish which patients are at risk of secondary hypogammaglobulinaemia.
In an ongoing study, the infection risk in an extended version of this cohort will be analysed in relation to immunoglobulin dynamics, B-cell subset repopulation patterns and late-onset neutropenia.
Patient characteristics, for example age at first rituximab infusion, sex, disease type (relapsing-remitting, RRMS; secondary progressive, SPMS; and primary progressive, PPMS), disease duration and Expanded Disability Status Scale (EDSS) level at inclusion were collected from SMSreg, with information of doses and dates of previous and current DMTs.Laboratory results were obtained from patient charts, collecting all available immunoglobulin G (IgG) and immunoglobulin M (IgM) measurements.Laboratory parameters were paired with a subsequent infusion within 45 days.Censoring Patients receiving IVIG infusions for any cause were right censored, contributing with time on treatment and laboratory results until the start of IVIG.Females with a registered pregnancy contributed with time on treatment and laboratory results until conception.Patients transferred to a non-participating centre were censored at the transfer date.Follow-up was ended on the day of the last registered rituximab dose.Patients in the Northern, Gothenburg and Stockholm Danderyd cohorts were right censored on 25 January 2021 and in the Stockholm Karolinska/ASC cohort on 14 May 2020.
els in relation to time on rituximab treatment.Secondary outcomes were IgG and IgM decrease per 1000 mg of accumulated dose of rituximab and IgG and IgM decrease in relation to the measured covariates baseline IgG value, previous exposure to immunomodulatory drugs, age when starting rituximab treatment, centre and sex.
ables (age, sex, centre, previous DMT) were analysed by linear regression at baseline, adjusting for covariates.Mean values were used to describe normally distributed data with standard deviations (SD).Medians and interquartile ranges (IQRs) were used to describe non-normally distributed values.p values under 0.05 were considered significant.Statistical analyses were performed in R 4.2.3.Generalized estimating equation models were used to examine the relationships between the outcomes IgG and IgM levels and the main exposure time on rituximab (in years) whilst adjusting for additional predictors, including age, sex, previous DMTs and centre.Confidence intervals were corrected for repeated measurements on individuals through robust 'sandwich' standard errors, implemented by setting the correlation matrix to 'independence'.
individuals with higher baseline IgG and IgM values tended to experience a greater rate of reduction of IgG and IgM over time in our study.There were no significant differences in follow-up dropout rates in relation to baseline IgG level.

F I G U R E 1
Serum levels of IgG (a) and IgM (b) decline as a function of time on treatment with rituximab.The boxes display the median and interquartile (IQR) range at each time point.Whiskers denote 1.5 × IQR and dots indicate outliers or individual values outside this range.The dashed black line indicates the lower limit of normal (LLN) for IgG at our laboratory (6.7 g/L).The red dash-dot line in (a) indicates the level of severe hypogammaglobulinaemia (<4 g/L) and in (b) LLN (0.27 g/L).diseases[7,44].Repeated dosing of anti-CD20 therapies not allowing for B-cell repopulation has been associated with a more rapid decrease of immunoglobulin levels than individual doses.Several recent MS and NMOSD studies report a correlation between time on treatment, treatment dose and the risk of developing hypogammaglobulinaemia[6,9,19,20,39].IgG decrease rates of approximately 0.4 g/L yearly are described with induction protocols (2000 mg of rituximab) and re-treatment immediately after B-cell repopulation[9,19,20].In a 14-year NMOSD follow-up study by Kim et al.[19], 41% experienced IgG measurements below <6.0 g/L (assigned LLN).A majority (11/15) of participants developed IgG levels below <7.0 (assigned LLN) during a 70-month median follow-up in an Italian NMOSD cohort[9].With similar rituximab treatment for up to 3.7 years for myelin oligodendrocyte TA B L E 3 Analyses of factors associated with decline of IgM during rituximab treatment assuming time on treatment as an independent variable.

F I G U R E 2
Individuals with higher baseline IgG (a) or IgM (b) tended rather to decline faster in IgG upon a similar rituximab challenge over time compared to individuals with lower baseline values.The boxes display the median and interquartile range (IQR) at each time point.Whiskers denote 1 .5 × IQR and dots indicate outliers.The dashed black line indicates the lower limit of normal (LLN) for IgG at our laboratory (6.7 g/L).The red dash-dot line in (a) indicates the level of severe hypogammaglobulinaemia (<4 g/L) and in (b) LLN (0.27 g/L).
of patients included.Including virtually all patients from three major geographical areas of Sweden treated with rituximab increased the external validity.Data quality was improved by high data density, few switches to other therapies and infusions within a 6-or 12-month timeframe, although the different prevailing infusion schemes over time remain a weakness.Future studies can evaluate whether deescalation strategies reduce the risk of developing hypogammaglobulinaemia.In this real-world setting, there may be different reasons for treatment discontinuation.Only a few patients were followed up on treatment for more than 5 years.Selection bias and healthy survivor bias are some potential confounders in this observational study setting.Patients remaining on therapy may have been selected for not developing side effects, for example frequent infections.No differences in baseline IgG measurements were found between the whole cohort and those continuing therapy.This strengthens our result of a uniform moderate and gradual decrease in mean IgG levels.CON CLUS IONAssessment of possible hypogammaglobulinaemia is important when treating MS.The rate of diminishing immunoglobulins in a large contemporary Swedish cohort during rituximab treatment is described, finding an association with time on treatment and with the total dose of rituximab, with a linear trend towards lower means of IgG levels.At the same time, the IgM decrease is more rapid and nonlinear.The mean IgG levels decrease by 0.27 g/L per year on rituximab therapy.After treatment with rituximab therapy for 3 years and 5 years, 7.7% and 9.5% have a measurement of IgG below LLN (<6.7 g/L), respectively.Previous treatments with natalizumab and fingolimod are associated with lower mean baseline IgG values, which may predict the risk of developing secondary hypogammaglobulinaemia.Switching from interferons and dimethyl fumarate was associated with a more rapid IgG decline, at present of unknown reason and clinical significance. ,B).
developed severe hypogammaglobulinaemia (IgG <4.0 g/L).Low IgM levels were experienced by 8.3%, with a nadir value below LLN (<0.27 g/L).In patients on rituximab therapy for 3 and 5 years, 7.7% and 9.5% had nadir measurements of IgG below LLN, respectively.Sixteen patients received IVIG infusions.Eleven individuals received repeated IVIG therapy for hypogammaglobulinaemia or infections and five patients received a single treatment and then continued rituximab infusions.There was a trend for previously natalizumab-and fingolimod-treated patients to reach IgG measurements below LLN earlier.No statistically significant difference in time to IgG levels below LLN or severe hypogammaglobulinaemia between sexes was seen.
Analyses of factors associated with decline of IgG during rituximab treatment assuming time on treatment as an independent variable.The reference group for sex is females, and for age, it is the lowest quantile.The statistical model employed, Generalized Estimating Equations (GEE), was chosen due to its robustness in handling the correlated data typical of longitudinal studies, assuming linear relationships among variables.A significance level of p < 0.05 was adopted, but given the 23 comparisons made, a Bonferroni-adjusted threshold of 0.002 (0.05/23) could be used to mitigate the risk of Type I error.This adjustment necessitates a cautious interpretation of findings near this significant boundary.
B L E 2 Note: a The comparison group is female pwMS.bThecomparison group is the lowest quantile of age at inclusion.cThecomparison group is previously untreated pwMS.
Since the decline of IgM is not linear a square term "Rate of immunoglobulin change in untreated, females and age of lowest quantile" to adjust for this non-linearity.The Bonferroni-adjusted threshold of 0.002 (0.05/24) because of the added squared term.
a The comparison group is female pwMS.bThe comparison group is the lowest quantile of age at inclusion.c reported an increased risk of infections requiring hospitalization or antibiotics or antivirals, compared to healthy controls or MS patients treated with injectables, natalizumab and fingolimod.Importantly, these data do not take rituximab treatment length or accumulated dose into Mitoxantrone, and Alemtuzumab.These are consolidated due to smaller patient numbers for each treatment and are referred to as 'Other' in the dataset.Age and sex effects on baseline IgG are presented as changes in g/L.p-Values <0.05 are considered statistically significant and indicate the likelihood that the observed effects are not due to random chance.The analysis includes 2630 individuals contributing to a total of 13076 observations, allowing for a comprehensive evaluation of the impact of rituximab dosage, demographic variables, and treatment history on IgG levels.
a The comparison group is female pwMS.b The comparison group is the lowest quantile of age at inclusion.c The comparison group is previously untreated pwMS.account nor analyse the relation with IgG or B-lymphocyte values.