Despite improvements in genetic and functional studies, delayed diagnosis of common variable immunodeficiency (CVID) remains challenging. To overcome this, an exploratory study to evaluate the diagnostic performance of a panel of biomarkers for CVID, such as the sum of κ+λlight chains and the soluble B-cell maturation antigen (sBCMA) levels, switched memory B cells (smB) and VISUAL score, through logistic regression models compared to gold-standard tests (specific antibody responses) was carried out.ANOVA and bivariate analysis were performed between different groups and logistic regression models were fitted using CVID biomarkers between CVID and selective IgA deficiency (SIgAD). A total of 88 subjects were studied: 27 CVID patients, 23 SIgAD patients, 20 secondary immunodeficiency (SID) patients and 18 healthy controls. We validated the diagnostic performance of individual biomarkers sBCMA and sum κ+λ, with Se 89% and Spe 89%, versus Se 90% and Spe 99%, respectively. sBCMA strongly correlated with all other three variables (sum κ+λ, smB cell and VISUAL). By contrast, sum κ+λ did not correlate with either smB cells or VISUAL, and could provide added diagnostic value. By multivariable tree decision model, only 2 two factors proved to be independent signature biomarkers of CVID, namely specific antibody responses and sum κ+λ. The resulting model had an AUC of 0.946, Se 0.85, and Spe 0.95. The tree-decision model can increase diagnostic efficiency. Sum κ+λ stood out over other CVID classifiers, further highlighting its potential as a diagnostic criterion.
Research Article
An Exploratory Approach of Clinically Useful Biomarkers of Cvid by Logistic Regression
https://doi.org/10.21203/rs.3.rs-3630707/v1
This work is licensed under a CC BY 4.0 License
Journal Publication
published 07 Jun, 2024
Read the published version in Journal of Clinical Immunology →
You are reading this latest preprint version
Biomarkers
diagnosis
decision-tree model
CVID
serum free light chains
switched-memory B cells
sBCMA
antibody vaccine response
VISUAL score
Common variable immunodeficiency (CVID) is the most clinically prevalent antibody deficiency, affecting 1:25,000 individuals. CVID is characterized by its wide clinical heterogeneity with a roughly unknown genetic basis, which sometimes harden and delay its diagnosis1–3.
In 1999, the European Society of Immunodeficiencies (ESID) and the Pan-American Group of Immunodeficiency (PAGID) established the original criteria for CVID diagnosis focused on marked hypogammaglobulinemia and altered vaccine response, which were later revised to be more accuarte4,5. In 2008, Chapel et al. emphasized the importance of categorizing clinical manifestations into two major phenotypes: infections only; and dysregulatory complications, namely autoimmune disease, malabsorption and/or lymphoproliferation6–8. The latter phenotypes associated with low proportion of switched-memory B cells (smB), which was proposed as a predictive marker9,10. Later, several authors proposed clinical severity scores11, 12. More recently, VISUAL score showed promising as a prognostic tool at CVID diagnosis based on five independent risk predictors, including smB phenotype, serum IgA and IgM, CD4+ T cells counts, and immunization antibody responses13.
Despite improvements in genetic and functional studies, the diagnostic delay of CVID remains challenging. Significant efforts have been directed towards developing novel biomarkers, including the sum of κ + λ light chains14–16 and the soluble B-cell maturation antigen (sBCMA) 17, both of them significantly decreased in CVID compared to other primary (PID) and secondary immunodeficiencies (SID).
Here we explored the diagnostic performance of recent biomarkers with respect to gold-standard tests by multivariate logistic and regression models to better and earlier identify CVID from other immunodeficiencies and to improve personalized management.
Study subjects
Patients from the Immunology Service of Hospital Clínico San Carlos (Madrid, Spain) were retrospectively included in this study. All patients fulfilled the ESID diagnostic criteria of CVID and SIgAD, respectively. The following routine clinical data were collected for eligible patients: demographics (age, gender), clinical manifestations according to Chapel phenotypes (infections and/or inflammatory complications), clinical laboratory results (hematology, serum immunoglobulins, smB cell phenotype, sFLC, specific antibody responses, CD4+ and CD8+ T lymphocyte counts), VISUAL score at diagnosis, imaging findings and final clinical diagnosis. Specific antibody responses to pure polysaccharide antigen (Typhim Vi, Sanofi Pasteur MSD), and to protein antigen after the conjugated anti-pneumococcal vaccine (Prevenar, Sanofi Pasteur MSD Limited), and tetanus-toxoid vaccine (Diftavax, Sanofi Pasteur) were evaluated. Serum samples pre- and post-vaccination were collected for all patients and evaluated using commercial ELISA kits (The Binding Site Group Ltd, Birmingham, UK).
Approval for the study was obtained from the hospital institutional Ethics Committee for PID and SID projects (19-284-E and 19/219-E), respectively. Written informed consent was obtained from all patients for inclusion in the study protocol.
Evaluation of sBCMA levels
Serum BCMA levels were determined determined using an ELISA- based assay with polyclonal anti-BCMA antibodies (Bio-Techne, R&D Systems, Minnesota, USA).
Evaluation of serum free light chains
sFLC (κ and λ) levels were quantified by nephelometry (FREELITE, The Binding Site Group Ltd., Birmingham, UK), according to the manufacturer’s instructions.
Statistical analysis
Descriptive and bivariate association analyses were performed using standard statistics and tests selected based on variable characteristics. When analysis of variance revealed significant differences, pairwise post-hoc analyses were conducted using Tukey's Honest Significant Difference test. The diagnostic value of individual variables to differentiate CVID from SIgAD was assessed with a series of logistic regression models, with the clinical group as the dependent variable. The performance of the models (area under the curve [AUC], sensitivity, and specificity) was evaluated through leave-one-out cross-validation (LOOCV). Pearson's linear correlations between quantitative predictors were assessed for both the combination of the CVID and SIgAD groups and separately for each of the two groups. Finally, a decision tree for distinguishing between CVID and SIgAD was constructed using the C5.0 algorithm and validated with the LOOCV method. Statistical analyses were conducted using R software (version 4.3.1) and the following packages: caret (6.0–94), C50 (0.1.8), and GGally (2.1.2).
Epidemiological characteristics of patients
We studied 88 subjects: 70 immunodeficiency patients from the Immunology Department of Hospital Clínico San Carlos (Madrid, Spain) diagnosed with CVID (n = 27), selective IgA deficiency (SIgAD) (n = 23) and SID with active haematological malignancy (chronic lymphocytic leukemia, CLL or multiple myeloma, MM) (n = 20). Eighteen healthy controls (HC) were also studied for sBCMA levels.
No significant differences were found in terms of sex. Proportion of women in the different groups was as follows: 55.56% in CVID; 56.52% in SIgAD; 60.00% in SID; and 61.11% in HC. Median age was as follows: CVID 50.19 years (from 19 to 81); SIgAD, 46.70 years (from 16 to 86); SID, 70.85 years (from 53 to 87); HC, 39.79 years (from 22 to 70). Age of the SID group was significantly higher than the remaining groups (p < 0.05).
sBCMA and sum κ + λ levels were significantly reduced in CVID patients compared to SIgAD, SID patients and healthy controls
sBCMA levels were significantly decreased in CVID patients (median 5.97) compared to SIgAD patients (median 23.30, p < 0.001), to SID patients (median 56.75, p < 0.001) and to HC (median 25.20, p < 0.001) (Fig. 1A). No differences were observed neither between SIgAD patients and controls nor between SID patients and controls.
Similarly, sum κ + λ was significantly decreased in CVID patients (median 9.10) compared to SIgAD patients (median 31.45, p < 0.001) and to SID patients (median 33.20, p < 0.001) (Fig. 1B).
We then evaluated the diagnostic performance by ROC curve analysis of these recent biomarkers compared to gold standard antibody responses and to prognostic biomarkers, specifically the proportion of smB cells and the VISUAL score in our cohort of CVID and SIgAD patients. The vaccine antibody response was evaluated by applied a Pearson's chi-square test (p < 0.001) (Table 1). As shown, all parameters showed adequate diagnostic performance, being the best two classifiers in our cohort antibody response and sum κ + λ.
|
Biomarker |
AUC ROC curve |
Se |
Sp |
Cut-off |
|---|---|---|---|---|
|
Specific Ab response |
NA |
100% |
78.95% |
NA |
|
smB phenotype |
0.9253 (CI95 0.8447-1.000) |
69.23% |
94.12% |
≤ 2.00 |
|
VISUAL score |
0.8952 (IC95 0.7836-1.000) |
80.00% |
95.24% |
> 10.00 |
|
sBCMA |
0.9452 (IC95 0.8838-1.000) |
88.89% |
86.96% |
≤ 15.00 |
|
∑κ + λ |
0.9380 (IC95 0.8616-1.000) |
88.00% |
100% |
< 20.20 |
| NA: not applicable | ||||
sBCMA levels correlate with other diagnostic and prognostic markers of CVID
We analyzed the relationship between smB phenotype, sBCMA, sum κ + λ and VISUAL score, and the occurrence of CVID diagnosis by Pearson’s correlation coefficient. Statistically significant correlation coefficients (r) were obtained for four parameters (Table 2). As shown, sBCMA positively correlated with sum κ + λ (r = 0.63) and smB phenotype (r = 0.407); while inversely with VISUAL score (r = -0.58). smB strongly correlated with VISUAL (r = 0.60). Sum κ + λ did not correlate with smB cells or VISUAL.
|
sBCMA |
||||
|---|---|---|---|---|
|
Sum κ + λ |
P |
0.6310 |
||
|
p value (bilateral) |
< 0.0001 |
|||
|
N |
50 |
Sum κ + λ |
||
|
smB |
P |
0.4078 |
0.1455 |
|
|
p value (bilateral) |
0.001 |
0.18 |
||
|
N |
50 |
50 |
smB |
|
|
VISUAL |
P |
-0.5873 |
-0.2959 |
-0.6054 |
|
p value (bilateral) |
< 0.0001 |
0.06 |
< 0.0001 |
|
|
N |
50 |
50 |
50 |
|
Clinical spectrum of patients. Characteristics of outliers and their relationship to prognostic biomarkers
We then evaluated whether outliers for sBCMA > 15.00ng/mL and the sum κ + λ ≥ 20.20 in 5 patients could predict severe clinical evolution, as shown in Table 3. In particular, 3 of the 5 patients had haematological cancer. Serum IgA values at diagnosis were significantly higher in CVID patients with sBCMA and sum κ + λ outliers compared with the remaining CVID patients (p = 0.0162; p = 0.0250, respectively).
Table 3. - Further clinical evaluation of patients with outliers, including those in the IgAD patient cohort.
|
|
|
|
|
Infections |
Autoimmune disease |
Enteropathy |
Malignancy |
|
|
CVID |
High sBCMA |
Pt #1 |
15.90 |
|
|
|
|
|
|
Pt #2 |
21.50 |
|
|
|
|
|||
|
Pt #3 |
19.70 |
|
|
|
|
|||
|
Sum κ+λ |
34.90 |
|||||||
|
Pt #4 |
22.00 |
|
|
|
|
|||
|
Pt #5 |
52.60 |
|
|
|
|
|||
|
IgAD |
Low sBCMA |
Pt #6 |
5.38 |
|
|
|
|
|
|
Pt #7 |
14.40 |
|
|
|
|
|||
|
Pt #8 |
14.80 |
|
|
|
|
|||
By contrast, 3 SIgAD patients disclosed sBCMA values below the cut-off of 15, which might suggest progression to CVID, to be confirmed in prospective follow-up. Of note, the 3 SIgAD patients showed significantly lower IgM at diagnosis than the remaining SIgAD cohort (p = 0.0457).
As expected, all CVID patients with VISUAL > 10 (n = 16; 72.72%) had severe complications according to clinical severity scores of Ameratunga11 and Grimbacher12. CVID patients with smB ≤ 2% (n = 8, 30.77%) presented with autoimmune disease, enteropathy and/or lymphoproliferative disease.
Multiple regression and best predicting decision algorithm.
As a final step, we confronted all the studied biomarkers in our cohort of CVID and SIgAD patients to assess the relevance of their contribution in the definition of the disease and to develop the best-fit algorithm for CVID diagnosis without pre-selection. Multivariate regression was tested through decision-tree model, and then validated using Leave-One-Out Cross-Validation (LOOCV).
The parameters that optimized the diagnostic performance of the C5.0 decision tree, according to LOOCV, were as follows: method = tree, window = false, and trials = 10. The resulting model had an AUC of 0.946, sensitivity of 0.85, and specificity of 0.95. The decision tree obtained in the first trial of the C5.0 algorithm included the global response to vaccines and the sum κ + λ, as shown in Fig. 2. A decision tree composed of global response to vaccines, smB percentage, and sBCMA yielded slightly improved results, albeit with a significant increase in complexity.
Laboratory findings can help to delimit the diagnosis in cases with unclear or atypical features, but they can also lead to misdiagnosis. For instance, absence of vaccine responses, considered as a major criterion, arises measurement and interpretation difficulties, making it challenging to be used alone2,4,18. The need to overcome these difficulties led us to analyze the individual performance versus combinations of biomarkers for CVID identification in our cohort of patients12.
The novelty of this exploratory works lies on the use of logistic regression to develop an accurate diagnostic model for CVID. Even today, CVID diagnosis is based on exclusion criteria without any findings considered pathognomonic of the disease. Timely diagnosis and prompt and effective treatment are extremely important for patients, reducing morbidity and mortality19,20. By comparing several classifiers, we were able to determine the optimal model for the data. Overall, the tree decision model was the most consistent, with specific antibody responses and sum κ + λ showing ideal diagnostic performance. Sum κ + λ was one of the two best CVID classifiers, highlighting its potential as a diagnostic criterium for the disease and its added value to current tests.
We first validated previous studies on the diagnostic value of sBCMA17 and sum κ + λ14–16 for CVID against other PID and SID patients. Interestingly, increase in any of sBCMA and/or sum κ + λ above cut-off levels might alert of inflammatory or lymphoproliferative complications during follow-up and personalize management.
Our results may extend the understanding of CVID pathophysiology and interdependencies by condensation of the information of the clustered biomarkers. Interestingly, a strong association was found between sBCMA and sum κ + λ, two apparently different biological events. Very low sum κ + λ suggest an alteration during early bone marrow B cell ontogeny at pro-B to pre-B cell stages, whose etiology remains unclear21. By contrast, sBCMA expression is restricted to advanced B cell maturation stages towards full plasma cell (PC) differentiation. Increased sBCMA values have been related to conditions where an expansion of these cell types occur, whereas its decrease may identify severe humoral immunodeficiencies22,23. This association might suggest that alterations at early maturation stages within the bone marrow have repercussions in ulterior activation and differentiation stages. In this context, CVID typically shows slow and progressive clinical and immunological onset, eventually from SIgAD, whereby these biomarkers may represent a real-time alert sign of progression24. Although sBCMA expression is weak in pre-PC maturation stages, we observed a significant correlation between sBCMA values and smB percentage, further supporting an alteration at germinal center.
Our results should be interpreted with caution due to the limited sample size and need to be validated in a wider prospective study. However, the results were quite uniform and the ideal cut-off levels validated those previously described by other authors.
In conclusion, our exploratory study suggests that combining the measurement of specific antibody responses to immunization with sum κ + λ enhances early CVID diagnosis. This integrated approach could generate patient-specific, personalized signatures for CVID diagnosis and prognosis.
|
Abbreviation |
Definition |
|
AUC |
Area under the curve |
|
CLL |
Chronic lymphocytic leukemia |
|
CVID |
Common variable immunodeficiency |
|
ELISA |
Enzyme-linked immunosorbent assay |
|
ESID |
European Society of Immunodeficiencies |
|
HC |
Healthy controls |
|
LOOCV |
Leave-One-Out Cross-Validation |
|
MM |
Multiple myeloma |
|
PAGID |
Pan-American Group of Immunodeficiency |
|
PC |
Plasma cell |
|
PID |
Primary immunodeficiencies |
|
ROC curve |
Receiver operating characteristic curve |
|
sBCMA |
Soluble B-cell maturation antigen |
|
sFLC |
Serum free light chains |
|
SID |
Secondary immunodeficiencies |
|
SIgAD |
Selective IgA deficiency |
|
smB |
Switched-memory B cells |
Acknowledgments
We are immensely grateful to the patients and controls that made possible this study. The authors thank Marina Lorne for graphical abstract design.
Statements and Declarations
Ethics Approval
Approval for the study was obtained from the hospital institutional Ethics Committee for PID (project #19 284 E) and SID projects (project #19/219 E), respectively.
Consent to Participate
Written informed consent was obtained from all patients for inclusion in the study protocol.
Consent for Publication
Not applicable.
Conflict of Interest
The authors declare no competing interests.
Author Contribution
TG-G and SS-R conceived the study, collected data, and drafted manuscript. AJ-H performed logistic regression model. All authors (TG-G, MP-O, AJ-H, KG-H, MCC, AV-M, MDM-R, NS, EDF, ARP, MG-F, MF-A, RPD, SS-R) provided critical revisions of the manuscript and final approval of the version to be published. All authors contributed to the article and approved the submitted version.
Funding
The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.
Availability of data and materials
-Not applicable
-The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.
- Tangye SG, Al-Herz W, Bousfiha A, Cunningham-Rundles C, Franco JL, Holland SM, Klein C, Morio T, Oksenhendler E, Picard C, Puel A, Puck J, Seppänen MRJ, Somech R, Su HC, Sullivan KE, Torgerson TR, Meyts I. Human Inborn Errors of Immunity: 2022 Update on the Classification from the International Union of Immunological Societies Expert Committee. J Clin Immunol. 2022;42(7):1473–507.
- Dahl C, Petersen I, Ilkjær FV, Westh L, Katzenstein TL, Hansen AE, Nielsen TL, Larsen CS, Johansen IS, Rasmussen LD. Missed Opportunities to Diagnose Common Variable Immunodeficiency: a Population-Based Case-Control Study Identifying Indicator Diseases for Common Variable Immunodeficiency. J Clin Immunol. 2023 Sep 28.
- Ameratunga R, Woon ST, Gillis D, Koopmans W, Steele R. New diagnostic criteria for common variable immune deficiency (CVID), which may assist with decisions to treat with intravenous or subcutaneous immunoglobulin. Clin Exp Immunol. 2013;174(2):203–11.
- Conley ME, Notarangelo LD, Etzioni A. Diagnostic criteria for primary immunodeficiencies. Representing PAGID (Pan-American Group for Immunodeficiency) and ESID (European Society for Immunodeficiencies). Clin Immunol. 1999;93(3):190–7.
- Seidel MG, Kindle G, Gathmann B, Quinti I, Buckland M, van Montfrans J, Scheible R, Rusch S, Gasteiger LM, Grimbacher B, Mahlaoui N, Ehl S, ESID Registry Working Party and collaborators. The European Society for Immunodeficiencies (ESID) Registry Working Definitions for the Clinical Diagnosis of Inborn Errors of Immunity. J Allergy Clin Immunol Pract. 2019 Jul-Aug;7(6):1763–70.
- Chapel H, Lucas M, Lee M, Bjorkander J, Webster D, Grimbacher B, Fieschi C, Thon V, Abedi MR, Hammarstrom L. Common variable immunodeficiency disorders: division into distinct clinical phenotypes. Blood. 2008;112(2):277–86.
- Chapel H. Common Variable Immunodeficiency Disorders (CVID) - Diagnoses of Exclusion, Especially Combined Immune Defects. J Allergy Clin Immunol Pract. 2016 Nov-Dec;4(6):1158–9.
- Chapel H, Lucas M, Patel S, Lee M, Cunningham-Rundles C, Resnick E, Gerard L, Oksenhendler E. Confirmation and improvement of criteria for clinical phenotyping in common variable immunodeficiency disorders in replicate cohorts. J Allergy Clin Immunol. 2012;130(5):1197–1198e9.
- Wehr C, Kivioja T, Schmitt C, Ferry B, Witte T, Eren E, Vlkova M, Hernandez M, Detkova D, Bos PR, Poerksen G, von Bernuth H, Baumann U, Goldacker S, Gutenberger S, Schlesier M, Bergeron-van der Cruyssen F, Le Garff M, Debré P, Jacobs R, Jones J, Bateman E, Litzman J, van Hagen PM, Plebani A, Schmidt RE, Thon V, Quinti I, Espanol T, Webster AD, Chapel H, Vihinen M, Oksenhendler E, Peter HH, Warnatz K. The EUROclass trial: defining subgroups in common variable immunodeficiency. Blood. 2008;111(1):77–85.
- Sánchez-Ramón S, Radigan L, Yu JE, Bard S, Cunningham-Rundles C. Memory B cells in common variable immunodeficiency: clinical associations and sex differences. Clin Immunol. 2008;128(3):314–21. 10.1016/j.clim.2008.02.013. Epub 2008 Jul 11.
- Ameratunga. Assessing Disease Severity in Common Variable Immunodeficiency Disorders (CVID) and CVID-Like Disorders. Front Immunol. 2018;9:2130.
- Yong PFK, Thaventhiran JED, Grimbacher B. A rose is a rose is a rose’, but CVID is Not CVID common variable immune deficiency (CVID), what do we know in 2011? Adv Immunol. 2011;111:47–107.
- Guevara-Hoyer K, Jiménez-Huete A, Vasconcelos J, Neves E, Sánchez-Ramón S. Variable immunodeficiency score upfront analytical link (VISUAL), a proposal for combined prognostic score at diagnosis of common variable immunodeficiency. Sci Rep. 2021;11(1):12211.
- Unsworth DJ, Wallage MJ, Sarkar E, Lock RJ. Abnormalities of serum-free light chain in patients with primary antibody deficiency in the absence of B lymphocyte clonality. J Clin Pathol. 2012;65:1128–31. 10.1136/jclinpath-2012-201044.
- Scarpa R, Pulvirenti F, Pecoraro A, Vultaggio A, Marasco C, Ria R, Altinier S, Compagno N, Firinu D, Plebani M, De Carli M, Matucci A, Vianello F, Vacca A, Spadaro G, Quinti I, Agostini C, Milito C, Cinetto F. Serum Free Light Chains in Common Variable Immunodeficiency Disorders: Role in Differential Diagnosis and Association With Clinical Phenotype. Front Immunol. 2020;11:319. 10.3389/fimmu.2020.00319.
- Guevara-Hoyer K, Ochoa-Grullón J, Fernández-Arquero M, Cárdenas M, Pérez de Diego R, Sánchez-Ramón S. Serum Free Immunoglobulins Light Chains: A Common Feature of Common Variable Immunodeficiency? Front Immunol. 2020;11:2004.
- Maglione PJ, Ko HM, Tokuyama M, Gyimesi G, Soof C, Li M, Sanchez E, Chen H, Radigan L, Berenson J, Cunningham-Rundles C. Serum B-Cell Maturation Antigen (BCMA) Levels Differentiate Primary Antibody Deficiencies. J Allergy Clin Immunol Pract. 2020;8(1):283–291e1.
- Ameratunga R, Edwards ESJ, Lehnert K, Leung E, Woon ST, Lea E, Allan C, Chan L, Steele R, Longhurst H, Bryant VL. The Rapidly Expanding Genetic Spectrum of Common Variable Immunodeficiency-Like Disorders. J Allergy Clin Immunol Pract. 2023;11(6):1646–64.
- Cabañero-Navalon MD, Garcia-Bustos V, Nuñez-Beltran M, Císcar Fernández P, Mateu L, Solanich X, Carrillo-Linares JL, Robles-Marhuenda Á, Puchades-Gimeno F, Pelaez Ballesta A, López-Osle N, Torralba-Cabeza MÁ, Bielsa Masdeu AM, Diego Gil J, Tornador Gaya N, Pascual Castellanos G, Sánchez-Martínez R, Barragán-Casas JM. González-García A, Patier de la Peña JL, López-Wolf D, Mora Rufete A, Canovas Mora A, Forner Giner MJ, Moral Moral P. Current clinical spectrum of common variable immunodeficiency in Spain: The multicentric nationwide GTEM-SEMI-CVID registry. Front Immunol. 2022;13:1033666.
- Ramirez NJ, Posadas-Cantera S, Caballero-Oteyza A, Camacho-Ordonez N, Grimbacher B. There is no gene for CVID - novel monogenetic causes for primary antibody deficiency. Curr Opin Immunol. 2021;72:176–85.
- Winkler TH, Mårtensson IL. The Role of the Pre-B Cell Receptor in B Cell Development, Repertoire Selection, and Tolerance. Front Immunol. 2018;9:2423.
- Cho S-F, Anderson KC, Tai Y-T. Targeting B Cell Maturation Antigen (BCMA) in Multiple Myeloma: Potential Uses of BCMA-Based Immunotherapy. Front Immunol. 2018;9:1821.
- Yu B, Jiang T, Liu D. BCMA-targeted immunotherapy for multiple myeloma. J Hematol Oncol. 2020;13(1):125.
- Seppänen M, Aghamohammadi A, Rezaei N. Is there a need to redefine the diagnostic criteria for common variable immunodeficiency? Expert Rev Clin Immunol. 2014;10(1):1–5.
