Page 70 - 2022_03-Haematologica-web

O. Benjamini et al.
CoV-2), which was both safe and highly effective. In December 2020, results of a clinical trial using the BNT162b2 mRNA COVID-19 vaccine in a large cohort of people (≥16 years old) demonstrated a 95% efficacy in pre- venting symptomatic infection.2 This study prompted an emergency use authorization from the Food and Drug Administration.3 In a real-world setting, nationwide vacci- nation data from Israel documented high efficacy of the vaccine even in patients with multiple comorbidities.4 However, the trial excluded immunosuppressed patients, as their immune response to vaccination is usually blunted.
Chronic lymphocytic leukemia (CLL) is associated with varying degrees of immune deficiency due to the primary disease or to the therapy administered. These include functional defects of B- and T-lymphocytes, natural killer cells, neutrophils and macrophages,5 as well as hypogam- maglobulinemia, which is evident in treatment-naïve patients and develops even more frequently following therapy for CLL.5,6
Large, retrospective studies from both Europe7,8 and the USA9 have already shown that patients with CLL have an increased rate of COVID-19 infection, a higher hospital- ization rate and a greater risk of dying from the virus irre- spective of disease stage or prior treatment status. The role of vaccination in these patients is therefore of major importance. However, several studies have already shown that patients with CLL mount limited responses to other more frequently used vaccines, such as those for influen- za,10 pneumococcal infection11 and varicella zoster.12 Furthermore, only limited data are available on the response to vaccines in the era of novel biological agents now used to treat CLL, such as BTK and BCL2 inhibitors in combination with anti-CD20 antibodies.13-15 A previous publication described a reduced serological response rate to the BNT162b2 vaccine in patients with CLL, compared to that in healthy controls, particularly during therapy.16
The aim of the study we report here was to investigate the safety and efficacy of the BNT162b2 mRNA COVID- 19 vaccine in patients with CLL and the effect of therapy on the serological response to the vaccine, given in nine medical centers in different parts of Israel.
Methods
Patients
This was a prospective, interventional, multicenter study that was performed on behalf of the Israeli CLL study group. The data retrieved from nine centers in Israel provid- ed information on a total of 400 CLL patients who had been vaccinated with the BNT162b2 mRNA COVID-19 vaccine.
The diagnosis of CLL was established according to International Workshop CLL criteria.5 Data were collected from medical records after approval from all of the individ- ual institutes’ ethics committees and all patients who par- ticipated gave written informed consent. The study was registered at ClinicalTrials.gov NCT04862806 The referring physicians collected demographic and clinical data from the patients’ medical records, including patients’ characteristics (gender, age, Binet stage, medical history including Cumulative Illness Rating Scale [CIRS] score), previous treatments (number and type), and biological features of the CLL (serum IgG, IgA, IgM levels, IGHV mutation status, fluorescence in situ hybridization [FISH] cytogenetic profile, and TP53 mutation) whenever available.
Vaccination and immune response assessment
All patients received two 30 mg doses of BNT162b2 vac- cine (Pfizer), administered intramuscularly 3 weeks apart. Blood samples for immune response evaluation to the vac- cine were assessed 2-3 weeks (median, 19 days) after the second dose using anti-spike (S) antibody tests. Anti-spike antibody tests were performed in each hospital using one of three available commercial kits: The Liaison SARS-CoV-2 S1/S2 IgG (Diasorin, Saluggia, Italy), with a positive cutoff of >15 U/mL; the Architect AdviseDx SARS-CoV-2 IgG II (Abbot, Lake Forest, IL, USA), with a positive cutoff of >50 U/mL. and an enzyme-linked immunosorbent assay (ELISA) that detects IgG antibodies against the receptor- binding domain (RBD) of SARS-CoV-2 (positive value >1.1; range 1.1-10).17,18
A surrogate viral assay was used to test antiviral humoral response based on a highly infectious recombinant vesicu- lar stomatitis virus (VSV) bearing the SARS-CoV-2 spike glycoprotein S. This recombinant virus, rVSV-SARS-CoV-2 or SARS-CoV-2 pseudo-virus (psSARS-2), closely resembles SARS-CoV-2 in its entry-related properties. The psSARS-2 neutralization assay was performed using a propagation- competent VSV-spike similar to the one previously pub- lished, which was kindly provided by Gert Zimmer (University of Bern, Switzerland).19
Safety
On the day of the serological test, patients were asked to report any adverse events and filled in a questionnaire relat- ed to the development of local and systemic adverse events. Patients reported in free text if they had had any adverse events after either vaccination and answered a multiple- choice questionnaire with a scale from zero to five, where zero indicated the lack of any adverse events.
Statistical analysis
The characteristics of IgG responders and IgG non- responders were compared using the Mann-Whitney test for continuous variables, while the Wald c2 test was used for the comparison of categorical variables. Some contin- uous variables were also tested as categorical variables using the thresholds indicated in the tables and text. A P value <0.05 was considered statistically significant.
For the predictive model, multivariate logistic regression was used to predict the response to the vaccine and deter- mine which variables were independently associated with the response. Least absolute shrinkage and selection oper- ator (LASSO) regularization was used to avoid over-fitting and obtain a simpler model which consists of only the informative variables while disregarding the remaining variables. We compared the predictive performance of the LASSO logistic regression model with the simple seven- factor score. We applied ten repeats of stratified 10-fold cross-validation to estimate the various predictive per- formance metrics (area under the curve, accuracy, speci- ficity and sensitivity) and their variance. This procedure helps to avoid over-estimation of the predictive perform- ance of LASSO logistic regression.
LASSO estimates regression coefficients by maximizing the log-likelihood function, like any other logistic regres- sion, but by adding a constraint that the sum of the absolute values of the regression coefficients is less than or equal to a positive constant.20,21 Thus, LASSO prefers a par- simonious model, penalizing models with too many vari- ables. In particular, if there is a subset of highly correlated
626
haematologica | 2022; 107(3)