Definition of factors associated with negative antibody response after COVID-19 vaccination in patients with hematological diseases

COVID-19 in patients with hematological diseases is associated with a high mortality. Moreover, preventive vaccination demonstrated reduced efficacy and the knowledge on influencing factors is limited. In this single-center study, antibody levels of the SARS-CoV-2 spike protein were measured ≥ 2 weeks after 2nd COVID-19 vaccination with a concentration ≥ 0.8 U/mL considered positive. Between July and October 2021, in a total of 373 patients (median age 64 years, 44% women) with myeloid neoplasms (n = 214, 57%), lymphoid neoplasms (n = 124, n = 33%), and other diseases (n = 35, 10%), vaccination was performed with BNT162b2 (BioNTech), mRNA-1273 (Moderna), ChADOx1 (AstraZeneca), or a combination. A total of 229 patients (61%) were on active therapy within 3 months prior vaccination and 144 patients (39%) were previously treated or treatment naïve. Vaccination-related antibody response was negative in 56/373 patients (15%): in 39/124 patients with lymphoid neoplasms, 13/214 with myeloid neoplasms, and 4/35 with other diseases. Active treatment per se was not correlated with negative response. However, rituximab and BTK inhibitor treatment were correlated significantly with a negative vaccination response, whereas younger age and chronic myeloid leukemia (CML) disease were associated with positive response. In addition, 5 of 6 patients with myeloproliferative neoplasm (MPN) and negative vaccination response were on active treatment with ruxolitinib. In conclusion, a remarkable percentage of patients with hematological diseases had no response after 2nd COVID-19 vaccination. Multivariable analysis revealed important factors associated with response to vaccination. The results may serve as a guide for better protection and surveillance in this vulnerable patient cohort.

In patients with hematological diseases, a compromised immune system may affect preventive vaccination as demonstrated by reduced efficacy not only for COVID- 19 as several studies have shown [9][10][11]. Mortality rate of COVID-19 after vaccination in this patient cohort is still high with around 12% [12]. Patients with non-Hodgkin lymphoma (NHL), higher age, and immune-compromising B cell-depleting therapy are at risk for reduced vaccination response against COVID-19 [7,9,13,14].
In addition, studies have shown a correlated time-dependent reduction in antibody titers over the time. Among persons of 60 years or older who became fully vaccinated in January 2021, the rate of breakthrough infections was 1.6 times higher compared to those who became fully vaccinated in March 2021 [15][16][17]. So far, little is known about this time-dependent reduction of antibody titers in patients with hematological disease as well as correlation to therapy and outcome after COVID-19 [8].
The aim of our study was to evaluate vaccination-related antibody response to BNT162b2 (BioNTech), mRNA-1273 (Moderna), and ChADOx1 (AstraZeneca) in patients with various hematological disorders and to identify prognostic factors influencing vaccination response.

Subjects
In this observational single-center study, data were collected from July to October 2021. Patients with hematological diseases and a scheduled appointment were registered regarding their vaccination status. Patients who received the 2nd COVID-19 vaccination at least 2 weeks prior the appointment were included. As a control group, patients with benign and autoimmune diseases were also registered.
As part of routine patient care, 7.5-mL serum and heparin samples were collected. After adequate clotting of serum samples during 1 h at room temperature, samples were centrifuged at 2000 g for 10 min at 18 °C. Plasma was similarly centrifuged/separated and used for the subsequent analyses. A Food and Drug Administration/Continuing Education (FDA/CE)-approved electrochemiluminescent assay (ECLIA) (Elecsys®, Roche, Mannheim, Germany) was used to quantify serum antibodies, pan Ig (including IgG) against the receptor binding domain (RBD) of the SARS-CoV-2 spike protein. The assay has a measurement range of 0.4 to 250 U/mL, with a concentration ≥ 0.8 U/mL considered positive. Data were analyzed for patients without detection of anti-N (nucleocapsid) SARS-CoV-2 antibody. As it is unclear which level of antibody represents a sufficient protection, we here focus on the analysis of patients with negative antibody response vs. patients with any response [18,19]. All tests were performed according to the manufacturer's instructions in an accredited laboratory at the University Hospital Mannheim.

Statistical analyses
Baseline covariates were compared between vaccination responders and non-responders using either the Mann-Whitney U test, chi 2 test, or Fisher's exact test, as appropriate. Each of these tests has been carried out as a two-sided test. Univariable logistic regression analyses have been performed in order to calculate odds ratios for several factors. Furthermore, a multiple logistic regression analysis for the binary outcome "responder" has been performed in order to analyze several variables simultaneously. In general, the significance level was set to 0.05. All statistical calculations were performed with SAS (release 9.4, SAS Institute Inc., Cary, NC).
Vaccination-related antibody response was positive in 317/373 patients (85%) with a median level of 197 U/ mL (range 0.8-250 U/mL) and negative in 56/373 patients (15%). Of the patients with positive seroconversion, 63/317 patients (20%) had an antibody level between 0.8 and 100 U/mL. Mean time from vaccination to measurement was not different in both cohorts. The average analysis was after 9 weeks, with a minimum of 2 weeks (both groups) and a maximum of 28 weeks (group with negative seroconversion) and 32 weeks (group with positive seroconversion).
In selected patients, antibody titers were measured twice at 2 and 4 months after the 2nd COVID vaccination: A decrease from 250 to 171 U/mL was noted in one patient and from 66 to 5 U/mL in another. Thus, antibody titers in both patients decreased significantly, even though they are considered positive.

Impact of gender and age
A total of 38/56 patients (68%) with negative antibody titer were male, and 18/56 patients (32%) were female. In patients with positive response, 171/317 (54%) were male. This difference slightly failed to reach statistical significance (p = 0.0531).
Older age was associated with a negative antibody titer. Medians for birth year were 1946 and 1958 for patients with negative or positive response, respectively (p < 0.0001). In particular, negative titers were found in elderly patients with aggressive and indolent NHL (Tables 2, 3, and 4).
Independent of the treatment status, lymphoid disorders (indolent and aggressive) were predominant in this subgroup. In patients with indolent NHL, negative antibody response was present on different treatment regimens: 7/8 patients were on BTK inhibitor (ibrutinib or acalabrutinib), and 5/8 patients were on rituximabbased chemotherapy. Moreover, 8/32 patients with indolent NHL without any therapy the last 3 months before 1st vaccination had no seroconversion. In the group of aggressive NHL, 5/8 patients with no antibody response were on active therapy.
A total of 28/373 patients were treated with ruxolitinib and most of them (n = 24) were diagnosed with MPN. Four/24 patients had a negative antibody response (age: 69-88 years), whereas 20/24 patients had a measurable positive seroconversion (age: 42-86 years). Nevertheless, antibody titers below 100 U/mL were detected in 9/20 patients with a positive immune response. In MDS patients, no correlation was found between specific therapy or age and negative vaccination response.

Univariable and multivariable analyses
In univariable analyses, the following factors were significantly associated with negative vaccination response: older age (p < 0.0001), indolent NHL (p < 0.0001), aggressive NHL (p = 0.0021), and mRNA-1273 (Moderna) vaccine (p = 0.0241; s. Figure 1). Patients with CML were significantly less likely to have negative response. These patients were on TKI treatment or in treatment-free remission [20]. MPN diagnosis slightly failed to reach statistical significance (p < 0.0001 and p = 0.0515), respectively.
Odds ratios and confidence intervals for some relevant factors are presented in Fig. 2. Therapies associated significantly with no antibody response were rituximab (OR = 14.7), BTK inhibitors (OR = 44.3), and immunoglobulin substitution (OR = 4.7).
In the multivariable logistic regression analysis using the forward-selection method, diagnosis of indolent (p < 0.0001, odds ratio (OR) = 1.6) and aggressive NHL (p = 0.0004, OR = 2.0) was found to increase the risk for negative vaccination response. On the other hand, protective impact was found for CML (p = 0.0312, OR = 0.105; s. Figure 2).
Regarding the year of birth, the risk for a negative vaccination response decreased with each year (p = 0.0008, OR = 0.956) indicating that older people (with a lower year of birth) have a higher risk.
Taking out the CML patient group as very good responders, factors influencing vaccination status did not change.
Time after vaccination, gender and being on active treatment in general had no significant impact on the seroconversion rate after vaccination.

Discussion
Our data of this single-center study comprise one of the biggest well-characterized patient group and correspond to a real-world cohort in a hematology ambulatory setting which enabled us to identify important factors influencing COVID-19 vaccination success in patients with hematological disorders. Fig. 1 Forest plots of univariable logistic regression models with odds ratio (OR) and confidence intervals demonstrating risk for negative vaccination response correlating with a the vaccine used and b treatments. Legend: CI, confidence interval; Chemoth, chemotherapy; TKI, tyrosine kinase inhibitors; Lenal., lenalidomide; JAK2, Jak2-inhibitors; Ig, immunglobulin substitution; Hypom., hypomethylating agents; HU, hydroxyurea; BTK, BTK inhibitors Fig. 2 Forest plots of the multivariable logistic regression model with odds ratio (OR) and confidence intervals demonstrating risk for negative vaccination response. Legend: CI, confidence interval; NHL_low, indolent non-Hodgkin lymphoma; NHL_high, aggressive non-Hodgkin lymphoma; CML, chronic myeloid leukemia The central finding of our study is that 99% of patients with CML had a positive seroconversion after the 2nd COVID vaccination. This confirms the results from a recently published smaller study [21].
Furthermore, in our cohort active treatment per se did not significantly correlate with negative vaccination response as reported in other studies summarized in a recent metaanalysis [7]. A possible explanation might be that our patient cohort included a substantial number of patients with negative seroconversion which have been treated earlier or were treatment naïve. Patients with different disorders including mainly indolent NHL belonged to this group indicating that not only patients on active treatment should be screened for antibody response.
In accordance to the abovementioned meta-analysis, patients diagnosed with lymphoid neoplasms (especially indolent and aggressive NHL) were at higher risk for negative seroconversion compared to patients with other hematological diseases [7]. However, any patients with other hematological disorders, especially MPN and MDS patients, had a high risk for negative vaccination result.
As described in the healthy population, older age was correlated with a negative antibody response to COVID-19 vaccination [22]. Age was related to negative antibody titers of all entities in a multivariable analysis. Regarding the year of birth, the risk for a negative vaccination response increased with each year.
In the univariable analysis, vaccination with mRNA-1273 (Moderna) correlated with negative antibody response. In contrast, in the multivariable analysis with addition of further influencing factors, no significance was discernible. As the number of patients vaccinated with Moderna was small and included predominantly NHL patients, no conclusion can be drawn out of this result.
As found in other studies, target-specific therapies were correlated with negative antibody response [9,10,23]. Therapies significantly associated with a negative seroconversion were BTK inhibitors like ibrutinib, acalabrutinib, rituximab, and immunoglobulin substitution. TKI, interferon, lenalidomide, hypomethylating agents, and other treatments did not demonstrate significance in our analysis. Patients with ivIG therapy represent a heterogenous disease group (patients with B-CLL, indolent NHL, multiple myeloma, etc.). Despite the fact that ivIG therapy was associated with reduced vaccination response, the therapy per se is unlikely the determining factor. Rather, in these patients, the underlying disease may have an impact on seroconversion.
However, most of the patients with MPN and no seroconversion received ruxolitinib. In patients with MDS, there was no correlation between specific therapy and negative response.
Allogeneic stem cell transplantation was not a negative predictor in our cohort as most patients were in longer follow-up after the procedure, which correlates with one recent published study [24]. The patients with negative response received treatment after complications.
Latest studies have shown a stronger decrease of vaccination response over the time associated with higher age [13,16,17] and antibody titers decrease every month [16,17]. Our preliminary results suggest that vaccine titers may decrease much more rapidly in patients with hematologic diseases compared to the general population. To clarify this, more systemic prospective data on a possible higher decline over time especially in hematologic patients are needed.
In this context, another important aspect is the influence of antibody titers lower than 250 U/mL or even below 100U/ mL. In our cohort, a substantial amount (20%) of patients were measured with titers below 100U/mL. Titers above 0.8 U/mL are considered positive but it is still unclear what level is representative for a sufficient protection. Thus, a classification of the effectiveness is important in order to be able to counteract with specific measures at an early stage, e.g., prioritization and earlier booster vaccination as already recommended for patients with immunodeficiency [9,23]. Other vaccination strategies [25] or treatments with monoclonal antibodies pre-/post expositional might be essential for these patients [8,26].
Neutralizing antibodies can prevent the host cell infection by binding to the spike protein. Although our study did not examine the cellular immune response, there is evidence that measurement of S-protein antibody levels correlates with neutralizing antibody titers as it was demonstrated in several studies [13,14,19,[27][28][29][30].
Since new variants of SARS-CoV-2 have emerged, the question is whether neutralizing antibodies can also capture viral variants and whether this correlates with antibody titers after vaccination [15]. It may be due to mutations in the RBD, as is the case with Omicron, that protection by neutralizing antibodies is not sufficient. Therefore, it is equally important to pay attention to the T cell response after vaccination in this context [31].
It seems like T cells show protection against severe COVID-19 and it appears that patients develop a cellular immune response after vaccination despite little or no seroconversion [32].
Vaccinated individuals do have a T cell immunity to the SARS-CoV-2 Omicron variant, potentially balancing the lack of neutralizing antibodies in preventing or limiting severe COVID-19 [31]. Booster vaccinations could be important to further restore cross-neutralization by antibodies [32]. However, patients with measurable antibody titers may still not be protected against infection because their antibodies may not be neutralizing as described above.
In summary, our data defined important predictive factors for negative antibody response after COVID-19 vaccination in patients with hematological diseases. Negative seroconversion was significantly correlated with lymphoid diseases, older age, and therapy with BTK inhibitors, rituximab, and immunoglobulin substitution. In addition, ruxolitinib therapy was an important negative predictor in MPN patients. Active antitumoral therapy per se was not a negative predictive factor. Vaccination of CML patients was significantly associated with positive seroconversion.
In conclusion, these data help to identify patients at risk for negative seroconversion after COVID-19 vaccination and to adapt further clinical decisions to prevent these patients more effectively from COVID-19.