Response to three doses of the Pfizer/BioNTech BNT162b2 COVID-19 vaccine: a retrospective study of a cohort of haemodialysis patients in France

The emergence of a novel form of infectious respiratory disease in late 2019 has since led to a global pandemic of coronavirus disease 2019 (COVID-19) and millions of death worldwide. In the absence of effective anti-infective medications and the relative inability of containment measures to stop the spread of the pathogenic severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) on the population level, public health strategies have relied on mass (voluntary) vaccination. Fortunately, the vaccines developed, tested clinically and approved in 2020 have given very encouraging results in terms of preventing clinical cases (and especially severe cases) of COVID-19 among vaccinated individuals [1].

Most countries have adopted a priority-based system to mass vaccination [2]. Given the high COVID-19 mortality rates observed in patients with chronic kidney disease (CKD) in general and those receiving renal replacement therapy in particular [3, 4], the nephrology community has emphasized the need to prioritize vaccination for patients [5, 6]. It must be borne in mind that the pivotal clinical trials required for marketing authorization of the forerunner vaccines did not include patients with physician-diagnosed CKD [7]. In France, patients on haemodialysis became eligible for vaccination with approved mRNA vaccines (BNT162b2 (Comirnaty) from Pfizer/BioNtech, or mRNA-1273 from Moderna) on January 18th, 2021 (https://​solidarites-sante.​gouv.​fr/​IMG/​pdf/​dgs_​urgent_​04_​vaccination_​patients_​a_​risque.​pdf). However, the initially published results on the immune response to COVID-19 vaccination in haemodialysis patients were inconsistent: some studies evidenced a weak (subnormal) response, whereas others evidenced an essentially normal response, with high titres of IgG against SARS-CoV-2’s spike (S1) protein [8‐14]. These disparate findings (for reviews, see [15, 16]) prompted us to investigate the vaccination rate and the vaccine response in a cohort of haemodialysis patients attending our dialysis centre in Antibes, France.

The primary objective of the present retrospective, real-life study was to describe the serological status of haemodialysis patients after vaccination with BNT162b2, with a focus on three doses vs. two doses. The secondary objective was to identify factors associated with a poor response to COVID-19 vaccination, in order to refine the advice given to our patients.

The main findings of the present retrospective study of haemodialysis patients attending a single dialysis centre in France were as follows: (i) two doses of BNT162b2 were not enough for a robust humoral immune response in all our patients on dialysis, (ii) the anti-S1 IgG titres increased significantly after a third dose of vaccine (as would be expected for a “booster” injection), (iii) the mean BMI and the mean serum albumin levels were slightly but significantly higher in responders than in non-responders with two doses, (iv) three factors (a history of COVID-19, the receipt of at least two doses of vaccine, and being on the French national waiting list for kidney transplantation) were found to be independently and significantly associated with the anti-S1 IgG titre at the end of the study, (v) the anti-S1 IgG titre at the end of the study was not correlated with age or immunosuppressant treatment, and (vi) the anti-S1 IgG titre was correlated with the anti-HBV titre after three doses of vaccine but not after two doses.

Our results for a group of 100 dialysis patients having received one, two or three doses of BNT162b2 are in line with a number of literature reports – some of which appeared after our analysis had been performed. Firstly, with regard to the difference between one dose and two doses, Attias et al. found that the seropositivity rate was 18% before the second injection and 82% afterwards [8]. Grupper et al. reported that patients (n = 56) on maintenance haemodialysis developed a substantial humoral response following vaccination with BNT162b2 but that the anti-S1 IgG titres of against SARS-CoV-2’s spike (S1) protein (measured a median of 30 days after the second dose) were significantly lower than in a group of (younger) healthy controls [11]. However, a report by Torregiani et al. suggested that only about one-third of patients on haemodialysis developed neutralizing antibodies after the first dose of BNT162b2 [13]. Simon et al. observed very low median anti-S1 titres in dialysis patients (n = 81) after two doses of BNT162b2 (171 U/ml, versus 2500 U/ml in healthy controls) [12]. Billany et al. reported that SARS-CoV-2 neutralizing antibodies against the receptor binding domain of the spike protein were not detectable in 19 out of 94 patients receiving maintenance haemodialysis and who had received the BNT162b2 or the AZD1222 (Oxford-AstraZeneca) vaccine (20.2%) [9]. In Israel, Yanay et al. reported a lower response rate to the vaccine, a lower anti–S1 antibody titre, and a higher rate of COVID-19 infection after vaccination in a group of dialysis patients, relative to controls [14]. In a study of 90 vaccinated HD patients (mean age: 69) of whom 19 (21%) had a history of SARS-CoV-2 infection, Giot et al. documented anti-S IgG seroconversion in 20% of patients after the first dose and in 77% after the second dose [22]. In Nacasch et al.’s study, 19% of double-vaccinated haemodialysis had low or undetectable antibody levels [23].

With regard to the difference between two doses and three doses, Ducloux et al.’s study of a group of 50 dialysis patients observed a vaccine response rate of 90% among those having received two doses of BNT162b2 [10]. The administration of a third dose enhanced the antibody titre markedly in almost all patients and especially in those with low titres after two doses. Marked increases in the anti-S1 antibody titre were also reported by Francken et al. after the administration of a third dose of BNT162b2 to a subset of patients with a value below 250 U/ml (according to the Roche Diagnostic Elecsys® enzyme immunoassay) after two doses [24]. In a follow-up report on the patients studied by Giot et al., Robert et al. notably described the response to a third dose in 10 “partial responders” (defined as positive for circulating anti-S1 IgGs but negative for neutralizing antibodies after two doses) [25]. After a third dose, 7 of the 10 partial responders still had anti-S1 IgGs and four had developed neutralizing antibodies. The median [IQR] anti-S1 IgG titre was 31.5 [17.8–41.8] BAU/mL after two doses and 776.7 [138.3–3038] after three. Like Robert et al., Frantzen et al. chose to administer a third dose of BNT162b2 only when the individual titre was below a particular threshold [24]. Bensouna et al. found that a third dose of BNT162b2 increased antibody levels substantially in patients on maintenance dialysis and appeared to be as well tolerated as the second dose [26]. Lastly, Stumpf et al. reported on the humoral and cellular immune responses to boost vaccination with mRNA vaccines in a cohort of dialysis patients, kidney transplant patients and medical personnel; after the boost, the seroconversion efficacy among dialysis patients (> 95%) was similar to that seen among medical personnel [27].

In contrast to studies in which only non-responders or partial responders were given a third dose, we recommended a third dose to all our patients in the spring of 2021. This was not long after the start of the vaccination campaign in France, at a time when a third dose was not being considered for general population. At the time of writing, a three-dose COVID-19 vaccination regimen is recommended by the French health authorities for all individuals over the age of 12, regardless of whether or not risk factors are present. Hence, our medical decision became a legal requirement; in France, members of the general population aged over 12 must receive a third dose if they want a valid “vaccine passport”.

The design of the present retrospective, observational study prevented us from determining the causal nature of relationships between the anti-S1 IgG titre on one hand and demographic, clinical and laboratory factors on the other. The only small (but statistically significant) differences between non-responders and responders after two doses of vaccine concerned the BMI and serum albumin level. However, the BMI and serum albumin values observed here were not unusual for a haemodialyzed population. Furthermore, two of the factors associated with anti-S1 IgG titre (a history of COVID-19 and being on the French national waiting list for kidney transplantation) are of little practical value with a view to increasing the response rate among vaccinated patients on haemodialysis. Although French and European guidelines are available, placement on the French national waiting list depends on each clinical team’s practices; these differ from one region of France to another and even from one centre to another in the same region [28].

In our patients with at least two doses of vaccine, age was not significantly associated with the anti-S1 IgG titre (as also observed by Irsara et al. [19]). In most literature reports, however, age is a major confounding factor in the reports on the response to COVID-19 vaccines. In the reports by Nacasch et al., Simon et al. and Grupper et al., greater age was associated with a lower anti-S1 IgG titre [11, 12, 23]. Jahn et al. found that almost all dialysis patients under the age of 60 showed an essentially normal response (compared with controls) to a second (but not first) dose of BNT162b2 (median [IQR] titre: 597.0 arbitrary units (AU)/mL [410.5; 800.0]; in contrast, dialysis patients over the age of 60 had significantly lower antibody titres (median [IQR] titre: 280.0 AU/mL (45.7; 477.0); p < 0.0001) [29]. However, the fact that our study population was quite elderly (median [interquartile range] age: 74 [64–81]) might explain the lack of a significant correlation. The moderate proportion of (often young) patients on the transplant list (33.8%) and/or the presence of contraindications to transplantation (e.g. immunosuppression and cancer) might also contribute individually or collectively to the lack of a significant association.

It is noteworthy that in our multivariate analysis, treatment with immunosuppressants was associated with lower anti-S1 IgG titre at Ser₁ but not at Ser₂. We note that according to Benotmane et al. 2021, kidney transplant recipients (rather than patients on dialysis, as in our study) treated with calcineurin inhibitors, mycophenolate mofetil, or steroids showed significantly lower anti-SARS-CoV-2 antibody titres after one and two doses of the Moderna mRNA-1273 vaccine [30]. In a study of dialysis patients and kidney transplant recipients, Stumpf et al. reported that the number of immunosuppressive drugs and the type (belatacept, mycophenolate mofetil-mycophenolic acid, and calcineurin inhibitors) were risk factors for seroconversion failure [27]. In a study of haemodialysis patients by Nacasch et al., long-term immunosuppressive therapy was the primary predictor of low antibody titres in double-vaccinated individuals (odds ratio: 30.4; p < 0.001) [23]. More generally, it is noteworthy that COVID-19 vaccination is not contraindicated in France for the major immunosuppressants (corticosteroids, methotrexate, azathioprine, mycophenolate mofetil, hydroxychloroquine, leflunomide, sulfasalazine, anti-tumour necrosis factor agents, anti-CD20 antibodies, and Janus kinase inhibitors); the benefits of vaccination – even when dampened by immunosuppressants – far outweigh the risks of non-vaccination.

According to the manufacturer of the in vitro assay used in the present study, an anti-S1 antibody titre of 7 U/ml is “neutralizing” for SARS-CoV-2 [17]. Anand et al. suggested a slightly higher value of 10 U/ml [31]. Among the 112 COVID-19-negative patients having received two doses of the BNT162b2 vaccine, the median anti-S1 antibody titre was 7.09 U/ml; this means that at least half of these patients had a titre below or barely above the supposedly neutralizing value and so were probably not sufficiently protected. In contrast, the median titre rose markedly after the third dose (to 93.26 U/ml), which probably corresponds to a good level of protection. More broadly, our results highlight the urgent need for reliable correlates of protection (CoP) in patient groups that have not been well represented in clinical trials of COVID-19 vaccines. A CoP is a measure of the immune response that is significantly correlated with protection against infection, disease and/or transmission in vaccinated individuals. The use of CoP may allow the prediction of clinical outcomes more rapidly than in clinical trials – especially for rapidly emerging new variants of SARS-CoV-2. However, CoP are difficult to define and may depend on the characteristics of each type of vaccine, including the antigen, the vector, the antigen presentation method, the presence or absence of adjuvants, and the vaccination regimen (number of doses) – all of which might affect the humoral response, the cellular response, or both. CoPs have not been clearly defined for SARS-CoV-2 vaccines in general, let alone in groups of vulnerable individuals like dialysis patients.

It will be interesting to (i) follow up changes over time in our patients’ IgG titres in particular and other markers of the humoral response in general, and (ii) compare these data with the literature data on the general population. Interestingly, there is evidence to suggest that dialysis patients mount a cellular immune response (albeit imperfectly) to SARS-CoV-2 [27, 32], which might help to compensate for a weaker humoral response. Lastly, novel variants of SARS-CoV-2 emerge frequently, and it will be essential to determine whether certain variants represent a particular risk for dialysis patients.

Given that we only administered the Pfizer/BioNTech BNT162b2 vaccine, we can only speculate about other vaccines on the basis of the literature data. At the time of our initial submission, there were few literature data on vaccines other than Pfizer/BioNTech, and most of the pivotal vaccine trials exclude participants on dialysis. Since then, a few studies have assessed the Moderna mRNA vaccine (mRNA-1273) in patients on dialysis. In a study of peritoneal dialysis patients having received the mRNA-1273 vaccine, Rodriguez-Espinosa et al. found that the mean ± SD anti-S1 IgG titre rose from 28.09 ± 52.2 after the first dose to 113.7 ± 56.9 after the second dose (a third dose was not described) [33]. Interestingly, Kaiser et al. performed a comparative study in haemodialysis patients: those vaccinated with mRNA-1273 showed higher anti-S titres than those vaccinated with BNT162b2 [34]. In their prospective, multicentre observational study of people vaccinated with mRNA-1273 or BNT162b2, Stumpf et al. found that the seroconversion rate for the anti-S titre was significantly higher in dialysis patients vaccinated with mRNA-1273 than (95%) than in those vaccinated with BNT162b2 (85%, p < 0.001) [27]. However, Hasmann et al. pointed out that according to Stumpf et al.’s data, the response rates were still significantly lower in dialysis patients than in controls [35].

The present study had several strengths. Firstly, it described the response to three doses of BNT162b2 in a well-documented, relatively large (n = 100) group of patients on dialysis. Secondly, it provided comparative data on patients having received one, two and three doses of the vaccine. Thirdly, we identified factors associated with a lack of response to one and two doses of BNT162b2. The study also had a number of limitations, most of which were inherently associated with its retrospective design. Firstly, we did not include a control sample from the general population or from non-dialysis patients with CKD. Secondly, the patients were not all vaccinated with the same number of doses and at the same time points relative to the serology assays, although this heterogeneity reflected care pathways and vaccination choices in a “real-life” population during an COVID-19 epidemic. However, the time interval between the third injection of vaccine and the most important serology assay (Ser₂) was fixed (35 or 36 days). Thirdly, the study was performed in a single centre, and our findings may not necessarily extend to other geographical areas and other healthcare systems. Fourthly, the study participants did not undergo a serology assay immediately prior to the start of the study (e.g. in January 2021). However, all but one of the participants were seronegative for anti-S1 IgG in May 2020, and we considered the sole seropositive (but asymptomatic) patient to be a false positive. Fifthly, we did not perform a neutralization assay, which might have provided more information about the patients’ likely degree of protection from infection. However, the neutralization assay is technically complex and is not available routinely in medical biology laboratories.

In conclusion, our present data suggest that dialysis patients vaccinated with two doses of BNT162b2 might not have a sufficient level of protection against SARS-CoV-2 and should receive a third dose (at least) as part of a personalized vaccination strategy. We consider that due to (i) the emergence of new, virulent SARS-CoV-2 variants and (ii) the frequently reported post-fall in anti- SARS-CoV-2 antibodies some months after vaccination, the need for long-term, regular boosts with new or modified COVID-19 vaccines is likely for this patient population. Indeed, we have now started to recommend a fourth dose to our dialysis patients.