Abstract
Background The diffusion of the SARS-CoV-2 delta (B.1.617.2) variant and the waning of immune response after primary Covid-19 vaccination favoured the breakthrough SARS-CoV-2 infections in vaccinated subjects. To assess the impact of vaccination, we determined the severity of infection in hospitalised patients according to vaccine status.
Methods We retrospectively analysed data from patients hospitalised in 10 centres with a SARS-CoV-2 infection (delta variant) from July to November 2021: i) all patients who had completed their primary vaccination at least 14 days before hospital admission; and ii) the same number of completely unvaccinated patients. We assessed the impact of vaccination and other risk factors through logistic regression.
Findings We included 955 patients (474 vaccinated and 481 unvaccinated). Vaccinated patients were significantly older, more frequently males, and with more comorbidities. They were less often admitted for Covid-19 (59·3% vs. 75·1%, p<0·001), showed fewer lung lesions, and required oxygen less frequently (57·5% vs. 73·0%, p<0·001), at a lower flow (3·0 vs. 6·0 L/min, p<0·001), and for a shorter duration (3 vs. 6 days, p<0·001). They less frequently required intensive care unit admission (16·2 % vs. 36·0 %, p<0·001). Mortality at day 28 was not different between the two groups (16·7% vs. 12.2%, p=0·075), but multivariate logistic regression showed that vaccination significantly decreased the risk of negative outcomes, including mortality, even when considering older patients, and those with comorbidities.
Conclusions Among patients hospitalised with a delta variant SARS-CoV-2 infection, vaccination was associated with less severe forms, even in the presence of comorbidities.
Introduction
Despite reaching relatively high Covid-19 vaccine coverage, during the second half of 2021 many countries faced a new wave of SARS-CoV-2 infections. It may be related to : i) a variable proportion of the population not yet immunised, ii) anti-Spike antibody levels decrease over time after two doses of either RNA or ChAdOx vaccines1, and iii) vaccination being less efficient against the delta (B.1.617.2) variant of concern (VOC), which effectively became the single circulating form of SARS-CoV-2 during early summer 2021, compared to the original strain 2. Several studies have characterised the rate of such breakthrough infections in vaccinated persons 3, and reported that this risk increased over time after primary vaccination 4–6. It has also been shown that even with the delta variant, in the general population, the rate of Covid-19-related hospitalisations 7,8, intensive care unit (ICU) admissions, and mortality 9 were much lower in those who received two doses of the aforementioned vaccines.
Nevertheless, no precise data are currently available regarding the characteristics of vaccinated patients hospitalised with a SARS-CoV-2 infection. We therefore aimed to determine whether vaccination, among other factors, influenced the outcome of hospitalised patients with a SARS-CoV-2 infection, in particular in terms of need for oxygen, ICU admission, and death within 28 days after hospitalisation.
Methods
Population
We conducted a retrospective, multicentre study across 10 hospitals in France (Ajaccio, Annecy, Caen, Clamart, Grenoble, Le Mans, Limoges, Nancy, Nantes and Poitiers). Hospitalised patients with a PCR-proven SARS-CoV-2 infection from July to November 2021, i.e., when the Delta variant was virtually the only circulating SARS-CoV-2 variant in France, either at the time of admission or after admission, were selected. Among them, two populations were identified and included in the study:
- all patients who had received at least either one injection of the Janssen vaccine or two injections of the AstraZeneca and/or RNA vaccines (with the second dose given at least 14 days before date of PCR);
- the equivalent number of patients without any vaccination before hospital admission. This last group was composed by including for each study month and in each hospital as many unvaccinated patients as vaccinated patients, by chronological order of hospitalisation.
All the centres screened their patients to include only the delta variant. For one centre, the inclusion was prolonged up to the December 13th, but screening guaranteed that only the delta variant was involved. The following data were collected: year of birth, gender, and pre-existing comorbidities (respiratory disease, cardiac failure, arterial hypertension, solid tumour, haematological neoplasia, and kidney failure, which was determined by the Cockcroft formula based on data collected at hospital admission); Covid-19 vaccination history; clinical data at admission and minimum Ct PCR level; maximum C-reactive protein plasmatic level; extension of lung disease on computed tomography (CT-scan (if performed); need for oxygen with duration and maximal flow; need for high-flow oxygen therapy; need for ICU admission; need for invasive mechanical ventilation; prescription of steroid, tocilizumab, and/or monoclonal antibodies; and mortality at day 28.
According to the French law on ethics, patients were informed during their hospital stay that their anonymised medical data could be used for research purposes; they were given the possibility to refuse this usage. As requested by French ethics and regulatory laws, the ethical committee of the French-speaking Society of Infectious Diseases (SPILF) (IRB00011642) gave its approval for the study (N°2022-0101), and the study was declared to the French National Commission for Informatics and Liberties (CNIL MR004: n°2224742).
Statistical analysis
Firstly, a descriptive analysis was performed in the cohort population according to vaccination status. Qualitative variables were described as counts (percentage) and frequency distributions were compared with the Chi square test or Fisher’s exact test when appropriate. Continuous variables were expressed as median (1rst quartile; 3rd quartile) and differences were tested with the independent t-test for normally distributed variables or otherwise the Mann-Whitney U test. Secondly, univariate and multivariate logistic regressions based on general linear models were performed with a stepwise variable selection according to the Akaike Information Criterion10. Factors associated with severe forms of infection defined by three outcomes, namely requirement for oxygen, ICU admission and death at day 28, were assessed, focusing on the impact of vaccination. To assess the impact of comorbidities and vaccination, we selected all the potential risk factors for severe outcome presented in Table 2 for the multivariate model (Table 3 and supplementary Table 1). In addition, to manage Missing data Not At Random (MNAR), multivariate analysis was initially performed only on the dataset with complete cases, after which the model was applied to the full dataset 11. Results are presented based on the full dataset, as they were similar to the analysis restricted to the complete case dataset. As an example, missing CT-scan results could equally be due to the fact that patients were too severe to benefit from a CT scan before high flow nasal oxygen or mechanical ventilation, or to the fact that the patient was mildly ill and went back home without any oxygen requirement; in some instances, the result was missing from the patient files. Finally, a multivariate analysis was similarly carried out on the dataset of vaccinated patients only, so as to independently assess the weight of comorbidities among vaccinated patients, taking into account the influence of the time elapsed since the most recent injection. Interactions were systematically searched and mentioned. All statistical analyses were performed with R version 4.1.2 12.
Results
Nine hundred seventy-four patients were initially included; 19 were excluded for i) being under 18 years (n=4), ii) not being fully vaccinated (n=12) or iii) being infected with a VOC other than delta or hospitalised when delta variant was not the only circulating variant (n=3). The characteristics of the 955 patients finally included, among whom 474 were vaccinated and 481 were not, are detailed in Table 1.
The two groups differed on several points: unvaccinated patients were significantly younger and had fewer comorbidities (reflecting the sequential prioritisation of the vaccine campaign during the early months of 2021), although they also had a higher body mass index. Among vaccinated patients, the median delay since the most recent dose was 125 days. The BNT162b2 (Comirnaty, Pfizer) vaccine had been used in 357 (76·1%) patients. Astra-Zeneca, Janssen (Johnson & Johnson), Moderna and Sputnik had been used in 62 (13·2%), 27 (5·8%), 22 (4·7%) and 1 (0·2%) patients respectively.
All in all, vaccinated patients had milder forms of SARS-CoV-2 infection (Table 1). They were less frequently hospitalised because of the infection itself; they had fewer extended lung lesions on CT-scan; they required oxygen therapy less frequently, with a lower oxygen flow, and for a shorter duration; they received steroid therapy less often; they were less frequently admitted to ICU; and they less often needed high-flow oxygen or mechanical ventilation. Vaccinated patients’ 28-day survival was not statistically different, even in patients presenting more comorbidities; in addition, when considering a negative outcome index combining death at day 28 and ICU requirement, there was a significant relative 32% risk reduction among vaccinated patients compared to unvaccinated patients. The protective effect of vaccination was also observed when considering only older patients (>65 years), as illustrated in Figure 1.
In univariate analysis, vaccination and a previous SARS-CoV-2 infection were protective toward ICU and oxygen requirement, but not significantly associated with 28-day mortality. Comorbidities, older age, male gender and a longer time lapse since the most recent vaccine injection (among vaccinated patients) were significantly associated with a negative outcome (Table 2). In addition, being hospitalised for COVID, an elevated CRP, a low Cycle threshold and extensive lung lesions on CT scanner were associated with a negative outcome (Supplementary Table 2).
Multivariate analysis confirmed the protective effect of vaccination regarding the need for oxygen, the need for ICU admission, and the risk of death. Vaccination was even able to offset the impact of many risk factors such as hypertension and overweight (Table 3 and Supplementary Table 1). Notably, in addition to factors with odds ratio not statistically different from 1 presented in Table 3, many risk factors were finally not kept in the final multivariate model due to their negligible impact. Moreover, in the multivariate regression model for death at day 28, the absence of significant interactions also showed that the protective effect of vaccination occurs for all patients, including those most at risk for severe forms such as older or immunosuppressed patients. In the multivariate regression model for oxygen requirement, an interaction between vaccination and a previous infection by COVID-19 was observed, suggesting that a previous COVID-19 infection decreased the influence of the vaccination; the large coefficient associated with having previous COVID-19 infection suggests that it takes away most of the protective effect on oxygen requirement, leaving a small but still significant influence to vaccination.
Discussion
The emergence of SARS-CoV-2 delta VOC jeopardised resolution of the Covid-19 pandemic. Indeed, the humoral immune response triggered by natural infection and/or vaccination is less likely to neutralise the VOC, which acquired key mutations in gene coding for the Spike protein. We therefore aimed to characterise the protection associated with vaccination towards the consequences of Delta VOC infection by studying the population of patients hospitalised with this infection. Including both patients hospitalised because of Covid-19 and those with incidental positive PCR while being admitted for another purpose allowed us to compare vaccinated and non-vaccinated patients not only with Covid-19, but all infected patients, including those who were asymptomatic. Indeed, vaccine-induced protection can manifest by the absence of symptoms in case of infection 13.
The emergence of the Delta variant was anticipated as being associated with breakthrough infections. Indeed, in vitro experiments showed that the neutralisation index of plasma from vaccinated individuals was lower for delta than for initial viral strain 14–16. Different studies and a meta-analysis assessing the impact of primary vaccination observed a ≃ 10% reduction in vaccine efficacy against asymptomatic and symptomatic infections with delta VOC, but no reduction of vaccine efficacy against severe forms 2,17. Some authors have noted that the higher incidence of infections with this VOC may also be due to the fact that when delta became predominant, those infected had received their primary vaccination 4 to 10 months before, meaning that the infection may have been related to the waning of the level of vaccine-induced antibodies 18.
Recent studies showed 86-88% vaccine-associated reduction in the hospital admission rate in those infected with the delta VOC 19–21; Our observations confirm the positive impact of vaccination on the severity of the SARS-CoV-2 delta infection, even in hospitalised patients. Moreover, we observed that the positive impact was maintained in populations at risk of severe Covid-19: those aged over 65 years, and those with pre-existing cardiac failure, kidney failure, or a chronic respiratory disease (at least for one or several severity criteria). All in all, this legitimates the prioritisation of these populations during the early months of the 2021 vaccine campaign: they benefited from valuable protection, even with the delta variant. Taken with previous fatality studies of vaccinated patients 21–23, this also confirms the perception shared by many physicians that vaccinated fragile people survived a form of Covid-19 in 2021 that they would not have survived without vaccination. Taking into account that overweight was more frequent among unvaccinated patients, while vaccination nonetheless effectively protected the associated mortality, emphasis should be made on this risk factor for severe forms of Covid-19.
The association between vaccination and less severe SARS-CoV-2 infection was less clearly evidenced in immunocompromised patients. The absence of interaction between vaccination and immunosuppression and the absence of a significant effect of immunosuppression in the model applied specifically to vaccinated patients suggest that vaccination is also efficient in immunocompromised patients (suppl. Table 1). Nonetheless, these results should not be over-interpreted as most of the immunocompromised patients were vaccinated in our cohort, and there is probably a lack of power to detect an interaction between both variables, or in other words a diminished effect of vaccination in immunocompromised patients. Moreover, immunocompromised patients remained at high risk of a severe form (OR=2·55) even after adjustment for parameters such as other risk factors and vaccination (Table 1). Decreased efficacy in this population was previously noted for virtually all vaccines, particularly influenza vaccines 24, and was also observed with Covid-19 vaccines 25. This led some countries to recommend an additional dose of the vaccine to these patients, and the prophylactic use of passive immunisation by monoclonal antibodies in immunocompromised patients with no or insufficient antibody response to vaccines 26.
As older patients were not only at risk of severe forms, but were also vaccinated earlier in the vaccination campaign, a lack of protection may have existed due to the aforementioned waning of neutralising antibody titres over time after vaccination. The multivariate analysis we specifically applied to vaccinated patients showed that both age and time from vaccination were associated with a significant increase in 28-day mortality. Once again, this confirms the need for a booster vaccine injection, particularly in those most at risk.
Our study has several limitations. Firstly, due to the retrospective design of the study, some data retrieved from the medical files may have been erroneous or not completely accurate. However, the high number of studied patients, and comprehensive analysis on patients of known risk factors for adverse outcomes helped to reduce risk of biases. Secondly, we did not measure anti-Spike antibody titers in vaccinated patients at admission, a previously described predictive risk factor for SARS-CoV-2 infection and Covid-19 27; it would have helped to better understand whether severe infections in vaccinated subjects were due to waning of the neutralising antibody vaccine response. Moreover, we used ICU admission as a severity criterion; however, it is usually considered that older patients with many comorbidities would not benefit from ICU admission in terms of survival. Therefore, not being admitted to ICU does not necessarily mean that there is no severe form of Covid-19.
In conclusion, patients hospitalised with a SARS-CoV-2 infection with the delta VOC have less marked severity criteria when they are vaccinated more than 14 days before hospitalisation, especially in case of elderly patients. These data should be provided when communicating information about Covid-19 vaccination with subjects at risk of severe Covid-19 because of their age or comorbidities.
Data Availability
All data produced in the present study are available upon reasonable request to the authors