Real-world Effectiveness of the SARS-CoV-2 mRNA Vaccines in Preventing Confirmed Infection in Patients on Chronic Hemodialysis

Abstract Background Persons on chronic hemodialysis have a significantly diminished humoral immune response to SARS-CoV-2 vaccines. Whether this translates to reduced vaccine effectiveness (VE) is unknown. Methods We used the US Department of Veterans Affairs COVID-19 Shared Data Resource to identify all Veterans who were tested for SARS-CoV-2 between January 26, 2021 and August 31, 2021. Using International Classification of Diseases, 10 th edition codes and attendance at a dialysis clinic or center, we identified those who were on chronic hemodialysis. We used a test-negative, case-control design using a doubly-robust logistic regression model to determine the VE of the BNT-162b2 (Pfizer) or mRNA-1273 (Moderna) vaccines in preventing confirmed SARS-CoV-2 infection. Results Among 847,199 Veterans tested for SARS-CoV-2 between January 26, 2021 and August 31, 2021, there were 6,076 Veterans on chronic hemodialysis. Among those, we identified 1,270 cases (580 fully vaccinated) and 2,959 controls (2,120 fully vaccinated). The overall VE >14 days after the second dose in preventing documented infection was 68.2% (95% CI:62.6,72.9). VE was 68.9% (95% CI:61.9,74.7) for Pfizer-BNT-162b2 and 66.7% (95% CI:58.9,73.0) for Moderna-mRNA-1273 vaccine. There was no difference in VE by age (<70 vs. >70 years), race or sex. There were no events recorded in persons with a Charlson’s comorbidity index score of <2. Conclusion VE of two doses of current mRNA vaccines in preventing SARS-CoV-2 infection in persons on chronic hemodialysis is lower than historic VE rates in the general population. Effect of additional doses in improving VE in this special population needs further study.

A c c e p t e d M a n u s c r i p t In randomized, placebo controlled clinical trials, mRNA vaccines developed by Pfizer  and Moderna (mRNA-1273) demonstrated 94%-95% efficacy against symptomatic infection and nearly 100% efficacy against severe disease or death. [1,2] Subsequent effectiveness studies in the real-world settings have reported similarly high rates of effectiveness in preventing confirmed infection and nearly 100% effectiveness in preventing severe disease or death. [3,4] However, the effectiveness wanes with the passage of time and in patients infected with certain variants of concern. [5] Patients with chronic kidney disease who are on chronic hemodialysis constitute a particularly challenging group due to significant immune suppression. While being at a higher risk of adverse outcomes including intensive care unit admission and mortality, they have a diminished humoral response to SARS-CoV-2 vaccination. [6][7][8][9] They also have a higher rate of SARS-CoV-2 breakthrough infections after vaccination. [10] Despite such risk, during the time this study was conducted, persons on chronic hemodialysis were not included in the list of immunocompromised persons recommended for an additional dose of the vaccine. [11] Vaccine effectiveness of the SARS-CoV-2 vaccines in patients on chronic hemodialysis in the real-world setting is unknown. This information is critical in determining the optimal prevention strategies in this population. To address this gap in knowledge, we determined the effectiveness of the Pfizer-BNT-162b2 and Moderna-mRNA-1273 vaccines among Veterans in care in the United States Department of Veterans Affairs healthcare system (VA) who were on chronic hemodialysis.

Study Population
In response to the SARS-CoV-2 pandemic, the VA rapidly created a national VA COVID-19 Shared Data Resource. Using case definitions and data mapping which were validated collaboratively across the VA, it contains information on all Veterans with a confirmed laboratory diagnosis of SARS-CoV-2 A c c e p t e d M a n u s c r i p t

Statistical Analyses
We used test-negative design to determine the effectiveness of vaccination against confirmed SARS-CoV-2 infection. This design is a widely accepted standard to determine vaccine effectiveness in a population after the introduction of a vaccine. [12][13][14] We have also used this design to report SARS-CoV-2 vaccine effectiveness in the real-world settings. [4,5,[15][16][17] We modeled the effect of vaccines on testing positive using a doubly robust approach, which protects against the misspecification of the propensity score model or the model for the vaccine's effect, by adjusting for covariates in both models separately. [18][19][20] The probability of vaccination is modeled using gradient-boosted regression models using age, sex, race, body mass index, testing facility and Charlson score as input variables. [21] We estimated the adjusted odds ratios (95% confidence M a n u s c r i p t intervals) from a logistic regression model weighted by inverse probability of vaccination weights, controlling for age, sex, race, body mass index, testing facility and Charlson score. Vaccine effectiveness (VE) was determined using the following formula: . We determined overall vaccine effectiveness >14 days after the second dose of the vaccine as our primary outcome of interest. We also determined vaccine effectiveness separately for patients who were only partially vaccinated by identifying partially vaccinated cases to controls never testing positive. Finally, we determined VE among subgroups by age, sex, race and Charlson's comorbidity index score. For all point estimates of VE, we calculated the corresponding 95% confidence intervals. [22,23] Nearly 97% of the patients in our final study group received the Pfizer-BNT-162b2 or the Moderna-mRNA-1273. We excluded those who received any other vaccine for our main analyses. As a sensitivity analysis, we performed a 1:1 propensity score matched analysis followed by conditional logistic regression (accounting for strata of matched pairs) to estimate VE. Details of propensity score modeling and imbalance evaluation are provided in supplementary materials, appendices 1 and 2. Propensity score matching was performed using the same propensity scores estimated for the weighted analysis. Nearest-neighbors type matching with caliper 0.25 was performed using R package matchit.
A c c e p t e d M a n u s c r i p t

Ethical Considerations
The study was approved by the Institutional Review Board at the VA Pittsburgh Healthcare System. A waiver of informed consent was granted for the study.

Results
We identified 847,199 Veterans who were tested for SARS-CoV-2 between January 26, 2021 and  There was no difference in vaccine effectiveness by age group (<70 vs. >70 years), race or sex. (Table   3) There were no events recorded in persons with a Charlson's comorbidity index score of <2.
In sensitivity analyses using a 1:1 propensity score matched approach, covariate imbalance was higher than for the primary analysis (supplementary table S1), and the estimated overall vaccine effectiveness was somewhat lower than the primary estimate (supplementary table S2 Patients on chronic hemodialysis are at a higher risk of SARS-CoV-2 infection, and when infected, are more likely to experience poorer clinical outcomes. This is compounded by the suboptimal humoral response to the currently available SARS-CoV-2 vaccines in these patients. While our results provide reassurance regarding the effectiveness of the mRNA vaccines in preventing documented infection, the observed vaccine effectiveness of 68% is lower than the 95% effectiveness in the general VA population [17] and observed in randomized clinical trials, [1,2] it is sufficiently high to significantly impact the course of disease in this highly vulnerable population. Two factors that significantly reduce vaccine effectiveness of the mRNA vaccines against documented infection are passage of time [5] and infection with the Delta variant. [24] The study period for the current study extended through August 2021, when the Delta variant accounted for the overwhelming majority of infections and a substantial proportion of the study population had been vaccinated several months earlier. [25] An overwhelming majority of our study population had comorbidities, the most common being hypertension, diabetes, and cardiovascular disease. Presence of these comorbidities is an independent risk factor for poor clinical outcomes. Vaccine effectiveness among persons with these comorbidities has not been studied. It is unclear whether the lower vaccine effectiveness rate in our chronic hemodialysis population was due to these comorbidities, the resultant renal failure, or being A c c e p t e d M a n u s c r i p t on hemodialysis itself. Recently, a third dose of the Pfizer-BNT162b2 vaccine has been approved for certain high-risk populations. Early data also demonstrate a significant increase in neutralization of the Beta and the Delta variants after the third dose, [26] and recent studies have demonstrated a significant increase in vaccine effectiveness in reducing infections and mortality among in persons who receive a third dose compared with those who only received two dose. [27,28] Whether a booster will enhance vaccine effectiveness in persons on chronic hemodialysis requires further study.
A limitation of our study is the lack of information on SARS-CoV-2 variants of concern. We have previously demonstrated a somewhat lower effectiveness of the Pfizer-BNT-162b2 vaccine against the Beta and Delta variants. [4,24] However, with an overall effectiveness of 68%, the benefit of the vaccine supports its use regardless of the variant type at this time. It is possible that future variants may be less amenable to the current vaccine, which would necessitate altering the vaccines to protect against such potential future variants. We also did not study clinical outcomes like severe disease or death. Again, with a vaccine effectiveness of 68% in preventing documented infection necessarily means that a large proportion of complications would be prevented. Varying testing and vaccination practices in different geographic region can impact the results of such studies. We mitigated this by matching the cases and control on the geographic location of testing. This matching was done to the level of the center or facility where testing was performed. Since we did not include patients on peritoneal dialysis in our study, our findings cannot be generalized to those patients.
Nearly 97% of the study subjects were male, therefore the results cannot be generalized to females.
Our study population was predominantly male and older compared with the general population, and the results should not be generalized to the larger national population. M a n u s c r i p t

Data Availability Statement
Requests for data must be directed to the Veterans Health Administration at the Department of Veterans Affairs. Any request must fulfil all requirements for data sharing according the existing laws, regulations, and policies of the Department of Veterans Affairs.

Diagnosis Codes
International Classification of Diseases (ICD) 9 th and 10 th edition codes related to dialysis are provided in supplementary materials appendix 4.
A c c e p t e d M a n u s c r i p t M a n u s c r i p t   A c c e p t e d M a n u s c r i p t 20