Authors: Xinning Tong, Le Gao, Ian C K Wong, Vivien K Y Chan, Angel Y S Wong, Judith C W Mak, Jacqueline K Y Yuen, Mark Jit, Ivan F N Hung, Kai Hang Yiu, Xue Li
Categories: Original Article, PCV13, PPSV23, Pneumococcal vaccine, cardiovascular diseases, older adults, AcademicSubjects/MED00860
Source: International Journal of Epidemiology
Doi: 10.1093/ije/dyae005
Recommendations around the use of 23-valent pneumococcal polysaccharide vaccine (PPSV23) and 13-valent pneumococcal conjugate vaccine (PCV13) seldom focus on potential benefits of vaccine on comorbidities. We aimed to investigate whether sequential vaccination with PCV13 and PPSV23 among older adults would provide protection against cardiovascular diseases (CVD) compared with using a single pneumococcal vaccine.
We conducted a Hong Kong-wide retrospective cohort study between 2012 and 2020. Adults aged ≥65 years were identified as receiving either a single or sequential dual vaccination and followed up until the earliest CVD occurrence, death or study end. To minimize confounding, we matched each person receiving a single vaccination to a person receiving sequential vaccination according to their propensity scores. We estimated the hazard ratio (HR) of CVD risk using Cox regression and applied structural equation modelling to test whether the effect of sequential dual vaccination on CVD was mediated via the reduction in pneumonia.
After matching, 69 390 people remained in each group and the median (interquartile range) follow-up time was 1.89 (1.55) years. Compared with those receiving a single vaccine, those receiving sequential dual vaccination had a lower risk of CVD [HR (95% CI): 0.75 (0.71, 0.80), *P < *0.001]. Post-hoc mediation analysis showed strong evidence that the decreased CVD risk was mediated by the reduction in all-cause pneumonia.
Sequential dual pneumococcal vaccination was associated with lower risk of CVD compared with single-dose PCV13 or PPSV23 in older adults. Such additional CVD benefits should be considered when making decisions about pneumococcal vaccination.
Keywords: Pneumococcal vaccine, PCV13, PPSV23, cardiovascular diseases, older adults
Respiratory infection is a well-recognized risk factor associated with atherosclerosis that is capable of triggering cardiovascular diseases (CVD) with long-term comorbidities due to oxidative stress.^1^^,^^2^ Pneumonia is a major acute respiratory infectious disease with severity ranging from mild to life-threatening for people of all ages. It is associated with CVD both in the short term (during the acute phase of respiratory infection) and long term (from several weeks up to 10 years after the pneumonia episode).^1^^,^^3^Streptococcus pneumoniae is a major causative pathogen for pneumonia and invasive diseases,^4^ but pneumococcal vaccination is effective in preventing pneumococcal pneumonia caused by S. pneumoniae.^5^
Older adults are at high risk of both pneumonia and CVD.^3^^,^^6^ Preventing these two life-threatening diseases would help alleviate two major sources of adult disease burden. Vaccines targeted at respiratory pathogens can reduce the risk of CVD by decreasing systemic inflammatory cytokines caused by respiratory infection.^7^^,^^8^ Epidemiological studies have shown that influenza vaccination protects against both acute myocardial infarction and mortality from cardiovascular disease and stroke.^8^^,^^9^ Currently, the two main vaccines used for preventing infection caused by different serotypes of S. pneumoniae are the 13-valent pneumococcal conjugate vaccine (PCV13) and the 23-valent pneumococcal polysaccharide vaccine (PPSV23). Recommendations on the use of PCV13 and PPSV23 among older adults still differ between countries. The main question is whether immunocompetent older adults should receive sequential doses of PCV13 and PPSV23 or a single dose of either.^10–13^ Evidence supporting either recommendation mainly focuses on vaccine effectiveness against invasive pneumococcal diseases and vaccine-type community acquired pneumonia without consideration of other benefits of preventing infection-related comorbidities.^14^^,^^15^
A combination of both vaccines may induce a higher immune response than a single vaccine,^7^ thereby decreasing the risk of infection and adverse health consequences. In this population-based cohort study using electronic medical records (EMRs) in Hong Kong, we aimed to investigate whether sequential dual pneumococcal vaccination for older adults would provide additional protection against CVD compared with using a single vaccine, in order to inform evidence-based vaccine policy.
We used the Hong Kong territory-wide EMR database—Clinical Data Analysis and Reporting System. The database covers all citizens who used public health services with 11 million cumulative patient records.^16^ Anonymized records of demographics, dates and registered causes of death, immunizations, diagnoses and prescriptions from outpatient, inpatient and emergency settings are included in the database for research and audit purposes. Detailed descriptions and analyses of record accuracy are available in other publications.^9^^,^^17^^,^^18^
This study used a retrospective cohort study design (Figure 1). We reported this study in accordance with the STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) guideline. PCV13 was recommended for older adults starting from year 2014 in Hong Kong^19^ and was added to the local Vaccination Subsidy Scheme for eligible adults in 2017. Hence PCV13 uptake was low prior to 2017.^20^ We recruited all eligible ≥65-year-old adults who received at least one dose of pneumococcal vaccination between 1 January 2017 and 31 December 2020. Since an interval of ≥5 years is needed between two doses of PPSV23 and no repeated PCV13 dose is recommended in this age group,^10–13^ we traced the vaccination records of recruited older adults for ≤5 years. We excluded participants (0.05%) with multiple doses of the same type of pneumococcal vaccine and divided the remaining cohort into a single-dose group (PPSV23 or PCV13 only) and a sequential dual vaccination group (both PCV13 and PPSV23), based on their vaccination records within 5 years before recruitment until 31 December 2020. The index date was the date of the first recorded vaccine received for the single-dose group and the date of the second recorded vaccine dose received for the sequential group. Consequently, those in the sequential group could have received a prior vaccine before the age of 65 years whereas all individuals would be ≥65 years old when they were included in the cohort. As we focused on immunocompetent older adults, we excluded participants with the immunocompromised conditions listed in Supplementary Table S1 (available as Supplementary data at IJE online). The remaining participants were followed up from the index date to the first occurrence of any CVD outcome of interest, death or study end date (31 December 2020), whichever came earliest.
Figure 1. Schema of study design. (a) Single pneumococcal vaccine (PV) group—date of cohort entry. Sequential PV Group 1, who received the first PV before cohort entry—date of cohort entry. Sequential PV Group 2, who received the second PV after cohort entry—date of the second PV vaccination. (b) Cardiovascular diseases of interest (atrial fibrillation, acute coronary syndrome, congestive heart failure, stroke or CVD-related death), death or study end
The main outcome was hospitalization for CVD, including atrial fibrillation, acute coronary syndrome, congestive heart failure, stroke or CVD-related death. The secondary outcomes were individual CVD events. The CVD outcomes, excluded conditions, comorbidities and all-cause pneumonia, were identified using International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes and death records were identified using International Classification of Diseases, 10th Revision (ICD-10) codes (Supplementary Table S1, available as Supplementary data at IJE online). Subgroup analyses were conducted by age group (65–74, 75–84 and ≥85 years) and by the presence of chronic disease (any chronic heart, liver and lung disease or diabetes, Supplementary Table S1, available as Supplementary data at IJE online).
We used 1 propensity score (PS) matching with a caliper width equal to 0.2 times the standard deviation of the propensity score to minimize confounding and balance the baseline characteristics.^21^ The propensity scores were derived using logistic regression with variables including age, sex and medical history from 1993 until time of recruitment. Variables with a between-group standardized mean difference of <0.1 were considered well balanced.^22^ We report the descriptive statistics of baseline characteristics before and after propensity score matching. We used means (SD) or medians [interquartile ranges (IQRs)] to report continuous variables and used number (percentage) to report dichotomous variables. We estimated hazard ratios (HRs) using Cox regression with the proportional hazard assumption checked using Schoenfeld residuals with the R ‘survival’ package. The likelihood ratio test between the crude regression model and the adjusted model was used to test the significance of variables.
We further conducted several sensitivity analyses to assess the robustness of the main
We chose hospitalization for dehydration as the negative outcome control to assess the impact of residual confounding.
We further tested whether the effect of sequential pneumococcal vaccination on the risk of CVD was mediated via the reduction in all-cause pneumonia using structural equation modelling based on binary probit link with the R ‘lavaan’ package. Individuals were considered to have new-onset all-cause pneumonia if they had relevant diagnoses from the index date to death, incidence of CVD or end of study (censoring), whichever came first, and if they had no pneumonia-related diagnoses during the 1-year washout period before the index date.^24^ Model goodness-of-fit was assessed using the root-mean-square error of approximation, comparative fit index and Tucker–Lewis index.
Data analysis and visualization were conducted using the R program (Version 4.1.2) and cross-checked individually by X.T., L.G. and V.C.
We identified 262 421 older adults in total from the database who received PPSV23 or/and PCV13 during the study period. After participant selection, we included 230 119 people (single 157 244; sequential 72 875). A flowchart of participant recruitment is shown in Supplementary Figure S1 (available as Supplementary data at IJE online). The baseline characteristics of all recruited participants before and after matching are listed in Table 1. Before matching, participants in the sequential dual vaccination group had a higher proportion of males (54.5%) and were younger (mean ± SD: 72.12 ± 5.84) than those in the single-dual-vaccine group (48.5%; mean ± SD: 76.09 ± 7.57). Generally, participants in the sequential group had more comorbidities than those in the single group with a higher Charlson Comorbidity Index score (mean ± SD: 0.62 ± 0.83 vs 0.53 ± 0.87). After 1 propensity score matching, 69 390 patients remained in each group with baseline variables all well balanced.
Tables 2 and 3 show the results of the main and subgroup analyses. The median (IQR) follow-up times for participants were 1.96 (1.62) years in the single group and 1.83 (1.37) years in the sequential group. The crude incidence rates (95% CI) of CVDs were 30.91 (29.94, 31.91) and 21.45 (20.61, 22.31) per 1000 person-years in the single and sequential vaccination groups, respectively. The incidences of individual CVD in descending order were atrial fibrillation [single 7.97 (7.49, 8.48); sequential 5.06 (4.67, 5.49)], congestive heart failure [single 7.80 (7.32, 8.30); sequential 5.16 (4.76, 5.58)], acute coronary syndrome [7.23 (6.77, 7.71); 4.91 (4.51, 5.32)], stroke [5.87 (5.46, 6.31); 4.43 (4.06, 4.82)] and CVD-related deaths [2.12 (1.88, 2.39); 1.24 (1.05, 1.45)].
Compared with those receiving a single dose, participants receiving sequential vaccination of PPSV23 and PCV13 had a significantly lower risk of adverse CVD outcomes [HR (95% CI): 0.75 (0.71, 0.80), *P < *0.001]. The analysis by individual diseases and a series of subgroup analyses all showed a lower risk of CVD associated with sequential pneumococcal vaccination (HR ranged from 0.65 to 0.85; Tables 2 and 3). The cardio-protective effect had a higher effect size for CVD-related death [HR (95% CI): 0.65 (0.51, 0.82), *P < *0.001] followed by atrial fibrillation [HR (95% CI): 0.70 (0.62, 0.79.), *P < *0.001], congestive heart failure [HR (95% CI): 0.72 (0.64, 0.81), *P < *0.001], acute coronary syndrome [HR (95% CI): 0.75 (0.67, 0.85), *P < *0.001] and stroke [HR (95% CI): 0.79 (0.69, 0.90), *P < *0.001]. In addition, similar effect sizes (P-value for interaction, 0.39) were found among older people with chronic comorbidities [HR (95% CI): 0.75 (0.68, 0.82), *P < *0.001] and those without chronic comorbidities [HR (95% CI): 0.73 (0.67, 0.79, *P < *0.001)]. The effect sizes of protective effect from sequential vaccination decreased with age (trend of P-value <0.001) in older participants [65–74 years old HR (95% CI): 0.75 (0.70, 0.82), *P < *0.001; 75–84 years 0.84 (0.75, 0.94), *P = *0.003; ≥85 years 0.85 (0.73, 0.98), *P = *0.031]. Sensitivity analyses (Supplementary Table S2, available as Supplementary data at IJE online) showed consistent protective effects of sequential vaccination compared with a single dose (HRs ranged from 0.74 to 0.77). The protective effect size of sequential pneumococcal vaccination was lower among participants without a history of CVD [HR (95% CI): 0.76 (0.68, 0.84), *P < *0.001] compared with our main analysis. Lower risk of all-cause pneumonia was observed among sequential vaccination recipients compared with those who received a single dose [HR (95% CI): 0.76 (0.70, 0.82), *P < *0.001]. Sensitivity analysis of hospitalization for dehydration showed no protective effect of sequential vaccination on dehydration [HR (95% CI): 1.08 (0.79, 1.49), *P = *0.624], which supports that large residual confounding would not result in significant selection bias using the PS-matching-based study design. Sensitivity analyses on the potential bias from late entry (Supplementary Table S3, available as Supplementary data at IJE online) showed that when considering the comparability of two groups of cohort entry through the use of time-dependent propensity score matching, the protective effect size of sequential pneumococcal vaccination is 0.14 [HR (95% CI): 0.86 (0.80, 0.91)]. In the time-varying exposure analysis, we found that, compared with single pneumococcal vaccination, sequential vaccination would help to decrease CVD risk by 12% [HR (95% CI): 0.88 (0.81, 0.96)]. These analyses, although still demonstrating a beneficial effect, were lower than the effect size found in the other sensitivity analyses.
Figure 2 illustrates the total effect and direct effect of sequential vaccination on the risk of CVD, and the mediated effect acting through new-onset all-cause pneumonia, compared with single-dose vaccination. The structural equation model achieved fair goodness-of-fit and confirmed the strong evidence of the total protective effect of sequential vaccination on the risk of CVD (*P < *0.001) in the main analysis. The reduction in CVD risk associated with sequential vaccination was mediated by a reduction in new-onset all-cause pneumonia (*P < *0.001), which accounted for 9.31% (proportion of total effect attributable to indirect effect) of the total effect.
Figure 2. Mediating effect of sequential pneumococcal vaccination with PPSV23 and PCV13 on the risk of CVD from all-cause pneumonia. The values are binary probit estimates of the total effect of sequential pneumococcal vaccination on the risk of CVD, the direct effects of sequential pneumococcal vaccination on the risk of CVD, sequential pneumococcal vaccination on all-cause pneumonia and all-cause pneumonia on the risk of CVD, and the indirect effect of sequential pneumococcal vaccination on the risk of CVD via all-cause pneumonia. CFI, comparative fit index; CVD, cardiovascular disease; PPSV23, 23-valent pneumococcal polysaccharide vaccine; PCV13, 13-valent pneumococcal conjugate vaccine; RMSEA, root-mean-square error of approximation; TLI, Tucker–Lewis index
Heart diseases are the leading cause of death globally and represent one-third of all global deaths, which renders CVD prevention meaningful in consideration of different risk factors.^25^ To the best of our knowledge, this is the first study to investigate the cardiovascular benefits from sequential dual pneumococcal vaccination with PCV13 and PPSV23 compared with a single dose of either among older adults—a high-risk group for both pneumonia and CVD.^3^ Our study found sequential pneumococcal vaccination using different vaccine types would provide protection against CVD in older adults and there was evidence that the beneficial effect was mediated through the reduction in all-cause pneumonia—a proxy for pneumococcal pneumonia in our study. A randomized clinical trial and its 4-year extension suggested that sequential vaccination with PCV13 and PPSV23 improved immunity responses to more serotypes compared with PCV13 or PPSV23 alone after a 3- to 4-year interval.^26^^,^^27^ Our sensitivity analysis on all-cause pneumonia and mediation analysis support the hypothesized protective mechanism of sequential vaccination on the reduced risk of CVD via effective prevention of pneumonia. Although these are indirect findings, a sequential combination of PCV13 and PPSV23 would have better prevention effects on pneumonia, potentially leading to a decrease in systemic inflammation caused by infection and therefore a reduction in the occurrence of CVD.
Currently, PCV13 and PPSV23 guidelines are inconsistent because the focus is on the effectiveness of pneumonia prevention but they rarely consider the effect of vaccination on alleviating comorbidities. Guidelines around pneumococcal vaccination of older adults in Canada and the UK both recommended a single dose of PPSV23, regardless of any chronic diseases.^11^^,^^28^ Meanwhile, Hong Kong and Australia recommend sequential pneumococcal vaccination among those who have chronic diseases.^10^^,^^19^ In fact, most adults >65 years old are not immunocompromised but have chronic comorbidities.^29^ Therefore, it would be meaningful to provide evidence in current guidelines for immunocompetent older adults from the perspective of comorbidity benefits. Previous epidemiological evidence on the effect of single pneumococcal vaccination on CVD risk showed that individuals at high risk of CVD (with chronic disease conditions or a history of CVD) have a lower risk of CVD, CVD-related deaths or all-cause mortality if they receive at least one dose of PPSV23 or PCV13 compared with those without any type of pneumococcal vaccination.^30^^,^^31^ Our study supports that, compared with receiving only one dose of PCV13 or PPSV23, sequential vaccination with both provides protection by reducing CVD risk, although we did not observe a significant difference between older adults with and without chronic diseases (P-value for interaction 0.39). Additionally, we observed that the protective effects of sequential vaccination decreased with age. One reason could be that vaccination is insufficient to overcome other risks factors of CVD, such as very advanced age, impaired immunity and inferior health conditions. Nevertheless, our findings support that elderly patients would benefit from sequential pneumococcal vaccination in view of cardiovascular disease alleviation. Notably, the decision and recommendation for single or sequential vaccination should consider multiple factors such as the risk–benefit, acceptability and preferences, and cost-effectiveness.
Our 9-year retrospective cohort study that covered 230 119 eligible participants suggests a protective effect from sequential PPSV23 and PCV13 vaccination on CVD compared with receiving one dose of PPSV23 or PCV13 only. This study could provide evidence to inform pneumococcal vaccination recommendations among immunocompetent older adults, although a few limitations exist. First, we used EMRs to identify patients’ immunization status, so the accuracy of immunization information cannot be validated. Misclassification for the single-vaccination group may exist and we may have underestimated the effect size from the sequential vaccination group. Second, our study lacks data about participants’ demographic variables such as dietary, smoking status, alcohol consumption and healthcare-seeking behaviour (e.g. individual social economic level, accessibility to healthcare resources), which may have given rise to residual confounding. We used propensity scores to minimize the health differences between groups and some included covariates could reflect the effects of lifestyle factors that were not adjusted for. The negative outcome control analysis also supports a small residual confounding without a significant impact on the study conclusion. Third, we did not adjust for the seasonal influenza vaccine as it was not usually recorded in routinely collected EMRs. We assumed the uptake of the influenza vaccine would be similar between two groups since the influenza vaccine is usually given at the same time as the pneumococcal vaccine in Hong Kong. Fourth, the generalizability of our finding that sequential pneumococcal vaccination has long-term beneficial effects on CVD should be interpreted with caution. As PCV13 was supplied in the local Vaccination Subsidy Scheme for eligible adults in 2017, receiving sequential vaccination was prevalent after year 2017. In consequence, our follow-up period was only ∼2 years, which may not have been long enough for cardiovascular disease that results from a chronic inflammatory cascade to develop. Future studies with longer study periods are recommended to test our findings. In addition, our study was not designed to examine which vaccine should be taken first in the sequential group for better CVD prevention. Instead, we aimed to test the hypothesis that sequential vaccination will provide a stronger immune response and broader serotype coverage that would further prevent the infection-triggered CVD and be more effective than a single pneumococcal vaccination alone. It would also be meaningful to further investigate whether the sequential order of PPSV23 and PCV13 would influence the beneficial effects on CVD. Fifth, individuals in the sequential group were followed up since the later vaccination date, which would have induced the potential for ‘late-entry bias’ as they had to survive until the second vaccination. This situation may have meant that people in the sequential group were in better health than those in the single group, so we may have overestimated the protective effects of the sequential vaccination. Our sensitivity analysis (Supplementary Table S3, available as Supplementary data at IJE online) accounted for this late-entry bias. It supports the assumption that this bias had a diminishing effect on the observed effect size compared with our main analysis. Despite this potential bias, we still observed significant protective effects on CVD risk resulting from the implementation of sequential pneumococcal vaccination. Lastly, due to the limitation of our database, the majority of records of patients diagnosed with pneumonia do not include laboratory records about causative pathogens; we were therefore unable to ascertain whether the recorded pneumonia was pneumococcal-related. We used all-cause pneumonia as a proxy outcome for pneumococcal pneumonia, which may have led to an underestimation of the effect size and proportion of mediation in the overall relationship.
Sequential pneumococcal vaccination with PCV13 and PPSV23 is associated with a reduced risk of acute coronary syndrome, congestive heart failure, stroke and CVD-related deaths compared with a single vaccination with either PCV13 or PPSV23. The effect is partially mediated by the prevention of all-cause pneumonia.
This study obtained ethics approval from the Institutional Review Board via The University of Hong Kong/Hospital Authority Hong Kong Western Cluster (UW 21–271).
We thank Professor Esther W. Chan from Department of Pharmacology and Pharmacy and Dr Celine S.L. Chui from School of Nursing, University of Hong Kong for advice on study design; we thank Mr Kuan Peng from Department of Medicine, University of Hong Kong for assistance with data collection in this study; we thank Ms Yin Zhang from Department of Medicine, University of Hong Kong for her administrative support. We also thank Ms Lisa Lam for English proofreading.
Xinning Tong, Department of Medicine, School of Clinical Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China; Department of Orthopaedics, The Eighth Affiliated Hospital, Sun Yat-sen University, Shenzhen, China.
Le Gao, Department of Pharmacology and Pharmacy, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China.
Ian C K Wong, Department of Pharmacology and Pharmacy, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China; Laboratory of Data Discovery for Health (D^2^4H), Hong Kong Science and Technology Park, Sha Tin, Hong Kong SAR, China; Aston School of Pharmacy, Aston University, Birmingham, UK.
Vivien K Y Chan, Department of Pharmacology and Pharmacy, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China.
Angel Y S Wong, Laboratory of Data Discovery for Health (D^2^4H), Hong Kong Science and Technology Park, Sha Tin, Hong Kong SAR, China; London School of Hygiene and Tropical Medicine, London, UK.
Judith C W Mak, Department of Pharmacology and Pharmacy, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China.
Jacqueline K Y Yuen, Department of Medicine, School of Clinical Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China.
Mark Jit, Laboratory of Data Discovery for Health (D^2^4H), Hong Kong Science and Technology Park, Sha Tin, Hong Kong SAR, China; London School of Hygiene and Tropical Medicine, London, UK.
Ivan F N Hung, Department of Medicine, School of Clinical Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China.
Kai Hang Yiu, Department of Medicine, School of Clinical Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China.
Xue Li, Department of Medicine, School of Clinical Medicine, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China; Department of Pharmacology and Pharmacy, Li Ka Shing Faculty of Medicine, The University of Hong Kong, Hong Kong SAR, China; Laboratory of Data Discovery for Health (D^2^4H), Hong Kong Science and Technology Park, Sha Tin, Hong Kong SAR, China.
We are unable to directly share the patient-level data since the data custodian, the Hong Kong Hospital Authority, which manages the Clinical Data Analysis and Reporting System, has not given permission. However, researchers can apply to access the data via the Hospital Authority Data Sharing Portal (https://www3.ha.org.hk/data) for research purposes. The statistical procedures and R codes used in this study are available upon request.
Supplementary data are available at IJE online.
Study X.L. Study X.L., X.T., L.G., V.C., M.J. Data extractions, analysis and cross-check: X.T., L.G., V.C. Drafting of the X.T., V.C., X.L. Data all authors. Clinical M.J., J.Y., K.Y., I.H. Critical revision of the all authors. Resource and funding X.L., I.C.K.W. Study X.L.
M.J.’s posts were partly funded by the Laboratory of Data Discovery for Health (D^2^4H). Hence, AIR@InnoHK, administered by the Innovation and Technology Commission, partly supported this work.
X.L. received research grants from Hong Kong Health and Medical Research Fund (HMRF, HMRF Fellowship Scheme, HKSAR), Research Grants Council Early Career Scheme (RGC/ECS, HKSAR), Research Grants Council Research Impact Fund (RGC/RIF, HKSAR), Janssen and Pfizer; internal funding from the University of Hong Kong; consultancy fee from Merck Sharp & Dohme; she is also an non-executive director of Advanced Data Analytics for Medical Science (ADAMS) Limited Hong Kong, all are unrelated to this work. I.C.K.W. reports research funding outside the submitted work from Amgen, Bristol-Myers Squibb, Pfizer, Janssen, Bayer, GSK Novartis, the Hong Kong RGC and the Hong Kong Health and Medical Research Fund, National Institute for Health Research in England, European Commission, National Health and Medical Research Council in Australia, and also received speaker fees from Janssen and Medice in the previous 3 years. He is also an non-executive director of Jacobson Medical Hong Kong, an non-executive director of Advanced Data Analytics for Medical Science (ADAMS) Limited, Hong Kong, all are unrelated to this work. All other authors have no reports on conflict of interest.
We are unable to directly share the patient-level data since the data custodian, the Hong Kong Hospital Authority, which manages the Clinical Data Analysis and Reporting System, has not given permission. However, researchers can apply to access the data via the Hospital Authority Data Sharing Portal (https://www3.ha.org.hk/data) for research purposes. The statistical procedures and R codes used in this study are available upon request.