Persistent increase of cardiovascular and cerebrovascular events in COVID-19 patients: a 3-year population-based analysis

Abstract

Aims

We evaluated the incidence and relative risk of major post-acute cardiovascular consequences of SARS-CoV-2 infection in a large real-world population from a primary care database in a region at moderate cardiovascular risk followed up in the period 2020–22.

Methods and results

This is a retrospective cohort analysis using data from a cooperative of general practitioners in Italy. Individuals aged >18 affected by COVID-19 starting from January 2020 have been followed up for 3 years. Anonymized data from 228 266 patients in the period 2020–22 were considered for statistical analysis and included 31 764 subjects with a diagnosis of COVID-19. An equal group of subjects recorded in the same database in the period 2017–19 was used as propensity score-matched comparison as an unquestionable COVID-19-free population. Out of the 228 266 individuals included in the COMEGEN database during 2020–22, 31 764 (13.9%) were ascertained positive with SARS-CoV-2 infection by a molecular test reported to general practitioners. The proportion of individuals with a new diagnosis of major adverse cardiovascular and cerebrovascular events was higher in the 2020–22 COVID-19 group than in the 2017–19 COMEGEN propensity score-matched comparator, with an odds ratio of 1.73 (95% confidence interval: 1.53–1.94; P < 0.001). All major adverse cardiovascular and cerebrovascular events considered showed a significantly higher risk in COVID-19 individuals. Incidence calculated for each 6-month period after the diagnosis of COVID-19 in our population was the highest in the first year (1.39% and 1.45%, respectively), although it remained significantly higher than in the COVID-19-free patients throughout the 3 years.

Conclusion

The increase of cardiovascular risk associated with COVID-19 might be extended for years and not limited to the acute phase of the infection. This should promote the planning of longer follow-up for COVID-19 patients to prevent and promptly manage the potential occurrence of major adverse cardiovascular and cerebrovascular events.

Graphical Abstract

SARS-COV-2, Atrial fibrillation, Myocardial infarction, Heart failure, Stroke, Viral infection, Cardiovascular disease recovery

Time of primary review: 35 days

1. Introduction

Numerous observational studies have consistently shown that SARS-Cov-2 infection (COVID-19) is often associated with the development of major adverse cardiovascular (CV) and cerebrovascular events (MACCE).^1–3 Higher risk is associated with male gender, the presence of CV risk factors, and a positive CV history,^4–9 although most of the previous studies were restricted to hospitalized individuals, had a relatively limited follow-up, and were often focused on a specific CV outcome. Post-acute CV consequences of SARS-CoV-2 infection (PASC-CVD) include acute coronary syndromes, myopericarditis, heart failure (HF), arrhythmias, and vascular complications (e.g. arterial and venous thromboembolism and vasculitis).¹⁰ At present, it is unclear whether the increased risk of PASC-CVD in the general population persists over a longer time.^11–18

In the present study, we evaluated the incidence and risk of major PASC-CVD in a real-world large population study of subjects from a primary care database followed up for up to 3 years during the pandemic in the period 2020–22 and compared with a pre-COVID-19 pandemic population as derived from the same database in the period 2017–19. The study was conducted in a region at moderate risk according to the European SCORE classification.¹⁹

2. Methods

2.1 Data source

We conducted a retrospective cohort analysis using data from COMEGEN, a cooperative of general practitioners (GPs), which includes 140 GPs working in the city of Naples (ASL Napoli 1 Centro), in Southern Italy. All GPs in the network use the same computerized medical record software and cover approximately 5% of the resident population. In addition to diagnoses of diseases (ICD-IX), anthropometric data, and vital parameters, chronic conditions, medical visits, hospitalizations, emergency department accesses, prescription drug dispensations, testing, and vaccinations are recorded.

2.2 Study population

The database used for all analyses included resident subjects aged >18 years. Subjects who decided to change GPs during the study or moved to a different region were excluded from the analysis. Anonymized data from electronic records of 228 266 patients in the period 2020–22 were considered for statistical analysis and included 31 764 subjects with a diagnosis of COVID-19. An equal group of subjects (n = 31 764) recorded in the same database in the period 2017–19 was used for comparisons as an unquestionable COVID-19-free population. Quality control included checking for repeated data, extreme values for quantitative variables, and the correspondence of dates. The cohort of subjects with COVID-19 included patients who had a MACCE on or after January 2020 up to December 2022 (index event). Index CV events in the comparator group were recorded from January 2017 to December 2019. Diagnoses of COVID-19 came from documentation of a positive polymerase chain reaction test in an outpatient laboratory or after a former hospital admission with a diagnosis of COVID-19. Our investigation conforms to the principles outlined in the Declaration of Helsinki.

2.3 The 2017–19 comparator group

The comparator group was composed of adult subjects in the COMEGEN database in the 3 years immediately preceding the outbreak of SARS-Cov-2 pandemic (2017–19). The same inclusion and exclusion criteria applied to the study population (COMEGEN 2020–22) were adopted.

2.4 Study outcomes

The following diagnosis and the corresponding ICD-IX codes were considered in both COMEGEN databases: myocardial infarction (MI) (410), stroke (434 and 436), atrial fibrillation (AF) (420 and 427.31), HF (428), myocarditis (429), and pericarditis (423). The MACCE composite score accounted for these selected diseases. Whenever more than a MACCE occurred in the same individual, the index date of the first incident event was considered. Due to the small numbers recorded, myocarditis and pericarditis were considered as a single diagnosis.

2.5 Study variables

Potential confounders or effect modifiers were extracted from the database of each single GP. The list of comorbidities or clinical conditions that may influence the risk of MACCE considered in the study outcomes included hypertension, hypercholesterolaemia, hypertriglyceridaemia, chronic obstructive pulmonary disease, diabetes, and cancer. Given the relatively low frequency of missing values (<5%), no imputation procedures were considered. The Ethics Board of the ‘ASL Napoli 1 Centro’ approved this study and granted a waiver of informed consent in view of the anonymized collection of data (protocol number 257/22-2023).

2.6 Statistical methods

2.6.1 PS matching

The primary cohort was defined as all patients who had a confirmed diagnosis of COVID-19. We constructed a propensity score (PS)-matched comparator cohort of patients from the COMEGEN database 2017–19. PS was constructed according to sex, age, smoking, and clinical conditions/comorbidity (neoplasms, chronic obstructive pulmonary disease, diabetes hypertension, hypercholesterolaemia, and hypertriglyceridaemia). Cohorts were PS matched (1:1) for 10 covariates:socio-demographic factors and comorbidities. We used a nearest matching approach with a distance of 0.1 pooled SDs of the glm of the PS. Variables with a standardized mean difference between cohorts lower than 0.1 were adequately matched. This approach generated two balanced cohort groups as tested by the standardized mean difference (see Supplementary material online, Table S1). Due to the large size of the control population and to the limited number of matching covariates, the baseline characteristics of the two groups were largely overlapping.²⁰

2.7 Data analysis

Demographic and comorbidity features were compared between the two unmatched databases with the t-test and the Pearson χ² test. The comparison of the same variables in the matched cohorts was done using a paired t-test and the McNemar test, respectively. Risks for a specific outcome in subjects diagnosed with COVID-19 were estimated by comparing disease incidence in this group with the corresponding value in the comparator cohort. Odds ratio (OR) was calculated along with their 95% confidence interval (CI). Cumulative incidence curves were computed from the life table using the Kaplan–Meier approach. The proportional hazard assumption was tested with Schoenfeld residuals. Hazard ratios (HRs) and 95% confidence intervals were estimated by fitting Cox proportional hazard models using a robust variance estimator with clustering for matched pairs. All data were censored at the end of the follow-up, i.e. 31 December 2022 for the COVID-19 patients and 31 December 2019 for the comparator cohort.

To estimate MACCE incidence for each semester of the follow-up, cases in the COVID-19-infected cohort and in the comparator cohort were stratified based on the number of days elapsed after the index date. We reported the risk difference as the difference between the cumulative incidence calculated by the Kaplan–Meier estimator in the SARS-CoV-2 group and the comparator group, multiplied by 100. All analyses were performed using R version 4.2.2 and SPSS 28.0

3. Results

No significant differences of CV risk factors and comorbidities were found between the 2017–19 and 2020–22 COMEGEN (Table 1). Out of the 228 266 individuals included in the COMEGEN database during 2020–22, 31,764 (13.9%) were diagnosed with SARS-CoV-2 infection. Compared with the non-infected population, the infected population was younger, more frequently smoking, diagnosed with cancer, and with CV risk factors (Table 1). In this group, a total number of 1096 MACCE were diagnosed in 752 individuals (2.37%), AF being the most common event, accounting for 545 diagnoses (1.72%).

Table 1

Demographics, life-style and cardiovascular risk factors, comorbidities, and CV outcomes in the COMEGEN databases (2017–19 and 2020–22) and in the group of COVID-19-infected individuals from the 2020–22 COMEGEN database

Variables (no and %)	2017–19 COMEGEN database	2020–22 COMEGEN database	2020–22 COMEGEN database COVID-19 infected
Total N	233.385	228.266	31.764
Sex
Male	107.788 (46.2)	105.403 (46.2)	14.291 (45.0)*
Female	125.597 (53.8)	122.863 (53.8)	17 383 (55.0)
Age (mean ± SD)	51.0 ± 20.7	50.8 ± 20.6	49.3 ± 19.3
≤65 years	170 966 (73.3)	167 870 (73.5)**	25 041 (79.1)*
≥66 years	62 419 (26.7)	60 396 (26.4)	6633 (20.9)
Smoking habit
Current/former	22.816 (9.8)	22.082 (9.7)	3.589 (11.3)*
Comorbidities
Neoplasm	48.382 (20.7)	47.021 (20.6)	8.282 (26.1)*
Chronic obstructive pulmonary disease	7.951 (3.4)	7.598 (3.3)	1.132 (3.6)*
Diabetes	20.672 (8.9)	19.870 (8.7)	2.528 (7.9)
Clinical factors
Hypertension	79.640 (34.1)	77.276 (33.9)	10.863 (34.2)
Hypercholesterolaemia	29.991 (12.9)	29.292 (12.8)	4.605 (14.5)*
Hypertriglyceridaemia	3.205 (1.4)	3.129 (1.4)	538 (1.7)*
CV diseases
Atrial fibrillation	1.713 (0.73)	3.639 (1.59)**	545 (1.72)
Hearth failure	523 (0.22)	1.186 (0.52)**	157 (0.49)
Stroke	383 (0.16)	787 (0.34)**	173 (0.54)*
Myocardial Infarction	380 (0.16)	490 (0.22)**	81 (0.26)
Myopericarditis	115 (0.05)	281 (0.12)**	50 (0.16)
MACCE Composite Endpoint	2863 (1.23)	5616 (2.46)**	752 (2.37)

^*P-values ≤ 0.05 comparing the subgroup of 2020–22 COVID-19 patients with the whole 2020–22 COMEGEN database.

^**P-values ≤ 0.05 comparing the 2017–19 COMEGEN database with the 2020–22 COMEGEN database.

After PS matching, the baseline characteristics of the COVID-19 group did not significantly differ from the comparator group of 31 764 subjects in any of the key variables considered in the PS identification (standardized mean difference < 0.10). Death rates were higher in the PS-matched control group than in COVID-19 subjects (14.4/1000 vs. 9.1/1000).

The proportion of individuals with a new diagnosis of MACCE was higher in the COVID-19 group than in the 2017–19 COMEGEN comparator, with an OR of 1.73 (95% CI: 1.53–1.94; P < 0.001) (Table 2). All MACCE considered showed a significantly higher risk in COVID-19 individuals, ranging from 1.54 to 3.12 for AF and myopericarditis, respectively.

Table 2

Comparison of MACCE incidence and OR in COVID-19 and in PS-matched COVID-19-free controls

CV diseases	2020–22 COMEGEN database COVID-19 infected	2017–2019 COMEGEN database COVID-19 free	Odds ratio	95% Confidence intervals	P-value
Total N (%)	31.764	31.764
Atrial fibrillation	545 (1.72)	297 (0.93)	1.54	1.33–1.78	<0.001
Hearth failure	157 (0.49)	77 (0.24)	2.04	1.56–2.69	<0.001
Stroke	173 (0.54)	52 (0.16)	1.60	1.13–2.26	<0.010
Myocardial infarction	81 (0.26)	35 (0.11)	2.32	1.56–3.45	<0.001
Myopericarditis	50 (0.16)	16 (0.05)	3.12	1.78–5.50	<0.001
MACCE composite endpoint	752 (2.37)	440 (1.39)	1.73	1.53–1.94	<0.001

Figure 1 shows HR for MACCE in the SARS-CoV-2-infected individuals vs. the 2017–19 PS-matched comparator group. All MACCE showed a significantly higher risk in the SARS-CoV-2-infected individuals. The change of MACCE incidence over time for COVID-19-positive and COVID-19-negative matched individuals during the 3 years of follow-up is reported in Figure 2. Differences between the curves are highly significant (P < 0.0001). A similar trend was observed for all MACCE (see Supplementary material online, Figure S1A–E). A quantitative estimate of the MACCE incidence over time in the two study groups is reported in Figure 3. Incidence calculated for each 6-month period after the diagnosis of COVID-19 in our population was the highest in the first two semesters (1.39% and 1.45%, respectively) and was substantially reduced in the following periods, although it remained significantly higher than in the COVID-19-free patients, with values ranging from 0.80% to 0.93%. A similar pattern was found for all MACCE, with more than 80% of the registered diagnoses occurring in the first two semesters. The sensitivity analysis comparing the risk of new diseases in PASC-CVD occurring during the first wave of infection (January 2020 to December 2020) vs. PASC-CVD occurring in the following period (January 2021 to December 2022) did not reveal significant risk modifications (2.25% vs. 2.38%).

Figure 1

HR for selected MACCE in COVID-19-affected individuals from the 2020–22 COMEGEN database (n = 31.764) compared with PS-matched COVID-19-free controls (n = 31.764).

Figure 2

Cumulative incidence of MACCE over time for selected diseases in COVID-19-affected individuals from the 2020–22 COMEGEN database (n = 31.764) compared with PS-matched COVID-19-free controls (n = 31.764).

Figure 3

Incidence (%) of MACCE (composite endpoint) in COVID-19-affected individuals from the 2020–22 COMEGEN database (n = 31.764) compared with PS-matched COVID-19-free controls (n = 31.764) for each 6-month period following COVID-19 diagnosis.

The OR of PASC-CVD calculated with the PS matching did not show any difference between males and females (see Supplementary material online, Table S2). On the contrary, testing for the homogeneity of OR across strata revealed significant differences between smokers and non-smokers, with smokers with COVID-19 experiencing a 3.47 higher risk of MACCE (95% CI: 2.30–5.23) when confronted with the comparator vs. a corresponding OR of 1.60 (95% CI: 1.41–1.81) in COVID-19 non-smoking cases (see Supplementary material online, Table S3). The influence of smoking as an effect modifier was particularly evident for AF and HF. Age also had an influence on the risk of CV in individuals affected by COVID-19. The overall risk of MACCE was higher in the age class < 65 years (OR = 2.36; 95% CI: 1.87–2.98) as compared with the age class over 65 years (OR = 1.56; 95% CI: 1.36–1.80) (see Supplementary material online, Table S4). This result is mostly driven by AF, which showed the most evident difference of risk between age classes (OR = 2.50 vs. OR = 1.35; P < 0.001)

The excess risk attributable to COVID-19 was of 1.2 MACCE cases per 100 subjects. The corresponding figures for specific outcome were 0.9 for AF, 0.3 for HF, 0.2 for stroke, 0.04 for MI, and 0.08 for myopericarditis.

4. Discussion

The results of our study confirm and extend previous observations showing that people with COVID-19 have an increased risk of newly diagnosed MACCE. We showed a persistent excess risk of MACCE in subjects infected by SARS-COV-2 over a 3-year observation period. These results, referring to a real-world setting such as the COMEGEN GPs database, show that the risk of MACCE extends well beyond the acute phase of COVID-19 and even longer of what was reported by Xie et al., thanks to the longer follow-up provided in our study.²¹ Recently, a retrospective cohort study found that the overall 2-year mortality risk was worse among patients infected with COVID-19, but no excess mortality was observed after 180 days. This mortality trend might be possibly explained by a tighter surveillance in subjects affected by acute COVID-19, as well as by the introduction of vaccination and the prevailing spread of less severe variants of the virus.²² In our analysis, overall 3-year death rates were lower in the COVID-19-positive population than in the PS-matched cohort, possibly because of the beneficial effect of a more accurate follow-up in patients affected by COVID-19. Moreover, our data were largely derived from non-hospitalized patients, whereas, in previous studies reporting higher death rates, most data were drawn from hospitalized patients. In addition, the mitigating effects of extensive vaccinations on the severity of COVID might have played a role. Since, in our study, causes of death could not be definitely assessed due to the nature of the data set, we did not further analyse this outcome.

Consistently with our findings, other observational studies have recently reported a significant rise in the incidence of CVD in patients affected by COVID-19, although all of them examined shorter periods in comparison with the 3-year follow-up provided by our study.^23–26 The study of Xie et al. found 23 excess cardiac events per 1000 patients, with HR for composite CV outcomes (HR 1.63, 95% CI: 1.59–1.68) comparable with our study, whereas a higher incidence of PASC-CVD (14% vs. 2,37%) was reported by Daugherty et al.,²⁷ most likely due to different data sources and baseline CV risk profiles.

Our analysis showed no effect modification due to sex in the long-term CV risk of MACCE. This result is in contrast with a recent meta-analysis, which reported a worse prognosis in females,²⁸ whilst consistently with our data, smoking habit was reported to significantly increase the risk of MACCE in COVID-19 patients.²⁸

Consistent with previous reports, our data confirmed arrhythmias as the most common PASC-CV² and were also in agreement with previous evidence on the incidence of HF,²⁹ although the risk of HF was lower.¹⁶ Our results showed that the incidence of myocarditis post-COVID-19 was smaller than that reported in a recent large meta-analysis.¹⁷ This may due to under-reporting of asymptomatic or mildly symptomatic myocarditis in the GPs practice.

The exact mechanisms underlying the association between COVID-19 and the vast range of CV-PACS evidenced in our current analysis as well as in the previous reports cannot be explored in observational studies. Persistent post-viral invasion of cardiac myocytes with consequent cellular necrosis and inflammation affecting endothelial cells can promote the development of myocarditis.³⁰ The persistent activation of inflammatory cascade with cytokine-mediated coagulopathy, microangiopathy, and dysregulation of renin–angiotensin–aldosterone system together with autonomic dysfunction might favour the development of thromboembolism ischaemic events and arrhythmias, respectively.^31,32 Moreover, the elevation of pro-inflammatory cytokines and activation of multiple growth factor signalling pathways could promote fibrosis of cardiac tissue and account for the persistently higher post-COVID-19 CV risk.³³

Some limitations of our study should be acknowledged. Our real-world-based analyses could not include measures of the severity of illness in COVID-19, although this was presumably higher in studies based on patients hospitalized for COVID-19. We also acknowledge that several variants of SARS-CoV-2 were dominant at different times during the study and these variants might be associated with varying clinical outcomes (although a sensitivity analysis did not support this possibility). Finally, at this stage, we did not collect data about SARS-CoV-2 vaccination.

We cannot exclude that other factors might have influenced the retrieved results (i.e. a more difficult access to healthcare systems during the pandemic), and we cannot rule out concomitant infections by other viruses, such as influenza for which subjects are not usually tested in clinical practice.

In conclusion, this study is the first with a long-term follow-up (3 years) in a large general population of patients from a geographical area at moderate CV risk. Our study also suggests that the risk associated with COVID-19 infection for the subsequent development of MACCE is independent of the severity of the COVID-19 infection. In fact, our findings are consistent with previous reports in which the study population was derived from hospital admissions and thus they may have reflected more severe clinical manifestations of COVID-19.

Our findings may have several implications for the daily clinical practice and future research. First, our data are not restricted to the population of patients with severe infection or hospitalized for COVID-19, hence providing a more general estimate of the CV consequences of COVID-19. Secondly, they indicate that the increase of CVD risk associated with COVID-19 might be extended for years and not limited to the acute phase of the infection. Our findings suggest that closer monitoring of CV health after COVID-19 may favour better preparedness to the potential consequences of future pandemics. They might also imply the need for extended follow-up programmes for COVID-19 patients to possibly prevent and promptly manage potential subsequent MACCE and promote earlier recovery. Since COVID-19 is associated with an excess of CVD especially in the first year after the infection, the potential long-term impact of the socio-economic burden of COVID-19 and the policies of the public healthcare systems might need to be reconsidered.

Translational perspective

Our findings demonstrate long-lasting excess of CV disease after SARS-CoV-2 infection and suggest that closer monitoring of CV health after COVID-19 may favour better preparedness to the consequences of future pandemics. They also imply the need for extended follow-up programmes for COVID-19 patients to possibly prevent and promptly manage potential major CV events and promote earlier recovery. Since COVID-19 is associated with an excess of CV diseases especially in the first year after the infection, the potential long-term impact of the socio-economic burden of COVID-19 and the policies of the public healthcare systems might need to be reconsidered.

Supplementary material

Supplementary material is available at Cardiovascular Research online.

Acknowledgements

All those listed as authors are qualified for authorship and all who are qualified to be authors are listed as authors on the byline. All authors have approved the manuscript and agreed with its submission. The present work has been endorsed and promoted by the Italian Society of Cardiovascular Prevention (SIPREC). The authors wish to thank all general practitioners of COMEGEN.

Funding

None declared.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Author notes

Conflict of interest: S.B. and S.P. have received financial support from the Italian Ministry of Health; project of Ministry of University and Research (MUR-PRIN) 2022: grant 2022FJK39Z. Other authors have nothing to disclose.