Cardiometabolic outcomes up to 12 months after COVID-19 infection. A matched cohort study in the UK.

BACKGROUND: Acute Coronavirus Disease 2019 (COVID-19) has been associated with new-onset cardiovascular disease (CVD) and diabetes mellitus (DM), but it is not known whether COVID-19 has long-term impacts on cardiometabolic outcomes. This study aimed to determine whether the incidence of new DM and CVDs are increased over 12 months after COVID-19 compared with matched controls. METHODS AND
FINDINGS: We conducted a cohort study from 2020 to 2021 analysing electronic records for 1,356 United Kingdom family practices with a population of 13.4 million. Participants were 428,650 COVID-19 patients without DM or CVD who were individually matched with 428,650 control patients on age, sex, and family practice and followed up to January 2022. Outcomes were incidence of DM and CVD. A difference-in-difference analysis estimated the net effect of COVID-19 allowing for baseline differences, age, ethnicity, smoking, body mass index (BMI), systolic blood pressure, Charlson score, index month, and matched set. Follow-up time was divided into 4 weeks from index date ("acute COVID-19"), 5 to 12 weeks from index date ("post-acute COVID-19"), and 13 to 52 weeks from index date ("long COVID-19"). Net incidence of DM increased in the first 4 weeks after COVID-19 (adjusted rate ratio, RR 1.81, 95% confidence interval (CI) 1.51 to 2.19) and remained elevated from 5 to 12 weeks (RR 1.27, 1.11 to 1.46) but not from 13 to 52 weeks overall (1.07, 0.99 to 1.16). Acute COVID-19 was associated with net increased CVD incidence (5.82, 4.82 to 7.03) including pulmonary embolism (RR 11.51, 7.07 to 18.73), atrial arrythmias (6.44, 4.17 to 9.96), and venous thromboses (5.43, 3.27 to 9.01). CVD incidence declined from 5 to 12 weeks (RR 1.49, 1.28 to 1.73) and showed a net decrease from 13 to 52 weeks (0.80, 0.73 to 0.88). The analyses were based on health records data and participants' exposure and outcome status might have been misclassified.
CONCLUSIONS: In this study, we found that CVD was increased early after COVID-19 mainly from pulmonary embolism, atrial arrhythmias, and venous thromboses. DM incidence remained elevated for at least 12 weeks following COVID-19 before declining. People without preexisting CVD or DM who suffer from COVID-19 do not appear to have a long-term increase in incidence of these conditions.

Introduction

Coronavirus Disease 2019 (COVID-19) is increasingly recognised as a multisystem condition [1]. Infection of the respiratory tract by the Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) virus triggers host immune responses that may have systemic effects through activation of inflammatory pathways [2,3]. COVID-19 may trigger a proinflammatory “cytokine storm” with dysregulated immune response, platelet activation, hypercoagulability, endothelial cell dysfunction, and thromboembolism affecting diverse systems with potential for end-organ damage [4]. While acute COVID-19 infection has been associated with new-onset cardiovascular disease (CVD) and diabetes mellitus (DM) [5], longer-term outcomes up to 1 year following the infection have not been well characterised. Cardiac manifestations of COVID-19 [6,7] include cardiac injury with elevated troponin levels, heart failure, and increased risk of mortality among patients hospitalised with COVID-19 [8,9]. COVID-19 may also be associated with acute myocardial infarction and ischaemic stroke in the first 4 weeks [10]. New-onset hyperglycaemia has been reported in COVID-19 patients, sometimes representing “stress hyperglycaemia,” and is associated with worse prognosis [5,11]. Complications of both preexisting and new-onset DM have been observed, including diabetic ketoacidosis and hyperosmolarity [12-14]. Possible mechanisms appear to be direct pancreatic damage from SARS-CoV-2 and/or the associated systemic inflammatory syndrome with severe COVID-19, as reflected in high levels of interleukin-6 (IL-6) and tumour necrosis factor alpha (TNFα), causing impaired pancreatic insulin secretion [15] and insulin resistance [16].

Since the first evidence of COVID-19 emerged in the beginning of 2020, there have been multiple waves of infection and multiple variants of SARS-CoV-2 have been identified. There were in excess of 1,000 deaths per day in the United Kingdom at the peak of the first wave in April 2020 and the peak of the second wave in January 2021 with many more suffering illness of varying severity [17]. The longer-term outcomes of COVID-19 are now receiving increasing attention. A high proportion of patients report experiencing symptoms for more than 4 weeks after initial presentation with COVID-19 [18]. The distinction has been made between “acute COVID-19” in the first 4 weeks after infection; “post-acute COVID-19” (or “ongoing symptomatic COVID-19”), from 5 to 12 weeks after the first infection; and “long COVID-19” (or “post-COVID-19 syndrome”) with symptoms persisting for more than 12 weeks after infection [19]. There is concern that cardiovascular and metabolic outcomes may be compromised in the aftermath of COVID-19 infection. However, susceptibility to COVID-19 and the severity of illness are known to be associated with cardiometabolic risk. Furthermore, public health restrictions during the height of the pandemic, including “lockdowns” or “stay-at-home” orders, were associated with profound changes in diet, exercise habits, and other health-related behaviours that might have impacted on CVD and diabetes in the general population even in the absence of COVID-19 infection. Controlled studies are therefore needed to evaluate the net long-term impacts of COVID-19 infection on cardiovascular and diabetes outcomes after allowing for premorbid differences between cases and controls and changes over time in control participants. There is concern for the possible burden of “long COVID-19” syndromes in patients recovering from COVID-19, but few studies have reported on long-term follow-up for large population-based samples. Al-Aly and colleagues [5] employed a data mining approach to Veterans’ Administration data and identified an increased burden of diverse health conditions between 30 days and 6 months after COVID-19. Knight and colleagues [20] suggested that arterial and venous complications remain elevated for 49 weeks after COVID-19. However, the timeline for recovery from COVID-19 remains poorly characterised.

Longitudinal datasets from electronic health records offer opportunities to analyse longer-term COVID-19 outcomes. We employed the Clinical Practice Research Datalink (CPRD), a database of anonymised primary care electronic health records, to identify a cohort of COVID-19 exposed patients in comparison with a matched cohort with no COVID-19 diagnosis. We aimed to estimate the net effect of COVID-19 on cardiometabolic outcomes over periods of 4 weeks, 3 months, and 12 months in order to inform research priorities, clinical services, and public health interventions that may be required following acute COVID-19 infection.

Methods

Ethical approval

The protocol was given scientific and ethical approval by the CPRD Independent Scientific Advisory Committee (ISAC protocol 20_00265R). CPRD holds over-arching research ethics committee approval for the conduct of research using fully anonymised electronic health records data from the CPRD databases. The study protocol has been published in summary form by the CPRD/MHRA and can be accessed here. The full protocol, including the prospective analysis plan, is included as S1 Protocol. This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist).

Data source and participant selection

We conducted a population-based, matched cohort study in CPRD Aurum. The CPRD Aurum is a large database that includes comprehensive medical record data for a current total of 1,356 family practices in England with approximately 13.4 million currently registered patients in the March 2022 release [21]. The database includes data for all registered patients at participating general practices, except for a negligible number of patients who opt out of data collection. Each patient has a unique anonymised numerical identifier that remains the same at each update of the database. Patients may therefore be tracked through successive releases of the database which is updated monthly. The March 2022 release data, which cover approximately 20% of the population of England, are considered to be generally representative of the general population with regards to geographical distribution, deprivation, age, and sex [21]. COVID-19 test results are automatically transmitted from laboratories to patients’ family practices. Coded records for CVD or DM derive from general practice consultations, hospital outpatient referrals, and inpatient admissions.

At the inception of the study, data for all patients with a diagnosis of COVID-19 were extracted from the February 2021 release of CPRD Aurum. The date of the first code for COVID-19 was the index date. Since confirmatory testing was not widely available during the initial phase of the pandemic, we included patients with clinical diagnoses of “confirmed” or “suspected” COVID-19 (S1 Table). However, we conducted a sensitivity analysis including only patients with a medical code recorded for polymerase chain reaction (PCR) test confirmed COVID-19. The COVID-19 cohort was compared with a sample of matched control patients who were not recorded with a COVID-19 diagnosis up to the case index date. Control patients were randomly sampled from the registered population of CPRD Aurum March 2021 release that provided the most up-to-date information on the database at the time of sampling. Controls were matched for year of birth, sex, and family practice, and their record was required to start no later than 18 months after the start of record for their matched case. Patients were excluded from eligibility as controls if their record ended before the matched case index date, or if they had prevalent CVD or DM recorded more than 12 months before the index date or within 1 year of the start of their record. Using the unique identifiers for patients in the sample, data were updated to the March 2022 release of CPRD Aurum for final analyses, providing follow-up to 31 January 2022. The records of any control patients diagnosed with COVID-19 after the index date were censored 30 days before the control COVID-19 diagnosis date (S1 Text).

Outcome measures

The study outcomes were first ever recorded CVD and DM diagnoses (S1 Table). CVD diagnoses were grouped into the categories: myocardial infarction and ischaemic heart disease; atrial arrhythmias, including atrial fibrillation and supraventricular tachycardia; heart failure; cardiomyopathy and myocarditis; pulmonary embolism; venous thrombosis; and stroke. DM diagnoses included diagnoses for type 1 and type 2 DM and initiation of oral hypoglycaemic drugs and insulin. Records of HbA1c were evaluated, and a second record of HbA1c ≥48 mmol/mol was considered diagnostic of diabetes. Patients were considered to have a type 1 DM phenotype if they were aged 35 years or less at diagnosis and were prescribed insulin within 91 days of diagnosis [22]. Mortality was evaluated from the CPRD date of death.

Covariates

Covariates were defined using data recorded in the study period before the index date. Covariates selected because of known associations with CVD and DM included smoking status (nonsmoker, current smoker, ex-smoker), body mass index (BMI, underweight <18.5 kg/m2, normal weight 18.5 to <25.0 kg/m2, overweight 25.0 to <30.0 kg/m2, obese ≥30 kg/m2, and not recorded), systolic blood pressure (SBP), and diastolic blood pressure (DBP) in categories of 10 mm Hg. Ethnicity was classified as “white,” “black,” “Asian” (of Indian subcontinent origins), “mixed,” “other,” and “not known.” We also employed the Charlson index to summarize comorbidities that may be associated with the severity of illness in COVID-19 [23]. Each component morbidity was classified as present or absent before the index date; the Charlson score was calculated with a maximum of 4 or more [24,25]. COVID-19 and control patients were matched on sex, year of birth, and family practice, with the latter offering control for area measures including level of deprivation. Prescriptions for glucocorticoids were evaluated in patients diagnosed with DM because this therapy for severe COVID-19 is associated with diabetes risk [26].

Analysis

The prospective plan envisaged analysis in a time-to-event framework, with the date of COVID-19 as the start date (S1 Protocol). Recognition of the importance of comparing risks from before the index date and adjusting for baseline values in a difference-in-difference analysis prompted us to employ Poisson models to evaluate incidence of DM and CVD in periods before and after the COVID-19 diagnosis. However, time-to-event and log-linear models are generally equivalent and give similar results [27].

Person-time at risk was evaluated in terms of person-weeks of follow-up, and incident events of CVD and DM were identified in each 4-week period. Follow-up time was divided into periods of 28 days, from 1 year before the index date to 1 year after the index date. Data were then aggregated for the periods before the index date (“Pre-Index”) and for 4 weeks after the index date (“acute COVID-19”), 5 to 12 weeks after the index date (“Post-acute COVID-19”) and from 13 to 52 weeks after the index date (“Long COVID-19”). Incidence rates, with exact Poisson confidence intervals (CIs), were estimated for COVID-19 patients and controls. Poisson regression models were fitted using the method of generalised estimating equations, employing an exchangeable correlation structure with robust standard error estimates (S2 Text). This approach allowed for clustering on the matched set of COVID-19 participant and control, as well as allowing for possible overdispersion. The effect of group (COVID-19 or control), time (pre-index, 4 weeks from index, 5 to 12 weeks, and 13 to 52 weeks), and the group-time interaction were estimated. The group-time interaction represented the net additional effect of acute COVID-19, post-acute COVID-19, or long COVID-19 in COVID-19 participants, net of the baseline values for COVID-19 participants, and the comparison with control participants. Analyses were adjusted for age, age-squared, sex, ethnicity, BMI, SBP, Charlson score, index month, and index month squared (S2 Text). Missing values were represented by indicator variables (Table 1); we compared unadjusted and covariate adjusted estimates. We recognised that incidence of CVD and DM might change during the period from 13 to 52 weeks after COVID-19 diagnosis. Therefore, in secondary analyses, we estimated adjusted rate ratios and their CIs for the comparison of each 4-week period following COVID-19 diagnosis with baseline, using the same analytical approach. Estimates were plotted with a smoothed curve fitted using the loess method. We conducted a sensitivity analysis including only COVID-19 participants with a positive PCR test for SARS-CoV-19 infection. A logistic regression model was fitted to evaluate variables associated with PCR confirmation. All analyses were implemented in the R program, version 3.6.1. Following peer review, we included a sensitivity analysis to evaluate whether adjusting for consultation frequency might explain association of COVID-19 with diabetes incidence.

Table 1

Characteristics of COVID-19 and matched control participants.

Figures are frequencies (percent of column total).

		Matched controls (428,650)	COVID-19 patients (428,650)
Sex	Male	190,401 (44)	190,401 (44)
	Female	238,249 (56)	238,249 (56)
Age (median, IQR years)		35 (22 to 50)	35 (22 to 50)
Ethnicity	“Asian”	25,978 (6)	38,164 (9)
	“Black African/Caribbean”	15,747 (4)	15,783 (4)
	“Mixed”	7,285 (2)	9,278 (2)
	Not recorded	130,560 (30)	83,850 (20)
	“Other”	14,760 (3)	17,205 (4)
	“White”	234,320 (55)	264,370 62)
Cigarette smoking	Nonsmoker	201,549 (47)	226,339 (53)
	Ex-smoker	51,333 (12)	64,320 (15)
	Current smoker	79,285 (18)	67,979 (16)
	Not recorded	96,483 (23)	70,012 (16)
BMI	Underweight	14,596 (3)	16,432 (4)
	Normal weight	89,903 (21)	102,241 (24)
	Overweight	66,400 (15)	84,393 (20)
	Obese	51,695 (12)	70,854 (17)
	Not recorded	206,056 (48)	154,730 (36)
SBP	<110	45,989 (11)	52,133 (12)
	110–119	67,188 (16)	74,917 (17)
	120–129	82,021 (19)	91,718 (21)
	130–139	64,381 (15)	70035 (16)
	140–149	28,642 (7)	29,874 (7)
	150–159	8,058 (2)	8,503 (2)
	≥160	5,067 (1)	5,307 (1)
	Not recorded	127,304 (30)	96,163 (22)
Charlson score	0	341,996 (80)	330,139 (77)
	1	70,581 (16)	80,337 (19)
	2	11,438 (3)	12,127 (3)
	3	3,197 (1)	4,141 (1)
	≥4	1,438 (0)	1,906 (0)

BMI, body mass index; COVID-19, Coronavirus Disease 2019; IQR, interquartile range; SBP, systolic blood pressure.

Results

There were 516,985 participants diagnosed with COVID-19. COVID-19 participants with prevalent CVD or DM, diagnosed within 1 year of the start of record or more than 1 year before the index date, were excluded leaving 431,193 COVID-19 participants. These were matched with potential controls and after excluding participants with indeterminate sex, age more than 104 years, prevalent CVD or DM recorded in their full medical record, there were 428,650 (99%) matched sets, with a median duration of follow-up of 12 months, for further analysis (Table 1 and S1 Text). COVID-19 participants included 347,011 (80.9%) with clinical or laboratory confirmation of diagnosis on the index date and 281,486 (65.7%) with a positive PCR test for SARS-CoV-2 on the index date (S3 Text). COVID-19 participants and controls were similar with respect to matching variables of age and sex, there was a slight female predominance with median age of 35 years (Table 1). COVID-19 participants were more likely to have defined values recorded for covariates. COVID-19 participants included slightly more of “Asian” ethnicity, fewer current smokers, more who were overweight and obese, and more with comorbidity than controls, while blood pressure distributions appeared to be similar. COVID-19 participants had a median of 11 consultations (interquartile range 6 to 21) after the index date, compared with 7 (3 to 15) in controls.

Fig 1 shows the incidence of DM and CVD from 52 weeks before the index date to 52 weeks after for COVID-19 participants (red) and controls (blue). In the period before the index date, the incidence of diabetes was slightly higher in patients who went on to be diagnosed with COVID-19 than control participants; a similar difference was also observed for the incidence of CVDs. In the first 4 weeks after COVID-19 diagnosis, there was a sharp increase in the incidence of diabetes and an even greater increase in CVDs. In the remainder of the first year following COVID-19 diagnosis, the incidence of DM for COVID-19 cases appeared to remain higher than for controls, while CVD incidence appeared to decline to premorbid values.

Fig 1

Incidence rates for DM and CVDs (per 100,000 patient weeks) for COVID-19 patients (red) and controls (blue) for 4-week periods.

Bars are 95% CIs. CI, confidence interval; COVID-19, Coronavirus Disease 2019; CVD, cardiovascular disease; DM, diabetes mellitus.

Table 2 shows the incidence of DM and CVD aggregated into 4 periods: before the index date, within 4 weeks of the index date, from 5 to 12 weeks after the index date, and from 13 to 52 weeks after the index date. In the period before the index date, the incidence of diabetes was 15.81 (95% CI 15.29 to 16.34) per 100,000 patient weeks in patients who were later diagnosed with COVID-19 but 11.32 (10.88 to 11.77) in controls who did not develop COVID-19; the incidence of CVD was 14.07 (13.58 to 14.58) per 100,000 patient weeks for cases and 7.58 (7.22 to 7.95) for controls, respectively. During the first 4 weeks from COVID-19 diagnosis, the incidence of DM increased to 23.79 (21.57 to 26.18) and the incidence of CVD increased to 76.92 (72.89 to 81.13) per 100,000 patient weeks. In the periods from 5 to 12 weeks and 13 to 52 weeks, the incidence of CVD declined to 22.06 (20.53 to 23.68) and 12.63 (12.10 to 13.18) per 100,000 patient weeks, respectively. Thus, in the period from 13 to 52 weeks after COVID-19 diagnosis, the overall incidence of CVD declined to premorbid levels or below. During the same periods, the incidence of DM appeared to remain elevated at 19.54 (18.10 to 21.06) per 100,000 in the post-acute period and 19.57 (18.91 to 20.26) per 100,000 in the long-COVID period.

Table 2

DM and CVD events and rates by period for COVID-19 cases and controls.

Figures are frequencies except where indicated.

Phase	COVID-19 or control	Patient weeks	CVD events	CVD incidence per 100,000 patient weeks (95% CI)	Diabetes diagnoses	DM incidence per 100,000 patient weeks (95% CI)
Before index date	COVID-19	21,894,712	3,081	14.07 (13.58 to 14.58)	3,461	15.81 (15.29 to 16.34)
	Controls	22,462,441	1,702	7.58 (7.22 to 7.95)	2,542	11.32 (10.88 to 11.77)
Acute: up to 4 weeks from index	COVID-19	1,765,392	1,358	76.92 (72.89 to 81.13)	420	23.79 (21.57 to 26.18)
	Controls	1,750,425	128	7.31 (6.10 to 8.69)	166	9.48 (8.10 to 11.04)
Post-acute: 5 to 12 weeks from index	COVID-19	3,485,790	769	22.06 (20.53 to 23.68)	681	19.54 (18.10 to 21.06)
	Controls	3,461,014	291	8.41 (7.47 to 9.43)	384	11.10 (10.01 to 12.26)
Long: 13 to 52 weeks from index	COVID-19	16,635,214	2,101	12.63 (12.10 to 13.18)	3,256	19.57 (18.91 to 20.26)
	Controls	16,351,115	1,487	9.09 (8.64 to 9.57)	2,153	13.17 (12.62 to 13.74)

CI, confidence interval; COVID-19, Coronavirus Disease 2019; CVD, cardiovascular disease; DM, diabetes mellitus.

Fig 2 shows the incidence of categories of CVD from 52 weeks before the COVID-19 index date to 52 weeks after. Pulmonary embolism showed the greatest increase during the acute COVID-19 period, with 34.67 cases per 100,000 patient weeks. The value was higher than for atrial arrhythmias (13.93 per 100,000 per week), venous thrombosis (8.84 per 100,000 per week), myocardial infarction and ischaemic heart disease (6.57 per 100,000 per week), stroke (6.68 per 100,000 per week), heart failure (3.80 per 100,000 per week), or cardiomyopathy and myocarditis (2.21 per 100,000 per week). While the relative effects differed for categories of CVD, the time course of changes in incidence was similar for each category of CVD.

Fig 2

Incidence rates for categories of CVD (per 100,000 patient weeks) for COVID-19 patients by 4-week periods.

COVID-19, Coronavirus Disease 2019; CVD, cardiovascular disease; IHD, ischaemic heart disease; MI, myocardial infarction.

Table 3 presents estimated incidence rate ratios both unadjusted and adjusted for age, sex, ethnicity, smoking, BMI, SBP, and Charlson comorbidity score. After allowing for underlying differences between cases and controls, and between the baseline pre-index period and follow-up, the net effect of acute COVID-19 on CVD incidence was estimated to be an adjusted rate ratio of 5.82 (95% CI 4.82 to 7.03), while for DM incidence the adjusted rate ratio was 1.81 (1.51 to 2.19). In the period from 5 to 12 weeks after diagnosis, the adjusted incidence rate ratio for CVD declined to 1.49 (1.28 to 1.73), while for DM, this was 1.27 (1.11 to 1.46) (Table 3 and S4 Text). In the period from 13 to 52 weeks after diagnosis, the adjusted incidence rate ratio for CVD was 0.80 (0.73 to 0.88), while for DM, this was 1.07 (0.99 to 1.16). Covariate adjustment resulted in little change in the estimated incident rate ratios (Table 3), with the difference-in-difference analysis already allowing for baseline differences in risk.

Table 3

Results of difference-in-difference analysis showing net effect of COVID-19 over baseline rate and comparison with controls.

(Estimates were adjusted for age, ethnicity, smoking, BMI category, SBP category, Charlson score, index month, and matched set).

		“Acute COVID-19” up to 4 weeks		“Post-acute COVID-19” 5 to 12 weeks		“Long COVID-19” 13 to 52 weeks
		RR (95% CI)	P value	RR (95% CI)	P value	RR (95% CI)	P value
DM	Unadjusted	1.80 (1.49 to 2.17)	<0.001	1.26 (1.10 to 1.45)	<0.001	1.07 (0.99 to 1.15)	0.09
	Adjusted	1.81 (1.51 to 2.19)	<0.001	1.27 (1.11 to 1.46)	<0.001	1.07 (0.99 to 1.16)	0.07
All CVD outcomes	Unadjusted	5.68 (4.70 to 6.86)	<0.001	1.42 (1.22 to 1.65)	<0.001	0.76 (0.69 to 0.83)	<0.001
	Adjusted	5.82 (4.82 to 7.03)	<0.001	1.49 (1.28 to 1.73)	<0.001	0.80 (0.73 to 0.88)	<0.001
Subgroups of CVD. Adjusted estimates
Pulmonary embolism		11.51 (7.07 to 18.73)	<0.001	2.29 (1.47 to 3.57)	<0.001	0.66 (0.50 to 0.87)	0.003
Atrial arrhythmias		6.44 (4.17 to 9.96)	<0.001	1.58 (1.10 to 2.27)	0.01	0.85 (0.68 to 1.05)	0.13
Venous thrombosis		5.43 (3.27 to 9.01)	<0.001	1.84 (1.27 to 2.64)	0.001	0.88 (0.71 to 1.10)	0.26
Heart failure		5.23 (2.04 to 13.44)	<0.001	2.30 (1.17 to 4.51)	0.02	0.72 (0.49 to 1.06)	0.10
Stroke		3.31 (2.05 to 5.35)	<0.001	0.87 (0.63 to 1.21)	0.41	0.73 (0.59 to 0.89)	0.002
Cardiomyopathy and myocarditis		2.96 (1.27 to 6.88)	0.01	1.37 (0.70 to 2.68)	0.35	1.11 (0.74 to 1.66)	0.62
Myocardial infarction and IHD		2.01 (1.34 to 3.00)	<0.001	0.97 (0.68 to 1.38)	0.86	0.80 (0.66 to 0.98)	0.03

BMI, body mass index; CI, confidence interval; COVID-19, Coronavirus Disease 2019; CVD, cardiovascular disease; DM, diabetes mellitus; IHD, ischaemic heart disease; RR, adjusted incidence rate ratio; SBP, systolic blood pressure.

When categories of CVD were analysed separately (Table 3), pulmonary embolism (RR 11.51, 7.07 to 18.73), atrial arrythmias (6.44, 4.17 to 9.96), and venous thromboses (5.43, 3.27 to 9.01) showed the greatest relative increases in acute COVID-19 but myocardial infarction (2.01, 1.34 to 3.00), heart failure (5.23, 2.04 to 13.44), and stroke (3.31, 2.05 to 5.35) were also increased. These conditions declined substantially in the period from 5 to 12 weeks after COVID-19 diagnosis. There was no evidence that any of these categories of CVD showed a net increase during the period from 13 to 52 weeks after COVID-19 diagnosis, with evidence of lower incidence for several, including pulmonary embolism and stroke.

When adjusted rate ratios were estimated by 4-week period (Fig 3 and S5 Text), there was no evidence that CVD events might be increased beyond 8 to 11 weeks after COVID-19 onset (RR 1.15, 95% CI 0.94 to 1.42, P = 0.18). There was evidence of a net reduction in CVD events, compared with baseline and controls, from weeks 20 to 23 onwards. There was evidence that DM incidence might be increased up to 20 to 23 weeks after COVID-19 onset (1.29, 1.08 to 1.55, P = 0.005) but not from 24 to 27 weeks onwards (1.03, 0.86 to 1.22, P = 0.78).

Fig 3

Adjusted rate ratios by 4-week period following COVID-19.

Points represent adjusted rate ratios, bands represent 95% CIs, line represents smoothed curve. (Estimates were adjusted for age, ethnicity, smoking, BMI category, SBP category, Charlson score, index month, and matched set.) BMI, body mass index; CI, confidence interval; COVID-19, Coronavirus Disease 2019; SBP, systolic blood pressure.

The median age at diagnosis of DM was 47 (interquartile range 34 to 57) years for COVID-19 cases and 44 (33 to 57) years for controls; for CVD, the median age at diagnosis was 60 (49 to 73) for COVID-19 cases and 60 (49 to 72) for controls (S6 Text). The distribution of sex, ethnicity, smoking, and obesity were generally consistent across stages of COVID-19 infection. Patients diagnosed with diabetes in the context of acute COVID-19 were more likely to be prescribed insulin with 91 days of diagnosis (52/420, 12%, compared with 11/166, 7%, for controls) but there was no clear evidence of an increase in type 1 DM at any stage of the illness (S6 Text). There were 7,724 (1.8%) control patients and 18,895 (4.4%) COVID-19 patients who received one or more primary care prescriptions for oral, enteral, intramuscular, or intravenous glucocorticoids from 4 weeks before the index date to the end of follow-up. There were 126 new diagnoses of diabetes in glucocorticoid-treated controls and 417 in glucocorticoid-treated cases during the same period.

Sensitivity analysis

There were 281,486 (65.7%) COVID-19 participants with a positive PCR test recorded on the index date. Compared with the entire sample, patients with PCR test confirmation of COVID-19 tended to be younger, were more often white, male, nonobese, and nonsmokers. The odds of PCR confirmation decreased by 1.6% per year increase in age. Results of the sensitivity analysis are shown in S7 Text. Although rates of CVD and DM were lower in the sample with PCR confirmed COVID-19 infection, consistent with their generally lower cardiometabolic risk profile, differences between COVID-19 cases and controls, and changes over time following COVID-19 infection, were consistent with those observed in the whole sample. In the analysis for DM, adjustment for consultation frequency had negligible effect on the first 2 decimal places of regression coefficients and their standard errors (S8 Text).

Discussion

Main findings

This study compared COVID-19 patients with matched population controls from 1 year before a COVID-19 diagnosis to 1 year after. The study showed that patients recorded with COVID-19 in primary care are at slightly greater risk of CVD and DM even before the COVID-19 diagnosis. While patients with preexisting DM or CVD may be at greater risk of symptomatic COVID-19, we excluded all patients with DM or CVD as preexisting conditions at study entry. Patients who had already died from CVD or DM were also necessarily excluded. Elevated cardiometabolic risk was associated with higher BMI, greater comorbidity, and a higher proportion of “Asian” ethnicity. After allowing for baseline differences between COVID-19 patients and controls, new DM diagnoses increased 1.8-fold during acute COVID-19 and continued to show evidence of increase for at least 12 weeks before declining. There was no evidence overall that COVID-19 was associated with any additional increase in DM and CVD events from 13 to 52 weeks, above that observed in controls. However, secondary analysis by 4-week period provided evidence that DM events might not return to baseline until 20 to 23 weeks after COVID-19. There was evidence of increased glucocorticoid exposure in COVID-19 patients, but this might be associated with only a very small proportion of new DM events. New CVD events were increased nearly 6-fold during acute COVID-19 and nearly 50% in post-acute COVID-19. This increase was largely driven by pulmonary embolism diagnoses, but atrial arrythmias, venous thrombosis, myocardial infarction, stroke, and heart failure also showed more modest increases. Cardiovascular conditions are often associated with acute presentations that may lead to early diagnosis; this is in contrast to diabetes which may remain undiagnosed for a variable period of time, possibly contributing to the more delayed decline in the latter condition with consultation rates changing over time [28].

Comparison with other studies

The long-term outcomes of COVID-19 infection represent an emerging area of research and public concern, but definitions are not universally agreed. This study drew on clinical recommendations that distinguish “acute” (4 weeks), “post-acute” (3 months), and “long” (1 year) periods after COVID-19 infection [19]. “Long COVID” is often self-identified by patients with the concept focusing on symptom burden and impacts on health-related quality of life [19,29]. These subjective outcomes of COVID-19 infection may be heterogenous in nature and their occurrence and persistence may depend on patient characteristics including age, sex, and comorbidity, as well as the severity of the initial COVID-19 illness and intensity and duration of therapeutic support in the acute illness [29,30]. It is well established that preexisting hypertension and ischaemic heart disease are associated with more severe illness and greater mortality in acute COVID-19 infection [31]. Hospitalised patients may experience a range of cardiovascular complications including arrhythmias, heart failure, and thrombotic complications [31], but there are few studies with long-term follow-up of patients without preexisting CVD. In a preprint, Knight and colleagues [20] reported on cardiovascular outcomes of a large population in England. Their analyses suggested that CVD outcomes might remain increased for up to 49 weeks following COVID-19 infection [20]. However, the analyses did not include a baseline period before the COVID-19 infection that might have provided more precise evaluation of baseline differences in risk between COVID-19 cases and controls.

Preexisting DM is also associated with more severe illness from COVID-19 [32] but some studies suggest that COVID-19 may be associated with new-onset diabetes. A systematic review of 8 reports of patients hospitalised in the early stages of the COVID-19 pandemic found that 14.4% of patients developed new-onset diabetes following the illness [33]. A possible effect of SARS-CoV-2 infection on pancreatic function is suggested by the finding that the virus infects pancreatic beta cells [34], reduces insulin production, and promotes beta-cell apoptosis [35]. COVID-19 may also lead to reduced physical activity and deconditioning [36] leading to greater insulin resistance. Contacts with medical care may also lead to increased opportunities to detect previously undiagnosed diabetes. Previous studies have often reported hospital-based cohorts, with smaller samples or shorter durations of follow-up. This large population-based study shows that patients diagnosed with COVID-19 have a slightly higher baseline risk of DM. There is evidence that new-onset diabetes is increased during the acute COVID-19 illness, and there is evidence that a net increase in diabetes persists for at least 12 weeks following the COVID-19 illness before declining. There was evidence that patients diagnosed during the acute illness might be more likely to be prescribed insulin but there was no evidence for increased type I diabetes overall, nor was there evidence that glucocorticoid exposure might account for a high proportion of cases. While this study took a population perspective, clinical research is needed to determine whether subgroups of patients at greater risk of hyperglycaemia, including those with pre-diabetes, pregnancy, or polycystic ovary syndrome, are vulnerable to post-COVID-19 DM.

Strengths and limitations

This study drew on a large, longitudinal population-based data resource that enabled us to conduct a matched analysis of mortality and new CVD and DM diagnoses for up to 1 year following COVID-19 [28]. The study drew on clinical records with several limitations. COVID-19 participants might predominantly include symptomatic cases as well as both symptomatic and asymptomatic contacts of cases who gave positive test results. We included patients with both confirmed and suspected COVID-19, but a sensitivity analysis found that restricting the analysis to PCR confirmed infections would not alter conclusions. PCR testing was associated with patient characteristics and reliance on PCR confirmation for participant selection might lead to bias. There might be a risk that the control group might be contaminated by undiagnosed or asymptomatic COVID-19 infections, this might be particularly so for the early stage of the pandemic when testing was not widely available. The Office for National Statistics Coronavirus (COVID-19) Infection Survey showed that at the height of the first wave of infection from 27 April and 10 May 2020, an average of 0.27% (95% CI: 0.17% to 0.41%) of the general population had COVID-19 [37]. From October 2020 to October 2021, fewer than 2% of the population was infected with COVID-19 in any given week. Thus, the probability of a control participant having undiagnosed COVID-19 around the index date of a matched case is less than 0.02. The database enabled adjustment for a range of important covariates, but these were not always completely recorded. In health records, values are commonly missing “not at random” making the application of imputation methods more difficult. Covariate values may change over time without being recorded at family practices. Estimates without covariate adjustment were of very similar magnitude to adjusted estimates, suggesting that there was no appreciable bias from missing or misclassified covariate values. We did not include a measure of deprivation, which is associated both with diabetes and CVD, but cases and controls were matched for family practice, providing partial control for area-based measures including deprivation. Analyses did not include measures of the severity of illness in COVID-19. However, the concept of “severity” might be difficult to operationalise in a study of COVID-19 complications because a greater number of complications might be indicative of more severe illness. We found that COVID-19 participants had higher rates of family practice consultations, consistent with their recent illness. This increased medical surveillance could be associated with more frequent opportunities for diabetes to be diagnosed. However, illness from diabetes could itself lead to increased consultations. We did not have access to data for secondary care prescribing, consequently exposure to glucocorticoids might be underestimated. We also did not have sufficient data for alcohol use, diet, physical activity, and lipid parameters that might have contributed as confounders of observed associations. Patients diagnosed with COVID-19 were generally less healthy than controls and this might, in part, account for differences in cardiometabolic outcomes. We employed a difference-in-difference analysis that allowed for baseline differences in risk, but there might have been residual confounding if there was misclassification of risk status from incomplete data recording. The observational nature of the study limits causal inferences concerning whether the increased risk of new CVD and DM diagnoses may result from COVID-19, or whether undiagnosed CVD and DM were more prevalent among COVID-19 cases, or whether COVID-19 aggravated or altered the natural history of preexisting disease. We also acknowledge that several variants of SARS-CoV-2 were dominant at different times during the course of the study (the original wild-type virus was superseded by the Alpha variant from December 2020 and the Delta variant from June 2021 [38]), and these variants might be associated with varying biomedical and clinical outcomes. The study employed matching but we also adjusted for matching variables in order to avoid bias [39].

Conclusions

This study provides evidence that DM incidence remains elevated for at least 12 weeks following COVID-19 infection before declining. Advice to patients recovering from COVID-19 should include measures to reduce diabetes risk, including diet, weight management, and physical activity levels, especially in view of heightened baseline risk. CVD is increased early after COVID-19 infection, mainly from pulmonary embolism, atrial arrhythmias, and venous thromboses, and these risks are increased for up to 3 months. However, people without preexisting CVD or DM who suffer from COVID-19 do not appear to have a long-term increase in incidence of these conditions.

The views expressed are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health.

STROBE checklist.

(DOC)

Click here for additional data file.

Study protocol.

(PDF)

Click here for additional data file.

Medical and product codes.

(XLSX)

Click here for additional data file.

Flowchart showing participant selection.

(DOCX)

Click here for additional data file.

Code for regression model.

(DOCX)

Click here for additional data file.

Medical codes associated with COVID-19 index date in 428,650 COVID-19 cases.

(DOCX)

Click here for additional data file.

Results of Poisson regression models showing main effects of group and time period and group by time interaction.

CI, confidence interval; RR, adjusted incidence rate ratio. (Estimates were adjusted for age, ethnicity, smoking, BMI category, SBP category, Charlson score, index month and matched set.)

(DOCX)

Click here for additional data file.

Data for Fig 3.

Adjusted rate ratios (95% confidence intervals) by 4-week period following COVID-19. (Estimates were adjusted for age, ethnicity, smoking, body mass index category, systolic blood pressure category, Charlson score, index month, and matched set.)

(DOCX)

Click here for additional data file.

Characteristics of case and control patients diagnosed with DM or CVD during follow-up.

(DOCX)

Click here for additional data file.

Results of a sensitivity analysis on 243,716 COVID-19 cases confirmed by polymerase chain reaction (PCR) test and matched controls.

(DOCX)

Click here for additional data file.

Sensitivity analysis for incident diabetes mellitus after COVID-19 adjusting for consultation frequency.

Figures are coefficients (standard errors).

(DOCX)

Click here for additional data file.

18 Jan 2022

Dear Dr Gulliford,

Thank you for submitting your manuscript entitled "Differential long-term impact of Covid-19 on incidence of diabetes mellitus and cardiovascular diseases: population-based cohort study" for consideration by PLOS Medicine.

Your manuscript has now been evaluated by the PLOS Medicine editorial staff and I am writing to let you know that we would like to send your submission out for external peer review.

However, before we can send your manuscript to reviewers, we need you to complete your submission by providing the metadata that is required for full assessment. To this end, please login to Editorial Manager where you will find the paper in the 'Submissions Needing Revisions' folder on your homepage. Please click 'Revise Submission' from the Action Links and complete all additional questions in the submission questionnaire.

Please re-submit your manuscript within two working days, i.e. by Jan 20 2022 11:59PM.

Login to Editorial Manager here: https://www.editorialmanager.com/pmedicine

Once your full submission is complete, your paper will undergo a series of checks in preparation for peer review. Once your manuscript has passed all checks it will be sent out for review.

Feel free to email us at plosmedicine@plos.org if you have any queries relating to your submission.

Kind regards,

Callam Davidson

Associate Editor

PLOS Medicine

17 Feb 2022

Dear Dr. Gulliford,

Thank you very much for submitting your manuscript "Differential long-term impact of Covid-19 on incidence of diabetes mellitus and cardiovascular diseases: population-based cohort study" (PMEDICINE-D-22-00177R1) for consideration at PLOS Medicine.

Your paper was evaluated by an associate editor and discussed among all the editors here. It was also discussed with an academic editor with relevant expertise, and sent to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. We cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

In addition, we request that you upload any figures associated with your paper as individual TIF or EPS files with 300dpi resolution at resubmission; please read our figure guidelines for more information on our requirements: http://journals.plos.org/plosmedicine/s/figures. While revising your submission, please upload your figure files to the PACE digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at PLOSMedicine@plos.org.

We hope to receive your revised manuscript by Mar 10 2022 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

We ask every co-author listed on the manuscript to fill in a contributing author statement, making sure to declare all competing interests. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. If new competing interests are declared later in the revision process, this may also hold up the submission. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT. You can see our competing interests policy here: http://journals.plos.org/plosmedicine/s/competing-interests.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

We look forward to receiving your revised manuscript.

Sincerely,

Callam Davidson,

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Please add continuous line numbering throughout the manuscript to facilitate future reviews.

Please update your title to ‘Cardiometabolic health outcomes after COVID-19 infection in the UK: A matched cohort study’, or similar.

The Data Availability Statement (DAS) requires revision. Note that a study author cannot be the contact person for the data.

Please structure your abstract using the PLOS Medicine headings (Background, Methods and Findings, Conclusions).

Abstract Background: Provide the context of why the study is important. The final sentence should clearly state the study question.

In the last sentence of the Abstract Methods and Findings section, please describe the main limitation(s) of the study's methodology.

Citations should be in square brackets, and preceding punctuation.

At this stage, we ask that you include a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract. Please see our author guidelines for more information: https://journals.plos.org/plosmedicine/s/revising-your-manuscript#loc-author-summary

Please ensure that the study is reported according to the STROBE guideline, and include the completed STROBE checklist as Supporting Information. Please add the following statement, or similar, to the Methods: "This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)."

The STROBE guideline can be found here: http://www.equator-network.org/reporting-guidelines/strobe/

When completing the checklist, please use section and paragraph numbers, rather than page numbers.

Please cite your protocol in the Methods section rather than the Acknowledgements.

Did your study have a prospective analysis plan? Please state this (either way) early in the Methods section.

a) If a prospective analysis plan (from your funding proposal, IRB or other ethics committee submission, study protocol, or other planning document written before analyzing the data) was used in designing the study, please include the relevant prospectively written document with your revised manuscript as a Supporting Information file to be published alongside your study, and cite it in the Methods section. A legend for this file should be included at the end of your manuscript.

b) If no such document exists, please make sure that the Methods section transparently describes when analyses were planned, and when/why any data-driven changes to analyses took place.

c) In either case, changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale.

The terms gender and sex are not interchangeable (as discussed in http://www.who.int/gender/whatisgender/en/ ); please use the appropriate term.

Please indicate in the figure caption the meaning of the bars in Figure 1.

Please refrain from including interpretation in the Results section (e.g. ‘The increase in CVD diagnoses was apparent in the last estimate for the pre-index period, which might imply that Covid-19 associated CVD might be recorded before Covid-19 diagnoses’ in paragraph 2 of the Results).

Please provide titles and legends for all figures (including those in Supporting Information files).

Please define any abbreviations used in Tables/Figures in the relevant legend.

Please present the unadjusted incident rate ratios as well as the adjusted ratios in Table 3.

Please relocate the Ethical Approval information to the Methods section.

The Funding, Competing interests, Contributors, and Data sharing sections should be removed from the manuscript and captured as part of your submission form responses (in the event of publication, this information will be included as metadata).

Please ensure references are formatted per our guidelines (e.g. use of et al. only after listing the first six authors, abbreviations per those found in the National Center for Biotechnology Information (NCBI) databases). Guidelines are available at https://journals.plos.org/plosmedicine/s/submission-guidelines#loc-references

Please include [preprint] in reference 25.

Comments from the reviewers:

Reviewer #1: It is essential to know the long-term cardiometabolic effects of COVID-19 so that measures can be taken early to prevent these after the acute phase of illness. In this context, this study's findings have important clinical implications.

1. What type of SARS-CoV-2 variant was prevalent during the study period?

2. Alcohol use, diet, physical activity, and lipid parameters (important risk factors for both CVD and diabetes) were not adjusted in the analysis. Are data not available? If yes, then acknowledge this as a limitation.

3. Cases and controls were matched on age, gender, and family practice, but the analyses were adjusted for age and gender. Isn't this over adjusting?

4. Apologies if I missed this - what % of patients were on glucose-lowering drugs, and what % were on insulin in the post COVID phase?

5. Comment on the % of participants with missing covariates.

Reviewer #2: In this study the authors examine acute, post-acute and long-term consequences of COVID-19 infection focusing primarily on cardiometabolic and pulmonary endpoints. They use cohort data from 1,473 family practices in UK and compare individuals with positive COVID-19 PCR test results to those who do not have a positive PCR test. For the analysis they remove prevalent cases of cardiometabolic diseases and examine incidence of cardiometabolic diseases after COVID-19. The study is well executed, and statistical methods are sound. The authors describe some of the limitations but there are a few additional limitations that I will address below that the authors should address.

1. The matched control individuals are healthier than those that get COVID-19. Therefore, some of the higher incidence of DM after COVID-19 may be related to higher baseline risk for DM. Please address this in the discussion.

2. Selection bias due to infection, severity of infection or availability for testing should be discussed.

3. The authors have information on BMI and smoking only in a subset of the cohort (Table 1). To account for missingness they add a new variable for missing data. It is possible that missing data is masking a worse cardiometabolic health (like higher BMI), or increased smoking. The authors should perform sensitivity analysis in a sample without any missing data.

Reviewer #3: Thanks for the opportunity to review your manuscript. My role is as a statistical reviewer so my comments focus on the design, data and analysis presented in the manuscript. I have put general questions first, and followed these with queries relevant to a specific section of the manuscript.

This study estimates whether people from England during 2021 were more likely to have a new diagnosis of DM or CVD in the 12 months after they were exposed to COVID-19. The data is sourced from routinely collected health data (CPRD). A cohort was created by matching patients with COVID-19 to unaffected patients, matching on age, gender, and home healthcare clinic.

Overall I think the study design and analysis is appropriate and the manuscript is clearly written. There are some additional details that need to be added to fully understand the limitations of the data sources used in the study. I was also not clear how missing covariate data was dealt with in the analyses.

There are some differences from the planned analyses (in the protocol on CPRD) - in particular the use of Poisson regression instead of Cox models. What was the rationale for the change?

One covariate that I would consider important to consider as a confounder between COVID-19 and CVD/DM is healthcare utilisation. My experience is that DM diagnosis can be sensitive to increased use of medical care, i.e. more ill patients seek care and fasting glucose is commonly part of the standard panel of blood tests (and subsequent A1C testing). Are measures of healthcare utilisation able to be derived from the CPRD data? These could explain the more modest IRRs for DM in the post-acute and long phases. This is briefly acknowledged in the discussion, but some way of checking this would be helpful to put these findings into context.

P2, Outcomes. Given that half of the sample were not exposed to Covid-19, i

P4, Paragraph 2. Have there been any studies with longer-term follow-up of these type of samples?

P5, Paragraph 1. When establishing an exposure-outcome association, representativeness is not strictly necessary, but it would helpful to point to a reference or table that shows this.

Does CPRD Aurum uniformly capture patients across England or are there regions that predominate? Not necessarily an issue - but I have seen previously studies with CPRD extracts that are more regionally focused (i.e. Northern England).

P5, Paragraph 2. How were the COVID-19 diagnoses linked to the CPRD data? Is this a routine linkage to patients from a notifiable diseases registry?

During this stage in England, were most COVID-19 diagnoses from PCR testing?

So just to clarify my understanding, effectively a pool of controls was created for each index date, including those not previously matched to a COVID-19 replacement (i.e. matched without replacement) and with no COVID-19 diagnosis before or on the index date? And stratified by home family practice?

Can the exclusions of patients with the numbers (e.g. excluding those with prevalent CVD/DM) be included in a flow diagram?

P5, Paragraph 3. From what sources do the CVD diagnoses come from? Patient self-report or linked hospitalisations?

Can you include a supplementary list (or reference) the drugs used to classify DM status? (i.e. ATC codes etc.). Were you able to separate the drug use by indication - e.g identify where these were for Gestational DM, or metformin use for PCOS?

P6, Paragraph 2. Generally how up to date were these covariates?

Strictly speaking, the outcome in the study is CVD or DM, so the patients were matched on exposure status not on outcome status. I would consider adjusting the terminology (e.g. 'case and control patients') as this does give the impression that this is a case-control study with COVID-19 as the outcome.

What was the rationale of examining glucocorticoids?

P6, Paragraph 7 (and next page). The first two sentences of this paragraph (periods of 28 days, and calculations performed in days) seem to contradict each other. Or are they referring to separate parts of the analysis?

Were the covariates in the analysis in the same form as in Table 1? Was age included as linear predictor? Was the fit of this checked?

What checks were made of the Poisson regression models - i.e. over-dispersion, distribution of residuals?

How was missing data (i.e. covariates) managed in the analyses?

Was the matched set the identity of each individual over the three time points? Or the pair of patients matched on age/gender/health centre?

Supp Table 2.

How was the overall difference estimated? Was this weighted according to follow-up time in each phase of the study?

Reviewer #4: This study aimed to quantify the risk of incident diabetes mellitus and cardiovascular disease in people registered in the CPRD Aurum primary care database as having had a positive test result for SARS-CoV-2 infection, compared to people who had no record of infection in the same data source.

The relevance of the research question is unquestionable, and the results of this study (if correct) would be very important for public health. However, I do have concerns about the suitability of the methods to address the research question. I list my main concerns below:

1. There is a high risk of information bias due to misclassification of the exposure (infection with SARS-CoV-2), particularly in the first wave. Access to testing was practically non-existent during the first months of the pandemic in the UK. Therefore, what is being mostly captured in electronic health records from primary care is symptomatic infection, rather than all infections with SARS-CoV-2. I acknowledge that the inclusion of 'suspected' cases in the analysis might have mitigated this effect to some extent, but the potential for controls to have been positive for SARS-CoV-2 during the first wave is non-negligible.

2. The authors state that they conducted a matched cohort study, but the results included two components:

2.1. In the first component, the authors describe the incidence of outcomes prior to the index date in the exposed and unexposed cohorts. There is vast evidence that people with comorbidities have higher risk of severe COVID-19 outcomes, should they get infected. As people with comorbidities are more likely to have more severe disease, they will also be more likely to present in primary care with symptoms and have the disease recorded in their health record. For this reason, I strongly disagree with the authors' claim in the discussion: "The study showed that patients recorded with Covid-19 in primary care are at slightly greater risk of CVD and DM even before the Covid-19 diagnosis." This study showed that people with history of CVD and DM are more likely to present with symptomatic infection of SARS-CoV-2, which we already knew. In addition, there is also a survival bias inherent in this analysis: the authors described the incidence of CVD and diabetes in two cohorts of people who survived long enough to present with symptomatic COVID. People who died from CVD are not included in the denominator.

2.2. I am also concerned with the execution of the second component of the study (i.e. the association between COVID-19 and incident CVD and DM). The authors stated that: "Matched sets that included case or control patients with prevalent CVD or DM, diagnosed within 12 months of the start of record or more than 12 months before the index date, were excluded." This is problematic. Patients with prevalent disease should have been excluded before matching, to avoid excluding pairs where the chosen control had CVD/DM history, as another control could have been selected instead of dropping the case. Also, the precise definition of the outcomes is not given (perhaps these could be included in appendix?) and thus is not clear to me what counted as an 'incident' event. For example, people with a DM diagnosis 5 years prior to COVID - were they excluded from the denominator in the descriptive 'incidence' analysis and dropped from the matched cohort study? This is unclear.

3. Following from the previous comment, could the authors provide the precise definitions of all variables used in this study and a flowchart for the patients' inclusion/exclusion. Some variables included in the models are prone to missingness but there is no mention to missing data in the methods. In line with best practice in the field, and to ensure transparency and reproducibility, could the authors make all code lists and patient selection information available in appendix?

4. The authors used 'long-COVID' to refer to patients who had SARS-CoV-2 infection 13 to 52 weeks after the index date. I acknowledge that this is clearly defined in the paper but 'long-COVID' has been used since the beginning of the pandemic to describe the state where symptoms linger after SARS-CoV-2 infection. I kept having to remind myself that 'long-COVID' in this paper was defined by time since index. Wondered if simply listing the time since index throughout the paper would make the manuscript clearer.

I have other minor comments:

Abstract

The abstract needs revision as it contains several typos. COVID-19 and SARS-CoV-2 infection are used interchangeably in the abstract - could this be corrected as the two are not equivalent. As mentioned above, I would also avoid the term 'Long covid'. The section for 'Outcomes' also includes the methods description - perhaps a mistake?

Introduction

The introduction gives a good rationale for this study. However, some sentences could be revised to improve the coherency of the argument being made:

(1) "Acute Covid-19 infection has been associated with new-onset cardiovascular disease (CVD) and diabetes mellitus (DM.)" Could you add a reference for this claim? This sounds like a fact and one wonders about what this study adds to the current body of research.

(2) "New onset hyperglycaemia has been reported in Covid-19 patients and is associated with worse prognosis" Please add a reference.

(3) "There were in excess of 1,000 deaths per day in the UK during April 2020 and January 2021" This is incorrect. According to the reference given by the authors, between April 2020 and January 2021, there were fewer than 1000 deaths registered in most of the days.

(4) "However, susceptibility to and severity of Covid-19 infection is known to be associated with cardiometabolic risk." What does susceptibility mean in this sentence? Susceptibility to the infection?

(5) "We employed the Clinical Practice Research Datalink (CPRD), a national database of anonymised primary care electronic health records, to identify a cohort of Covid-19 patients, comparing new CVD and DM diagnoses to a matched cohort with no Covid-19 diagnosis." Please note that none of the primary care databases at CPRD has national UK coverage. CPRD GOLD is no longer representative of the UK population in terms of geographical distribution of contributing practices. CPRD Aurum includes data from GP practices in England and Northern Ireland, and do not include practices from the devolved nations.

Methods

1. The number of practices varies across builds of the CPRD Aurum primary care database. Please clarify which build was used before stating the number of practices included in the database.

2. Please cross-check the number of patients registered in practices actively contributing with data to CPRD Aurum in the build used for this study. The % of the population covered is probably outdate too. I suggest that the author look for these numbers in the CPRD Release notes for the specific build used in this study.

3. Minor: please delete '18.5' prior to 'Kg/m2' in all categories.

4. Could the authors also include a DAG or some justification as to why they considered those variables as potential confounders of the association between COVID-19 and incident CVD / DM outcomes?

5. The information on PCR testing is interesting. Could the authors clarify how did they ascertain who got a PCR positive test?

6. What were mortality data used for?

Any attachments provided with reviews can be seen via the following link:

[LINK]

28 Feb 2022

Submitted filename: RezelPottsResponseToReviews28Feb2022.docx

Click here for additional data file.

1 Apr 2022

Dear Dr. Gulliford,

Thank you very much for submitting your revised manuscript "Cardiometabolic outcomes up to 12 months after COVID-19 infection. A matched cohort study in the UK" (PMEDICINE-D-22-00177R2) for consideration at PLOS Medicine.

Your paper was evaluated by an associate editor and discussed among all the editors here. It was also sent back to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of the reviewers' comments, we will not be able to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments. Obviously we cannot make any decision about publication until we have seen the revised manuscript and your response, and we plan to seek re-review by one or more of the reviewers.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

We hope to receive your revised manuscript by Apr 29 2022 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

We look forward to receiving your revised manuscript.

Sincerely,

Callam Davidson,

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Please provide an explanation as to why your use of the same data system is different from Rev #4.

Comments from the reviewers:

Reviewer #1: Thank you authors for addressing my comments and revising the manuscript accordingly. Please change "moderators" to "confounders" in your response to comment 2.

Reviewer #2: The authors have sufficiently addressed all my comments.

Reviewer #3: Thanks for the revised manuscript and response to my original queries. Overall the revisions and responses clarified all of my original review and apart from one very small change I recommend that the manuscript be accepted for publication. The editor can safely ignore the cut-off comment - this should have been removed before submitting the initial review.

The explanation for the change in analysis to allow for a difference-in-difference analysis makes sense to me, and I agree that generally you will see similarity between HRs and IRRs when each analysis is applied to the same data. The DID approach with this type of data really looks effective.

I agree with your reply to Reviewer 1, point 2 about the adjustment for age and sex. This is an approach often recommended as being 'doubly-robust' in that if one (e.g. matching) fails then the other (e.g. covariate) will be effective. Even if data is already well-matched and confounding with the exposure is 'eliminated', it is still a good idea if there is an association between these covariates and the outcome, as it shrinks the unexplained variance and you end up with more precise estimates of standard error (and hence narrower 95% CI) of other estimates in the model.

The one small pedantic change requested - for categories of BMI (L114), I would use "18.5 - <25 kg/m2", as if it were "18.5-24.9" strictly speaking someone with a BMI of 24.95 would not be classified into any category.

Reviewer #4: RE Comment 2.2:

Thank you for clarifying what was done. The authors stated that "CVD and DM status could only be determined after controls were selected and their full data was extracted from the database". This is untrue. The 'define' tool, in the CPRD data platform, allows researchers to obtain a list of all patients with a record of a given condition (as defined in the code list). This includes CVD and DM. Once you ran the define and obtain a list of patients with history of CVD and DM, this can be merged with the denominator files (the complete list of patients that you mentioned) to identify and exclude those that are not eligible for the study. Matching should only be done after creating the pool of potentially eligible controls.

I understand this is not what was done, and I acknowledge that the authors were transparent about this in the revised version of the manuscript. I do not know if the non-standard design of the study introduced bias, and if yes, of what magnitude. I suspect this won't change the study conclusion, but I do struggle with this error that could be easily avoided. A statistical reviewer may be better suited to advise.

Thank you for clarifying all other points and thank you for the opportunity to review this paper.

Any attachments provided with reviews can be seen via the following link:

[LINK]

21 Apr 2022

Submitted filename: RezelPottsResponseToReviews20April2022.docx

Click here for additional data file.

18 May 2022

Dear Dr. Gulliford,

Thank you very much for submitting your revised manuscript "Cardiometabolic outcomes up to 12 months after COVID-19 infection. A matched cohort study in the UK" (PMEDICINE-D-22-00177R3) for consideration at PLOS Medicine.

Your paper was re-evaluated by an associate editor and discussed among all the editors here. It was also discussed again with an academic editor with relevant expertise, and sent back to independent reviewers, including a statistical reviewer. The reviews are appended at the bottom of this email and any accompanying reviewer attachments can be seen via the link below:

[LINK]

In light of these reviews, I am afraid that we are still unable to accept the manuscript for publication in the journal in its current form, but we would like to consider a revised version that addresses the reviewers' and editors' comments.

In revising the manuscript for further consideration, your revisions should address the specific points made by each reviewer and the editors. Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments, the changes you have made in the manuscript, and include either an excerpt of the revised text or the location (eg: page and line number) where each change can be found. Please submit a clean version of the paper as the main article file; a version with changes marked should be uploaded as a marked up manuscript.

We hope to receive your revised manuscript by Jun 01 2022 11:59PM. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

Please use the following link to submit the revised manuscript:

https://www.editorialmanager.com/pmedicine/

Your article can be found in the "Submissions Needing Revision" folder.

We look forward to receiving your revised manuscript.

Sincerely,

Callam Davidson,

Associate Editor

PLOS Medicine

plosmedicine.org

-----------------------------------------------------------

Requests from the editors:

Please clarify your methodology in response to the comments of reviewer #4.

Please use track changes rather than highlighting to indicate changes in your marked up manuscript – the previous version provided did not highlight all the changes made between the R2 and R3 versions.

Please ensure all figures can be interpreted without the need to refer to the main text (Figure 1 for example needs either a key or information in the legend to define which line represents patients vs controls).

Please add the following statement, or similar, to the Methods: ""This study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guideline (S1 Checklist)."

Changes in the analysis-- including those made in response to peer review comments-- should be identified as such in the Methods section of the paper, with rationale.

Please ensure all supplementary tables are cited in the main text (I couldn’t locate a reference to Supplementary Table 2).

Comments from the reviewers:

Reviewer #3: I noted after the last review that excluding the matched sets for a members with prevalent DM/CVD shouldn't introduce bias, it's reassuring to see that when this step is applied earlier there is no appreciable difference in the pattern of DM incidence.

The additional analysis by 4-week period with the extended data is quite interesting - better not to speculate about mechanisms, but I will just add that I think this is a valuable addition.

One very small check, and an addition as well. For fig 3 the confidence intervals come from the SE estimates of the main model? If so then I would just then specify the method used to create the smoothed line.

Great paper - whoever was doing the analysis might need a day off after managing to update the paper with additional data so quickly

Reviewer #4: Thank you for addressing my comments. Unfortunately, I don't think the methods are clear. The authors described using the Feb 2021 version of Aurum to identify patients with COVID-19. They then mention that controls (I am assuming they meant to say controls when they said 'patients') 'were randomly selected from the registered population of CPRD Aurum March 2021 release' (p. 7, lines 88-89). Subsequently, the authors mentioned that 'data were updated to the March 2022 release of Aurum for final analysis'. I am unclear why three different versions of CPRD Aurum were used, when all information you need is provided in the most recent version of the database. This gives me no confidence that the authors handled the data properly, and therefore I cannot recommend this paper for publication.

Any attachments provided with reviews can be seen via the following link:

[LINK]

20 May 2022

Submitted filename: RezelPottsResponseToReviews19May2022.docx

Click here for additional data file.

10 Jun 2022

Dear Dr. Gulliford,

Thank you very much for re-submitting your manuscript "Cardiometabolic outcomes up to 12 months after COVID-19 infection. A matched cohort study in the UK" (PMEDICINE-D-22-00177R4) for review by PLOS Medicine.

I have discussed the paper with my colleagues and the academic editor and it was also seen again by one reviewer. I am pleased to say that provided the remaining editorial and production issues are dealt with we are planning to accept the paper for publication in the journal.

The remaining issues that need to be addressed are listed at the end of this email. Any accompanying reviewer attachments can be seen via the link below. Please take these into account before resubmitting your manuscript:

[LINK]

***Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.***

In revising the manuscript for further consideration here, please ensure you address the specific points made by each reviewer and the editors. In your rebuttal letter you should indicate your response to the reviewers' and editors' comments and the changes you have made in the manuscript. Please submit a clean version of the paper as the main article file. A version with changes marked must also be uploaded as a marked up manuscript file.

Please also check the guidelines for revised papers at http://journals.plos.org/plosmedicine/s/revising-your-manuscript for any that apply to your paper. If you haven't already, we ask that you provide a short, non-technical Author Summary of your research to make findings accessible to a wide audience that includes both scientists and non-scientists. The Author Summary should immediately follow the Abstract in your revised manuscript. This text is subject to editorial change and should be distinct from the scientific abstract.

We hope to receive your revised manuscript by EOB Wednesday 15th June. Please email us (plosmedicine@plos.org) if you have any questions or concerns.

We ask every co-author listed on the manuscript to fill in a contributing author statement. If any of the co-authors have not filled in the statement, we will remind them to do so when the paper is revised. If all statements are not completed in a timely fashion this could hold up the re-review process. Should there be a problem getting one of your co-authors to fill in a statement we will be in contact. YOU MUST NOT ADD OR REMOVE AUTHORS UNLESS YOU HAVE ALERTED THE EDITOR HANDLING THE MANUSCRIPT TO THE CHANGE AND THEY SPECIFICALLY HAVE AGREED TO IT.

Please ensure that the paper adheres to the PLOS Data Availability Policy (see http://journals.plos.org/plosmedicine/s/data-availability), which requires that all data underlying the study's findings be provided in a repository or as Supporting Information. For data residing with a third party, authors are required to provide instructions with contact information for obtaining the data. PLOS journals do not allow statements supported by "data not shown" or "unpublished results." For such statements, authors must provide supporting data or cite public sources that include it.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript.

Please note, when your manuscript is accepted, an uncorrected proof of your manuscript will be published online ahead of the final version, unless you've already opted out via the online submission form. If, for any reason, you do not want an earlier version of your manuscript published online or are unsure if you have already indicated as such, please let the journal staff know immediately at plosmedicine@plos.org.

If you have any questions in the meantime, please contact me or the journal staff on plosmedicine@plos.org.

We look forward to receiving the revised manuscript by Jun 15 2022 11:59PM.

Sincerely,

Callam Davidson,

Associate Editor

PLOS Medicine

plosmedicine.org

------------------------------------------------------------

Requests from Editors:

In the last subsection of the abstract, please update to “In this study, we found that CVD was increased early after Covid-19 mainly from pulmonary embolism, atrial arrhythmias and venous thromboses. DM incidence remained elevated for at least 12 weeks following Covid-19 before declining.”

Please ensure journal abbreviations in your references align with those found in the National Center for Biotechnology Information (NCBI) databases.

Comments from Reviewers:

Reviewer #4: Thank you for addressing my comments. The revised version of the manuscript is much clearer in the description of the methods.

Any attachments provided with reviews can be seen via the following link:

[LINK]

14 Jun 2022

Dear Dr Gulliford,

On behalf of my colleagues and the Academic Editor, Dr Weiping Jia, I am pleased to inform you that we have agreed to publish your manuscript "Cardiometabolic outcomes up to 12 months after COVID-19 infection. A matched cohort study in the UK" (PMEDICINE-D-22-00177R5) in PLOS Medicine.

Before your manuscript can be formally accepted you will need to complete some formatting changes, which you will receive in a follow up email. Please be aware that it may take several days for you to receive this email; during this time no action is required by you. Once you have received these formatting requests, please note that your manuscript will not be scheduled for publication until you have made the required changes.

When making the formatting changes, please also address the following comments:

* Apologies that my previous comment was unclear - please re-insert the final sentence from the previous draft of your manuscript into the Abstract Conclusions ('People without pre-existing CVD or DM who suffer from COVID-19 do not appear to have a long-term increase in incidence of these conditions.').

* Please remove 'Funding: UK National Institute for Health Research.' from the Abstract.

In the meantime, please log into Editorial Manager at http://www.editorialmanager.com/pmedicine/, click the "Update My Information" link at the top of the page, and update your user information to ensure an efficient production process.

PRESS

We frequently collaborate with press offices. If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximise its impact. If the press office is planning to promote your findings, we would be grateful if they could coordinate with medicinepress@plos.org. If you have not yet opted out of the early version process, we ask that you notify us immediately of any press plans so that we may do so on your behalf.

We also ask that you take this opportunity to read our Embargo Policy regarding the discussion, promotion and media coverage of work that is yet to be published by PLOS. As your manuscript is not yet published, it is bound by the conditions of our Embargo Policy. Please be aware that this policy is in place both to ensure that any press coverage of your article is fully substantiated and to provide a direct link between such coverage and the published work. For full details of our Embargo Policy, please visit http://www.plos.org/about/media-inquiries/embargo-policy/.

To enhance the reproducibility of your results, we recommend that you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. Additionally, PLOS ONE offers an option to publish peer-reviewed clinical study protocols. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols

Thank you again for submitting to PLOS Medicine. We look forward to publishing your paper.

Sincerely,

Callam Davidson

Associate Editor

PLOS Medicine