Incident allergic diseases in post-COVID-19 condition: multinational cohort studies from South Korea, Japan and the UK – Nature.com

Data source

The Kyung Hee University (KHUH 2022-06-042), the Korea Disease Control and Prevention Agency (KDCA), the National Health Insurance Service (NHIS; KDCA-NHIS-2022-1-632) of South Korea, JMDC (PHP-00002201-04), and UKB (94075) approved the study protocol.

Written informed consent was obtained from all participants at enrollment. We used three large-scale, nationwide and population-based cohort designs in this study: a South Korean nationwide cohort (K-COV-N cohort [main cohort]; total n=10,027,506), a Japanese claims-based cohort (JMDC cohort [replication cohort A]; total n=12,218,680) and a UK prospective cohort from the UK Biobank (UKB cohort [replication cohort B]; total n=468,617). Both the K-COV-N and JMDC cohorts employ a universal health insurance system. The UKB, meanwhile, is a dataset comprised of voluntary participation, including biomedical samples and health information. Detailed explanations of the JMDC and UKB cohorts can be found in supplemental material section.

The K-COV-N cohort is a large-scale, nationwide, general population-based cohort in South Korea, covering 98% of the South Korean population34. The cohort was developed and provided by the NHIS of South Korea and KDCA focused on individuals aged 20 years between January 1, 2018, and December 31, 2021. It contained information on COVID-19 vaccination, SARS-CoV-2 test results, COVID-19-related outcomes, results of national health examination, death records, and health insurance data including outpatient and inpatient information. The following characteristics of the Korean database enable us to construct a well-designed cohort: (1) A comprehensive healthcare system, implemented by the Korean government, covers people who have been infected with SARS-CoV-2; (2) all information was anonymized by the Korean government34; (3) It includes SARS-CoV-2 test results, vaccination status, and COVID-19-related hospital records; and (4) the overall predictive value for diagnostic records of the NHIS was 82% according to a previous study6,35,36.

We included all individuals aged 20 years with COVID-19 and non-infected participants from 2020 to 2021 (total n=10,027,506). We precluded those who meet the following criteria: (1) insufficient socioeconomic information or died before; and (2) history of allergic diseases in the pre-observation period, defined as two years (n=4,335,150). Eventually, 5,692,356 individuals were included from South Korea in this study.

The exposure was SARS-CoV-2 infection, which was defined if the participants tested positive for COVID-19 either by real-time reverse transcriptase polymerase chain reaction or rapid antigen testing of nasopharyngeal swabs. We considered the original SARS-CoV-2 if the initial infection was before July 31, 2021, and the delta variant was from August 1, 202137. Patients who were admitted to an intensive care unit and those who required oxygen therapy, extracorporeal membrane oxygenation, renal replacement, or cardio resuscitation were perceived as having moderate to severe COVID-1938. The others were considered having mild COVID-19. The COVID-19 vaccination status was categorized according to dosage (unvaccinated, 1, and 2 times). Individuals who were vaccinated with the Johnson & Johnson/Janssen vaccine were considered twice vaccinated after the single dose.

The primary outcome was the onset of allergic diseases, including: asthma, AR, AD, and FA7. Also, the term allergic diseases refers to a diagnosis of any of the following condition: asthma, AR, AD, or FA39,40. Allergic asthma was identified as asthma combined with an additional allergic disorder (AR, AD, or FA), while non-allergic asthma was classified as asthma occurring in the absence of any allergic diseases7. We defined patients with allergic diseases as those having at least two claims during the observation period and were taking relevant medications. We provided a list of the ICD-10 codes and medications used to define each disease in this study (TableS1).

The demographic characteristics of the participants were obtained from the health insurance database as followings: sex, age (2039, 4059, and 60 years), household income (low [039 percentile], middle [4079 percentile], and high [80100 percentile]), and region of residence (urban and rural)34. The information on body mass index (underweight [<18.5kg/m2], normal [18.523.0kg/m2], overweight [23.025.0kg/m2], obese [25.0kg/m2], and unknown), blood pressure (systolic blood pressure <140mmHg and diastolic blood pressure <90mmHg, systolic blood pressure 140mmHg or diastolic blood pressure 90mmHg, and unknown), fasting blood glucose (<100, 100mg/dL, and unknown), serum total cholesterol (<200, 200240, 240mg/dL, and unknown) and glomerular filtration rate (<60, 6090, 90mL/min/1.73m2, and unknown) were included from the fasting serum samples of national health examination41. The CCI, history of cardiovascular disease, chronic kidney disease, and chronic obstructive pulmonary disease, history of medication use for diabetes, hyperlipidemia, and hypertension, smoking status (non-, ex-, and current smoker), alcoholic drinks (<1, 12, 34, 5 days per week, and unknown), and aerobic physical activity (sufficient [600 Metabolic Equivalent Task scores], insufficient, and unknown) were collected based on ICD-10 code and/or results of national health examination12,42. Additionally, to minimize bias related to missing data, we focused on the missing indicator method, generating missing indicator variables and incorporating them into the adjustment variables43.

We executed 1:5 exposure-driven propensity score matching to balance the distribution of covariates in the two groups. We used a greedy nearest-neighbor algorithm with random selection without replacement within caliper widths of 0.001 standard deviations44,45. We assessed the adequacy of matching by comparing SMDs. A SMD<0.1 indicated no major imbalance in the two groups44,45. We constructed the following covariates as matching variables for South Korea: age, sex, household income, region of residence, CCI, body mass index, blood pressure, fasting blood glucose, serum total cholesterol, glomerular filtration rate, smoking status, alcoholic drinks, aerobic physical activity, and history of medication use for diabetes mellitus, dyslipidemia, and hypertension. For the replication cohorts of Japan and the UK, we also used similar covariates as matching variables (Supplement Material). All covariates were regarded as adjustment variables in further statistical models. After propensity score matching, a total of 836,164 individuals were included in the study (FigureS1 and Table1).

The same ICD-10 codes, definition of exposures and outcomes, observation period, and propensity score matching were utilized for the JMDC and the UKB cohorts as well (Supplement Material). Due to the absence of SARS-CoV-2 vaccination data41, the JMDC and the UKB cohort were used only to validate the main findings of the K-COV-N cohort. After propensity score matching, the JMDC and the UKB cohorts consisted of 2,541,021 and 325,843 individuals, respectively (Figs.S2 and S3).

As aforementioned, SARS-CoV-2 infection was defined as primary exposure and the incident allergic diseases after at least 30 days of infection was defined as the primary outcome in the general population-based cohorts of South Korea, Japan and the UK (TablesS2S3). To overcome immortal time bias, the date of the first diagnosis of SARS-CoV-2 was perceived as the individual index date. We considered 20182019 the pre-observation period to observe the history of medical diagnosis. The observation period of the Korean cohort was between January 1, 2020, and December 31, 2021. The follow-up ended on December 31, 2021, or upon the death of the subject (Fig.S4).

We performed 1:5 exposure-driven propensity matching in the nationwide cohorts of South Korea, Japan, and the UK (Table1 andS4, S5). A Cox proportional hazard regression model with estimates of HRs and 95% CIs was used to explore incident overall and four subtypes (asthma, AR, AD, and FA) of allergic diseases associated with post-COVID-19 conditions45. We further assessed the time attenuation effect of allergic diseases following SARS-CoV-2 infection (<3, 36, and 6 months) to reduce reverse causation. This refers to the duration it took for patients infected with COVID-19 to be diagnosed with allergic diseases, and includes individuals who had not been diagnosed during the pre-observation period. We performed several subgroup analyses to the following parameters: severity of COVID-19 (mild and moderate to severe), strain type (original and delta), and dosage of SARS-CoV-2 vaccination (0, 1, and 2 times). In addition, we executed stratification analyses according to sex, age, household income, CCI, body mass index, alcohol drinking status, aerobic physical activity and strain type of SARS-CoV-2 (TablesS11S20). We used SAS (version 9.4; SAS Institute Inc., Cary, NC, USA) to perform all statistical analyses in this study. A two-sided p-value less than 0.05 was considered statistically significant (TablesS23S25).

We conducted sensitivity analyses to assess the reliability of the findings from our primary analyses. First, to validate the study results and identify detection bias, we included tympanic membrane perforation disease as a negative control in our analyses for both the main and replication cohorts (TableS26)46. Second, to reduce misclassification bias due to dyspnea, we performed an analysis excluding symptoms of dyspnea in asthma cases. (TableS27). Third, we established a strict diagnostic criterion for asthma in the main cohort (TableS28). We conducted analyses on cases diagnosed with asthma, considering those with a history of emergency department visits or hospitalization47. Fourth, allergic asthma and non-allergic asthma were compared as distinct groups due to differences in the asthma phenotype (TableS29). Fifth, in order to examine the impact of COVID-19 severity on allergic diseases, the mild group and the moderate to severe group were analyzed as two separate cohorts (TablesS21 and S22). Sixth, we analyzed the onset of allergic diseases in relation to SARS-CoV-2 infection status among individuals with the same number of vaccine doses, for understanding the long-term immune protection provided by the COVID-19 vaccine and its effectiveness extent (TableS30). In the same context, we conducted a time attenuation analysis to identify potential impacts, including the decrease in immunity over time (TableS31).

In the case of the main cohort and replication cohort A, the outcome measures were determined independently, without any involvement from the participants. In contrast, for replication cohort B, the participants were directly involved in determining the outcome measures through a process of voluntary reporting. The study design and implementation were conducted without consultation. However, we plan to disseminate the results of this study to all study participants and wider relevant communities upon request.

Further information on research design is available in theNature Portfolio Reporting Summary linked to this article.

Go here to see the original:

Incident allergic diseases in post-COVID-19 condition: multinational cohort studies from South Korea, Japan and the UK - Nature.com