Investigating the occurrence of autoimmune diseases among children and adolescents hospitalized for Mycoplasma pneumoniae infections

Background Mycoplasma pneumoniae infection is common in the general population and may be followed by immune dysfunction, but links with subsequent autoimmune disease remain inconclusive. Objective To estimate the association of M. pneumoniae infection with the risk of subsequent autoimmune disease. Methods This retrospective cohort study examined the medical records of South Korean children from 01/01/2002 to 31/12/2017. The exposed cohort was identified as patients hospitalized for M. pneumoniae infection. Each exposed patient was matched with unexposed controls based on birth year and sex at a 1:10 ratio using incidence density sampling calculations. The outcome was subsequent diagnosis of autoimmune disease, and hazard ratios (HRs) were estimated with control for confounders. Further estimation was performed using hospital-based databases which were converted to a common data model (CDM) to allow comparisons of the different databases. Results The exposed cohort consisted of 49,937 children and the matched unexposed of 499,370 children. The median age at diagnosis of M. pneumoniae infection was 4 years (interquartile range, 2.5–6.5 years). During a mean follow-up time of 9.0 ± 3.8 years, the incidence rate of autoimmune diseases was 66.5 per 10,000 person-years (95% CI: 64.3–68.8) in the exposed cohort and 52.3 per 10,000 person-years (95% CI: 51.7–52.9) in the unexposed cohort, corresponding to an absolute rate of difference of 14.3 per 10,000 person-years (95% CI: 11.9–16.6). Children in the exposed cohort had an increased risk of autoimmune disease (HR: 1.26; 95% CI: 1.21–1.31), and this association was similar in the separate analysis of hospital databases (HR: 1.25; 95% CI 1.06–1.49). Conclusion M. pneumoniae infection requiring hospitalization may be associated with an increase in subsequent diagnoses of autoimmune diseases.


Introduction
Mycoplasma pneumoniae (M.pneumoniae) is one of the most frequent causes of respiratory diseases and is responsible for 10% to 40% of all cases of community-acquired pneumonia (CAP) in children.M. pneumoniae, a gram-negative bacterium lacking cell walls and characterized by its small size, depends on host cell association as an extracellular pathogen.It develops a specialized attachment organelle that potentially leads to the formation of capsular material outside the cell membrane.This bacterium is a significant respiratory pathogen in children (1).Although many infected individuals completely recover without complications, some experience sustained chronic diseases (2)(3)(4).More specifically, some children experience extrapulmonary manifestations that influence specific organs or whole-body systems after an infection.Some evidence suggests that a M. pneumoniae infection may lead to an increased development of autoimmune diseases (5).
There is increasing interest in the impact of M. pneumonia infection on autoimmune diseases (2,6), and many case reports have examined this relationship (7,8).However, there is no definitive evidence of a causal relationship because these studies had cross-sectional designs (9,10), observational studies lacking appropriate controls (11), and examined small numbers of patients who were mostly from specialist clinics (12).There are also conflicting results from case-control studies regarding the relationship of M. pneumoniae infection with common childhood autoimmune diseases, such as Kawasaki disease and Guillain-Barre's syndrome (13).We are unaware of any rigorous epidemiologic investigations that examined the effect of M. pneumoniae infection on subsequent autoimmune diseases.
The purpose of this study was to determine the association of M. pneumoniae infection with the risk of subsequent autoimmune diseases using a nationwide population-based database that has information on all medical diagnoses and use of healthcare resources, with control for confounding by comparison with hospital-based data.

Study design
This retrospective population-based cohort study enrolled all individuals born in South Korea between 2002 and 2005, as identified by the National Health Insurance Service (NHIS), with linked Statistics Korea census data.The NHIS provides healthcare services coverage for over 98% of the South Korean population and includes claims-based medical data.These data provided information on demographic characteristics, healthcare utilization (with diagnostic codes from the 10th version of the International Classification of Diseases, ICD-10), prescriptions, and relevant procedures.All guidelines for observational studies that use routinely collected health data were followed (Supplementary Table 1).

Study population
A total of 1,914,461 subjects were included and followed up from birth until 31/12/2017, using all included participants who were less than 18 years old on that date (Figure 1).The index date was defined as the date of first hospitalization due to M. pneumoniae infection.Individuals with underlying history of lower respiratory tract infection due to other pathogens (ICD-10 code J12X-J18X), respiratory and cardiovascular disorders specific to the perinatal period (ICD-10 code P20X-P29X), congenital malformations of the respiratory system (ICD-10 code Q30-Q34), chromosomal anomaly (ICD-10 code Q90-Q99), and lower respiratory tract infection due to solids and liquids (ICD-10 code J69X) were excluded.Individuals diagnosed as autoimmune diseases prior to the index date were also excluded.Lower respiratory tract infections due to other viral pathogens and infections due to ingestion of solids or liquids were also excluded, considering the potential inclusion of M. pneumoniae diagnoses or insufficient differentiation within these categories.Individuals in the exposed cohort were matched with unexposed controls at a 1:10 ratio based on birth year and sex using incidence density sampling.The unexposed cohort participants were randomly matched based on birth year and sex and had no M. pneumoniae infection or autoimmune diseases at the index patient's diagnosis date.

Exposure
The exposure in this study was defined as hospitalization due to M. pneumoniae infection, with relevant ICD-10 codes including J200 (acute bronchitis due to M. pneumoniae); J157 (pneumonia due to M. pneumoniae); B960 (M.pneumoniae as the cause of diseases classified elsewhere), or A493 (Mycoplasma infection, unspecified site) (14-16) (Supplementary Table 2).We referred to existing studies to define the exposure to enhance validity and minimize the possibility of misclassification (Supplementary Tables 3-5; literature search in Supplementary Table 6).Moreover, two pediatricians (Ha and Han) performed separate analyses to validate the predictive value of ICD-10 codes for the diagnosis of M. pneumonia infections (J200, J157, B960, or A493) by reviewing electronic medical records (EMRs) at Bundang CHA Medical Center and Kangnam Sacred Heart Hospital (17).The positive predictive values (PPVs) were 97.8% (95 CI, 95.8%-99.9%)and 97.5% (95% CI, 95.3%-99.7%),indicating that the diagnostic assessment used herein had high accuracy and was suitable for use in this analysis.
In addition, a Common Database Model (CDM) comparison, consisting of hospital-based data from four different hospitals that had data of patients with M. pneumonia infections and a matched unexposed cohort, was conducted to determine the robustness of the main results.These data were from different healthcare institutions and allowed large-scale distributed comparative effectiveness analysis (18).Evidence was generated using standard analytic tools by converting data to a CDM.For this analysis, patient-based retrospective cohort data were from four hospitals (Myongji Hospital [MJ], Kyung Hee University Hospital at Gangdong [KH], Bundang CHA Medical Center [CHA], Pusan National University Hospital [PS]).Heterogeneous data sources were standardized and analyzed using the analysis program that was distributed to these four institutions.

Follow-up
All participants were followed from 01/01/2002 until the first diagnosis of an autoimmune disease, death, or the end of the study (31/12/2017), whatever happened first.The follow-up of unexposed individuals included the determination of a M. pneumonia infection later; an individual who was censored due to a M. pneumonia infection was transferred to the exposed group.The first year of follow-up was excluded from the analyses to reduce the probabilities of reverse causality and surveillance bias.The exposed individuals were followed for an average of 9.9 years (SD 3.5) and the unexposed cohort for 10.1 years (SD 3.4).

Autoimmune diseases
Information on autoimmune diseases was from the NHIS database.The 41 autoimmune diseases (19,20) were identified by ICD-10 codes (Supplementary Table 2).

Covariates
Data were examined for age, sex, follow-up time, birth residence, household income, calendar year/season at birth, any medical condition during the perinatal period, asthma comorbidity, and inpatient and outpatient hospital visits during the first year after study entry (21)..The most updated information before the index date was used for all analyses, and data on inpatient and outpatient visits were collected as a surrogate for use of medical resources during follow-up.Asthma was defined as one or more physician diagnoses 6 months prior to the index date, with one or more events of asthma exacerbation requiring hospitalization and/ or systemic corticosteroids (22,23).Use of systemic macrolide antibiotics infections during any month after the index date, duration of hospitalization, and use of oxygen therapy for treating M. pneumoniae during the first year after study entry were also assessed.We removed the first month after study entry from this calculation considering that exposed patients are likely to receive intensive medical care during the first month after M. pneumoniae infection.

Statistical analysis
Conditional Cox models were used to estimate hazard ratios (HRs) with 95% CIs for developing an autoimmune disease due to previous M. pneumonia infection based on time after the index date.For this analysis, specific autoimmune diseases with fewer than 100 observed cases were not analyzed separately, but it was used in calculations of hazard ratios (HRs) for the corresponding main category.We performed stratified analyses by matching identifiers (birth year and sex) and adjusted for age group at index date (≤60 months or >60 months), time since index date (1 to <5 years or ≥5 years), calendar year of birth (2002-2003 or 2004-2005), birth residence (Seoul and metropolitan, city, or rural area), household income (by tertiles, low, middle, or high), calendar season at birth date (spring, summer, fall, or winter), any medical condition during the perinatal period (yes or no), asthma comorbidity (yes or no), inpatient hospital visits during the first year after study entry (yes or no), and outpatient hospital visits during the first year after study entry (≤13 or >13).All analyses were stratified by matching identifiers (birth year and sex) and adjusted for birth residence (Seoul/ metropolitan, city, or rural area), household income (low, middle, or high), and perinatal history (disorder related to the length of gestation and fetal growth, birth trauma, infections specific to the perinatal period, congenital malformation/deformation, and chromosomal abnormalities).HRs were separately calculated for sex, age at the index date (≤60 months or >60 months), time since the index date (1-5 years or ≥5 years), calendar year of birth (2002-2003 or 2004-2005), birth residence (Seoul/metropolitan, city, or rural area), household income (low, middle, or high), the season of birth (spring, summer, fall, or winter), disorders related to any perinatal status (yes or no), history of asthma (yes or no), history of hospital admission (yes or no), and the number of outpatient visits (≤13 or >14, based on the median) during the first year after study entry.In the stratified analysis, information from the first month after study entry was not considered, as medical service utilization within that month was likely attributed to the exposure itself.Differences in HRs by variables were assessed by introducing an interaction term into the Cox models.The difference in log (HR) between strata was used to calculate the z-score and P value for this comparison.The absolute rate differences with 95% CIs were also determined for all associations.
Sensitivity analyses were performed to examine the effect of using a more stringent definition of M. pneumonia infection (ICD-10 code J157 alone), two or more and three or more autoimmune syndromes, individual autoimmune diseases, any of the major groups of autoimmune diseases (Supplementary Table 2) (20), or a more stringent definition of the outcome (use of medication such as non-steroidal anti-inflammatory drugs [NSAIDs], systemic steroid, intravenous immunoglobulin (24,25), thyroid medications).
To determine the possible effects of reverse causality and surveillance bias, the analyses were repeated by excluding participants who were enrolled during the first 2 years (2002)(2003) or during the first 5 years (2002)(2003)(2004)(2005)(2006)(2007).In this analysis, the cumulative incidence curves of autoimmune diseases among participants with more than 1 year of follow-up were analyzed.
Furthermore, we performed an analysis using the CDM to assess the reliability of our results.We utilized four hospital-based cohorts that had been transformed into the OMOP-CDM (Observational Medical Outcomes Partnership Common Data Model) format.This format enabled the utilization of deidentified patient data extracted from EMRs.The hospitals included Myongji Hospital (MJ), Kyung Hee University Hospital at Gangdong (KH), CHA University Bundang CHA Medical Center (CHA), and Pusan National University Hospital (PS).This allowed for the multicenter analysis of disparate databases.This consisted of the following information: person, drug exposure, drug era, condition occurrence, condition error, observation period, observation, procedure occurrence, visit occurrence, death, drug cost, procedure cost, location, provider, organization, care site, payment plan period, and cohort (26).The analysis period and total number of patients were 2003 to 2020 and 880,392 from MJ, 2006 to 2017 and 822,183 from KH, 2006 to 2019 and 2,363,386 from CHA, and 2011 to 2018 and 1,753,001 from PS.In total, 5,818,962 patients were enrolled across all hospitals, consisting of 5,239 individuals in the exposed cohort (M.pneumonia-related hospitalization) and 37,640 in the unexposed cohort (Figure 2).All analyses were conducted in SAS statistical software version 9.4 (SAS Institute), and a two-sided P value below 0.05 was considered statistically significant.

Ethics statement
The

Baseline characteristics of the participants
All individuals who required hospitalization due to an initial M. pneumonia infection between 2002 and 2005 were identified (Figure 1).We excluded 348,741 individuals who met our predefined criteria, including those with a history of respiratory or cardiovascular complications during the intrauterine and perinatal period (n = 19,826), congenital malformations (n = 9,745), and chromosomal anomalies (n = 4,061).This left us with 52,805 children with M. pneumonia infections who met our criteria for inclusion in the study.We further excluded 2,868 children who had been diagnosed with autoimmune diseases, leaving us with 49,937 individuals in the exposed cohort.For the unexposed cohort, we identified individuals who had not been hospitalized for M. pneumonia infections and who met our inclusion criteria for population matching.Overall, there were 49,937 participants in the exposed cohort and 499,370 in the matched unexposed cohort.
The median age at diagnosis of M. pneumoniae infection was 3.67 years, and 48.54% of all patients were men (Table 1).Compared with the unexposed cohort, the exposed cohort had more inpatient and outpatient hospital visits and a higher rate of asthma comorbidity.The incidence of M. pneumoniae infection was greatest (0.8%) in children who were 2 and 3 years old and decreased steadily with age (Supplementary Figure 1).Trends in M. pneumoniae infection-related hospitalization from 2002 to 2016, analyzed using our database, are presented in Supplementary Figure 2. The variations in prevalence might be explained by the M. pneumonia epidemics in Korea during specific periods (27).

Relationship between hospitalization due to M. pneumoniae infection and increased risks to autoimmune diseases
Elevated HRs of autoimmune diseases in the exposed group with history of hospitalization due to M. pneumoniae infection was shown by the Cox model with adjustment of multiple confounders, compared with the unexposed group (Table 2).During a mean follow-up time of 10 years, we identified newly diagnosed autoimmune diseases in 3295 individuals in the exposed cohort (incidence rate: 66.5 per 10,000 person-years; 95% CI: 64.3-68.8)and in 26,145 individuals in the unexposed cohort (incidence rate: 52.3 per 10,000 person-years; 95% CI: 51.7-52.9).This corresponded to an absolute difference of 14.3 per 10,000 personyears (95% CI: 11.9-16.6),with an HR of 1.26 (95% CI: 1.215-1.308).

Effects of healthcare history, age at infection, and antibiotic use on autoimmune diseases hospitalization due to M. pneumoniae infection
We examined the effects of use of antibiotics, oxygen therapy, and hospital visits on the association between hospitalization due to M. pneumoniae infection and autoimmune disease.The association was influenced by the prescription of antibiotics from 1 month before to 1 month after the index date, whereas the significance interaction was marginal.Oxygen therapy and frequency of outpatient visits within a month after the index date did not show an effect (Supplementary Table 7).

Multiple sensitivity analyses
The CDM comparison corroborated the observed associations (Figure 2) with higher estimates noted for autoimmune diseases (HR: 1.25; 95% CI: 1.06-1.49)although there were differences in statistical values among the hospitals.
The results of further sensitivity analysis yielded consisted results (Table 3).The association of hospitalization for M. pneumoniae infection with autoimmune disease was stronger for those with two or more autoimmune syndromes (HR: 1.506; 95% CI: 1.314-1.727;P for difference = 0.0011) and with three or more autoimmune syndromes (HR: 1.789; 95% CI: 1.178-2.716;P for difference <0.001) (Table 3).The sensitivity analysis also indicated greater risk when excluding cases diagnosed 2 years after study entry or 5 years after study entry; when exposure was defined more stringently (only ICD-10 code J157); and when the outcome was defined more stringently (diagnosis and treatment for autoimmune disease).In addition, when participants were restricted as individuals followed up for more than 5 years after entry, the exposed group (history of hospitalization for M. pneumoniae infection) had a higher cumulative incidence of autoimmune diseases compared with their matched unexposed group across the follow-up period (Supplementary Figure 3).

Discussion
In this large-scaled population-based cohort study, we observed a higher risk of autoimmune disease following hospitalization for M. pneumoniae infection.This association was particularly pronounced among patients who were older at the time of admission (as determined by diagnosis data) and whose index date was beyond 60 months.We found that the association persisted regardless of hospitalization duration, oxygen therapy, and antibiotic use.To the best of our knowledge, this is the first study to examine the relationship of hospitalization for M. pneumoniae infection with 19 distinct autoimmune in a large pediatric population, using a population-based comparison.By shedding light on this possible link, our study underscores the Risk estimates for the association of M. pneumonia-related hospitalization with different types of autoimmune diseases.HR, hazard ratio.Autoimmune diseases with fewer than 100 cases were not analyzed separately but were used for calculations of hazard ratios (HRs) for the main categories.Cox models were stratified by matching identifiers (birth year and sex) and adjusted for birth residence (Seoul/metropolitan, city, or rural area), household income (low, middle, or high), and perinatal history (disorder related to length of gestation and fetal growth, birth trauma, infections specific to the perinatal period, congenital malformation/deformation, and chromosomal abnormalities).The first year of follow-up was excluded from all analyses.a Autoimmune diseases with fewer than 100 cases were not analyzed separately but were included in calculations of hazard ratios (HRs) for the main categories.b Includes autoimmune vascular disorders, such as polyarteritis nodosa and Churg-Strauss syndrome.
need for continued research regarding the potential link between severe M. pneumoniae infection and autoimmune and immunemediated diseases (28,29).
Our findings also suggest that degree of association differed for the various autoimmune diseases examined in our study.For instance, the hazard ratio was 1.40 for juvenile idiopathic arthritis and 1.26 for autoimmune thyroid disease.These differences may reflect variations in the prevalence and pathogenicity of different autoimmune diseases in children Moreover, our results are consistent with previous studies that have reported associations between M. pneumoniae infection and autoimmune diseases such as Guillain-Barre's syndrome (GBS) (30,31) and polyarthritis (32).However, these previous studies were limited by a cross-sectional design, which precluded causal inferences.While our study provides compelling evidence for a potential link between severe infection and autoimmune and immune-mediated diseases, it is important to acknowledge that further research is needed to fully understand the mechanisms underlying this association (28).
Our findings are consistent with previous evidence that reported links of severe cases of M. pneumoniae infection and related respiratory diseases with immune system dysfunction and suggest that autoimmune reactions may contribute to some of the extrapulmonary complications in patients who have these infections (33).Previous evidence suggests that this may be attributed to immune responses related to cytokine activation and immune cell reaction due to severe infection.For example, inflammatory reflexes precipitated by M. pneumoniae infection might activate the sympathetic nervous system, possibly leading to impaired immune function or a widespread inflammatory response.Moreover, the adhesin proteins of M. pneumoniae (P1, P30, and P116) enable the adhesion of this pathogen to host epithelial cells, and this leads to the production of antibodies that target cells in the brain, liver, kidney, smooth muscle, and lungs that could induce autoimmune disorders in the affected organs (34,35).The release of cytokines and inflammatory mediators during M. pneumoniae infection can lead to worse disease outcomes and overactivation of the immune system (1,36).This supports our finding that patients hospitalized for M. pneumoniae infection were more likely to develop autoimmune and immune-mediated diseases involving different organs and multiple autoimmune syndromes.This might also explain the stronger association of infection with an autoimmune disease in children exposed after the age of 60 months, whose clinical course is more severe.Immune reactions induced by M. pneumoniae infections might occur because of the amino-acid sequence homology of mycoplasmal adhesins with antigens (P1, P30, and perhaps P65 and P116) in diverse human tissues and the formation of immune complexes (37).M. pneumoniae infections can activate T cells and increase B-cell proliferation, and these play important roles in the pathogenesis of extrapulmonary manifestations during pneumonia and autoimmune diseases (3).However, another study found no significant difference in the age of occurrence of extrapulmonary symptoms possibly related with immune-mediated reactions between M. pneumoniae infection and respiratory infections caused by M. pneumoniae in children.Caution is warranted in interpreting this finding due to the inherent limitations of epidemiological studies (38).
The strengths of our study are that we used a population-based cohort design and performed a complete follow-up of nearly 50,000 patients who were diagnosed with M. pneumoniae infections during a 10-year period.Furthermore, the large sample size allowed us to perform several sensitivity analyses, so we analyzed the effect of age at exposure and the use of multiple outcome measures.The availability of detailed sociodemographic and medical information also allowed us to control for a large number of potential confounders.
This study has several limitations.First, there may have been some surveillance bias.To address this issue, we performed sensitivity analyses using extended lag times, measured different outcomes, and adjusted for estimated levels of medical surveillance.Second, we relied on diagnoses of infection registered in the NHIS.This could have been an underestimation if infection occurred as a comorbidity of a different severe condition, if there was delayed identification of the pathogen, or if there was a missed diagnosis.Furthermore, since exposures included inpatient cases, the associations between Mycoplasma pneumoniae infection and autoimmune diseases may be limited to severe cases and may not be applicable to non-hospitalized infections.Third, since this is an observational retrospective cohort study, we could not exclude potential confounders that may affect the causal relationship because of the nature of this data.As an example, changes in epidemics over time such as age, sex, and peak season that might influence the pathogenesis of M. pneumoniae infection were not assessable.Furthermore, unmeasured factors such as unreported health conditions preceding an autoimmune disease (39), alterations in health-related behaviors, or the use of certain medications should be addressed in future studies.Fourth, although we identified a stronger association for older children, we did not assess the association between M. pneumoniae infections with autoimmune diseases that occurred later in life.

Conclusions
This nationwide population-based cohort study may indicate that exposure to M. pneumoniae infection requiring hospitalization may be associated with an increase in subsequent diagnoses of autoimmune diseases.Although our findings provide insight into the relationship between hospitalizations due to M. pneumonia infection and the increased occurrence of autoimmune disease, further studies are needed to better understand the underlying mechanisms.

FIGURE 1
FIGURE 1 Study design.Individuals born in South Korea from 2002 to 2005 were identified from the database of the National Health Insurance Service (NHIS) and linked information on death from Statistics Korea.After application of exclusion criteria, there were 49,937 children in the exposed cohort and 1,277,613 matched children in the unexposed cohort.a There were 265,430 born in 2002-2003, 233,940 born in 2004-2005.b The first year of follow-up was excluded for calculation of accumulated person-years.c Eligible unexposed individuals had no M. pneumonia infection and no autoimmune disease at the time of diagnosis of the matched index patient.Matching was performed using density sampling.There were 26,543 born in 2002-2003, and 23,394 born in 2004-2005.d Among individuals in the unexposed cohort, 10,420 (2.1%) who received diagnoses of M. pneumonia infections during follow-up were reallocated to the exposed group after diagnosis.e For the CDM, the exposed cohort included hospitalized patients from four hospitals without M. pneumonia infection and without autoimmune disease at the date of diagnosis of the index patient.

FIGURE 3
FIGURE 3 study was approved by the Institutional Review Board and

TABLE 1
Characteristics of the study cohorts.

TABLE 1 Continued
a The period between the index date and the occurrence of the outcome among children who experienced the outcome.bIncomewas categorized into tertiles based on the amount of insurance co-payment.cThefirst month after study entry was removed from this calculation.dComplicationsduring the intrauterine and perinatal period that were diagnosed 3 months prior to the index date (i.e., diagnosis date of the exposed participant or the diagnosis date of the index participant for the matched unexposed participant).eDefinedas one or more physician diagnosis during 6 months prior to the index date, with 1 or more event of asthma exacerbation requiring hospitalization or systemic corticosteroids.fPrescriptions of systemic macrolide antibiotics for treating M. pneumoniae infections during any month after the index date.gEventsduring 1 month after the index date.hThemedian length of hospitalization was 6 days.

TABLE 2
Risk of autoimmune disease in the exposed cohort (patients hospitalized with M. pneumoniae infection) relative to the matched unexposed cohort.

TABLE 2 Continued
Cox models were stratified by matching identifiers (birth year and sex) and adjusted for birth residence (Seoul/metropolitan, city, or rural area), household income (low, middle, or high), perinatal history (disorders related to fetal growth and development, birth injury, congenital malformations), and prescription duration of systemic macrolide antibiotics.The first year of followup was excluded from all analyses.cP value was from an interaction test by incorporating an interaction term into the Cox model.dComplications during the intrauterine and perinatal period that was diagnosed 3 months prior to the index date (i.e., the date of diagnosis of an exposed participant or the date of diagnosis of the index participant with the matched unexposed participant); history of fetal growth and development disorders, birth injury, infections specific to the perinatal period, and congenital malformations.eDefined as one or more physician diagnoses during 6 months prior to the index date, with one or more events of asthma exacerbation requiring hospitalization or systemic corticosteroids.Removed the first month after study entry from this calculation.Bold values indicate statistical significance at the p < 0.05 level.
a Results not shown if the number of subjects was less than 10. b f

TABLE 3
Sensitivity analysis of the association between M. pneumonia-related hospitalization and autoimmune disease.