Association of weight change patterns across adulthood with incident asthma: a retrospective cohort study

This study aimed to investigate the relationship between weight change patterns across adulthood and the risk of incident asthma later in life using data from the National Health and Nutrition Examination Survey (NHANES) 2001–2018. In this retrospective cohort study, asthma was defined by self-report questionnaires, and age at diagnosis was used to estimate the time of asthma onset. Based on BMI at 25 years old (young adulthood) and BMI at 10 years before the survey (middle adulthood), patterns of weight change were divided into five categories including stable normal, non-obese to obese, obese to non-obese, maximum overweight and stable obese. A total of 27,359 participants (female 13,582, 49.6%) were enrolled in this study and during a mean follow-up of 9.8 years, 1035 subjects occurred asthma. After adjusting for age, gender, race, education, family income and smoking status, participants changing from non-obese to obese, stable obese had significantly higher risks of incident asthma than those with normal weight during adulthood (HR1.70, 95% CI 1.35–2.15, P < 0.0001; HR 1.66, 95% CI 1.21–2.19 P = 0.0019, respectively). The findings suggested that maintaining normal weight during adulthood may be important for preventing incident asthma in later life.

Associations of weight change patterns and the risk of asthma. By using the stable normal group as reference, the stable obese group and non-obese to obese groups are associated with a higher risk of asthma, with HR (95% CI) of 1.66 (1.21, 2.29) and 1.70 (1.35, 2.15) in model 3, respectively (Table 2). By using the stable obese group as the reference, the stable normal group significantly reduced the risk of asthma with HR (95% CI) of 0.60 (0.44, 0.83) as presented in Table 3. Supplementary Figure S3 presented the adjusted cumulative hazard curves of asthma for each weight change pattern, and weight change from non-obesity to obesity had the highest cumulative incidence. Table 1. Baseline characteristics of study participants in the NHANES 2001-2018 according to weight change patterns during adulthood. All estimates accounted for complex survey designs and the analyses used the sample weights, stratification and clustering suggested by the CDC. Twenty-three, 2298, and 15 participants had missing information for baseline education level, family income-poverty ratio, and smoking status respectively. www.nature.com/scientificreports/ Since the maximum overweight was heterogeneous, we further analyzed this relationship in participants with overweight at age 25 or 10 years before baseline, which is presented in Table 4. The stable overweight group and weight change from overweight to obesity had higher risk of asthma (HR 1.48, 95% CI 1.07-2.04 and HR 1.47, 95% CI 1.03-2.08) compared with the stable normal group.
The association between the absolute weight change and the risk of asthma is shown in Table 5 and Fig. 2. We also found a non-liner relationship between absolute weight change and the risk of asthma. As a continuous variable, a per 5 kg increment in weight change was associated with a 6% higher risk of asthma (HR 1.06, 95% CI 1.03, 1.09). As a categorical covariate, individuals with weight gain ≥ 20 kg had a higher risk of asthma (HR 1.45, 95% CI 1.11-1.89) compared with individuals with weight change within 2.5 kg. The adjusted cumulative hazard curves of asthma for each absolute weight change group are presented in Supplementary Fig. S4. The cumulative incidence of weight gain ≥ 20 kg was the highest obtained.  Table 2. Hazard ratios (HRs) and 95% confidence intervals (CIs) of incident asthma with weight change patterns across adulthood in NHANES 2001-2018 used the stable normal category as a reference. All estimates accounted for complex survey designs and the analyses used the sample weights, stratification and clustering suggested by the CDC. Significant results are in boldface type. Model 1: Non-adjusted model. Model 2: Adjusted for age at ten years before baseline, gender, race/ethnicity (Mexican American, non-Hispanic White, non-Hispanic Black, and other race). Model 3: Adjusted for age at ten years before baseline, gender, race/ ethnicity (Mexican American, non-Hispanic White, non-Hispanic Black, and other race), baseline education level (less than high school, high school or equivalent, college or above), baseline family income (family income-poverty ratio, 0-1.0, 1.1-3.0, > 3.0), baseline smoking status (ever smoke and never smoke).  Table 6. In each subgroup, non-obese to obese group was associated with a higher risk of asthma. Stable obesity significantly increased the risk of asthma in the female subgroup, baseline age < 50 years old subgroup and non-smoking subgroup. The results of interaction analyses Table 3. Hazard ratios (HRs) and 95% confidence intervals (CIs) of incident asthma with weight change patterns across adulthood in NHANES 2001-2018 used the stable obesity category as a reference. All estimates accounted for complex survey designs and the analyses used the sample weights, stratification and clustering suggested by the CDC. Significant results are in boldface type. Model 1: Non-adjusted model. Model 2: Adjusted for age at 10 years before baseline, gender, race/ethnicity (Mexican American, non-Hispanic White, non-Hispanic Black, and other race). Model 3: Adjusted for age at 10 years before baseline, gender, race/ ethnicity (Mexican American, non-Hispanic White, non-Hispanic Black, and other race), baseline education level (less than high school, high school or equivalent, college or above), baseline family income (family income-poverty ratio, 0-1.0, 1.1-3.0, > 3.0), baseline smoking status (ever smoke and never smoke).   Table 5. Hazard ratios (HRs) and 95% confidence intervals (CIs) of incident asthma with absolute weight change patterns across adulthood. All estimates accounted for complex survey designs and the analyses used the sample weights, stratification and clustering suggested by the CDC. Significant results are in boldface type. Model 1: Non-adjusted model. Model 2: Adjusted for age at 10 years before baseline, gender, race/ethnicity (Mexican American, non-Hispanic White, non-Hispanic Black, and other race). Model 3: Adjusted for age at 10 years before baseline, gender, race/ethnicity (Mexican American, non-Hispanic White, non-Hispanic Black, and other race), baseline education level (less than high school, high school or equivalent, college or above), baseline family income (family income-poverty ratio, 0-1.0, 1.1-3.0, > 3.0), baseline smoking status (ever smoke and never smoke).  . General additive models demonstrate the relationship between absolute weight change across adulthood and the risk of asthma. The resulting figures show the predicted log (relative risk) in the y-axis and the absolute weight change in the x-axis. The model was adjusted for gender, age at ten years before baseline, race/ethnicity, baseline family income, baseline education levels, and baseline smoking status. Table 6. Hazard ratios (HRs) and 95% confidence intervals (CIs) of incident asthma with weight change patterns across adulthood stratified by baseline age, gender and smoking status. All estimates accounted for complex survey designs and the analyses used the sample weights, stratification and clustering suggested by the CDC. Risk estimates were adjusted for gender (not adjusted in stratified analysis by gender), age at 10 years before baseline (not adjusted in stratified analysis by age), race/ethnicity, baseline family income, baseline education, and baseline smoking status (not adjusted in stratified analysis by smoking status). Significant results are in boldface type. www.nature.com/scientificreports/ that changing from non-obese to obese, and stable obese was also associated with a higher risk of asthma (HR 1.71, 95% CI 1.34-2.18 and HR 1.65, 95% CI 1.19-2.28, respectively), which is consistent with the results in Table 2.

Number of events
We separately investigated the relationship between weight change in the underweight group at age 25 and the risk of asthma. There were 1374 underweight participants at 25 years old, accounting for 5% of the total participants. Two hundred twenty of them maintained underweight at middle age (including nine cases of asthma), 825 (including 27 cases of asthma) of them became normal weight, 227 (6 cases of asthma) of them became overweight, and 102 of them (5 cases of asthma) became obese. Further, there was no significant increase in the risk of asthma in the underweight group compared with those who maintained normal weight (Supplementary  Table S5).

Discussion
In this study we used data from nine waves of the NHANES survey to investigate the relationship between weight change patterns and the risk of asthma. The results suggest weight change patterns across adulthood are significantly associated with the risk of asthma, with participants who enter young adulthood with a normal BMI and continue to maintain a normal BMI during middle adulthood having the lowest risk of asthma. Furthermore, participants who are stable overweight, stable obese and those who gain weight from non-obese to obese are associated with an increased risk of asthma.
Numerous studies have found adults with overweight and obesity are associated with a higher risk of asthma than adults with normal weight [16][17][18][19][20][21] . Recently, results from two studies (using Norwegian and European populations) suggest a causal relationship between BMI and the risk of asthma in adults by using bi-directional Mendelian randomized analysis, with the risk of asthma significantly increased with the increase in BMI 22,23 . These results are consistent with our findings. In this study, we found that overweight or obese participants in age 25 or middle age (age 10 years before baseline) had a significantly increased risk of asthma in later life.
Another finding of this study is that weight changes are common across adulthood. Only a few previous studies have examined the relationship between weight change and the risk of asthma in adults, and the results are controversial. The Nurses' Health Study II (1596 cases/85,911 participants, using a four-year follow-up) and the E3N Cohort Study (372 cases /67,299 participants, using a 3-year follow-up) found weight gain significantly increased the risk of developing asthma 24,25 , which are consistent with our findings. While other studies are inconsistent with our results, the Zurich Cohort Study which included 591 participants with 20-years cumulative asthma incidences (8.1%) and the Finnish Twin Cohort which included 10,597 participants with 9-years cumulative asthma incidences (1.2%) found no association with weight gain or obesity and later asthma in both men and women 17,26 . Reasons for those inconsistent results may include differences in baseline age, differences in sample size and differences in the definition of weight change. In sum, most of those studies focused on a specific population and some included only women, so the results may not be generalizable to whole populations. Our study used nationally representative data, which complemented the previous studies, and the results also confirm that weight gain across adulthood presents higher risk for asthma.
Two previous articles have reported the association between weight change patterns and asthma in adults. Kim et al. used the KNHANES data with 29,936 participants who developed asthma after a 3-year follow-up in 245,593 adults and found that weight gain increased the risk of asthma compared with maintaining a normal weight, again consistent with our results. They also reported that no association was found between weight gain from normal or underweight to obese and the risk of asthma 14 , which appears inconsistent with our results. The inconsistency might be due to two reasons. First, the two studies observed weight change over different periods of time. Our study assessed weight change across adulthood with an average period of 25 years, whereas their study assessed weight change for only three years between 2005 and 2008. Secondly, they used different reference groups, included different number of cases and there were also some differences in the definitions of weight change patterns 14 . Wang et al. found the risk of developing asthma increased 1.61 and 1.43 in the non-obese to obese group and the stable obese group, respectively by using the data from NHANES (1999-2016) 15 . Their study focused on obese participants. However, there were more overweight participants than obese ones. Of the 27,359 participants, 20.75% were overweight at age 25, rising to 36.66% at 10 years before baseline. There were 10,031 overweight participants and 6510 obese participants (36.66% versus 27.40%) at 10 years before the survey (middle adulthood). We further analyzed the association between weight change and the risk of asthma in overweight participants. The incidences of asthma are also significantly higher among these participants who were stable overweight and changed from overweight to obese in early adulthood. Therefore, we emphasize that weight changes not only in obese participants, but also in overweight participants are significantly associated with the risk of asthma, which complemented the previous study.
This study found no association between absolute weight loss and incident asthma, which is consistent with previous results 24,25,27 . This study also found weight loss from obese to non-obese was not significantly associated with the risk of asthma, possibly due to a relatively small sample size as only 0.9% participants lost weight from obesity to normal. Thus, we should interpret these results with caution.
Obesity and excess fat accumulation in the body may promote the development and progression of asthma through three pathways, including inflammatory, metabolic, and mechanistic pathways. Animal research suggested that pro-inflammatory factors increased in bronchoalveolar lavage fluid of high-fat diet-ovalbumin mice (HFD-OVA) compared with lean mice 7 . Other research found that obesity alters the lung function of participants by reducing the movement of the diaphragm to increase airway restriction 28 . In sum, in the inflammatory and metabolic pathways, adipose tissue promotes airway obstruction and affects lung function by secreting excessive pro-inflammatory factors IL-6, IL-1β, TNFα, and TGFβ, while the mechanical pathways affect lung function by increasing airway restriction that promotes asthma 7 www.nature.com/scientificreports/ In this study, we used a large and nationally representative sample, analyzed weight status at each time point, weight change patterns, absolute weight change, and the risk of asthma. In addition, we analyzed the association between weight change patterns and the risk of asthma in underweight, normal weight, overweight, and obese participants in detail. These are the strengths of this study. Also, considering that some potential factors (smoking, family income and education) could affect the results, we used different models to analysis the relationship and further adjusted these confounding factors on the basis of model 2. The result of the analysis showed that compared with the unadjusted model, the associations between body weight change and the risk of asthma did not change in the adjusted models.
However, there are some limitations in the current study. First, the self-reported weight rather than measured weight was used in this study. Though a high degree of consistency between self-reported weight and measured weight has been demonstrated in previous articles, some misclassification has likely occurred 30,31 . And we used self-reported scale instead of lung function to identify participants with asthma, which may also lead to misclassification. Second, the sample size of weight loss from youth to middle age was small, which may limit the precision of our estimates; Third, we could not adjust for some confounding variables including dietary factors, alcohol consumption, and physical activity because these data at 10 years before baseline were not collected. Lastly, we did not analyze the association between the change in other obesity related markers, such as fat mass and waist circumference and incident asthma, because that data was not collected at the two-time point intervals.

Conclusions
In conclusion, our investigation indicates that in a sample of U.S. adults, stable obesity, adult with overweight and weight gain from normal or overweight to obese during adulthood are associated with increased risks of asthma. Maintaining normal weight during adulthood had the lowest risk of asthma. We found no significant association between weight loss and asthma. This may be due to small sample size, so more studies are needed to assess this association. In addition, we emphasized that weight changes not only in obese participants but also in overweight participants are associated with the risk of asthma, and weight management in these participants should be emphasized.

Material and methods
Data source. All data used in this study for analysis were obtained from the NHANES (2001-2018), which was conducted by the National Center for Health Statistics of the Centers for Disease Control and Prevention (CDC). NHANES is a series of ongoing cross-sectional surveys and used a stratified multistage probability sample design to collect health-related data of nationally representative samples of the US population [32][33][34][35] .
All participants signed the informed consent forms during the period of recruitment, and all protocols were approved by the research ethics review board of the National Center for Health Statistics of CDC. We confirmed that all methods were performed in accordance with relevant guidelines and regulations.
Study design and population. We collected data on adults aged 40 or over at the time of the survey from the nine cycles of the consecutive NHANES 2001-2018. We defined the time at which participants had a physical examination as the baseline, and weight change was estimated from participants' self-reported weight at age 25 (weight age25 ) and weight at 10 years before baseline (weight 10prior ). We defined incident asthma from a self-administered questionnaire which was completed at the clinic visit. We used the reported age at diagnosis to establish the time of incident asthma. The present study design is shown in Figure S1 and the design of NHANES has been shown in previous studies 36-38 . Referring to several recent studies, the exclusion criteria were as follows 36,37 : participants who were younger than 40 years old; participants who did not have asthma diagnostic information and diagnosis time of asthma; participants who did not have data on BMI at age 25 (BMI age25 ) and BMI 10 years before baseline (BMI 10prior ); and participants whose asthma onset date was before the initiation of a 10-year follow-up. Finally, we enrolled 27,359 participants in the current study.
Weight change measures and covariates. Weight age25 , weight 10prior and height at age 25 were recalled when participants were interviewed. Measured height at physical examination is used to calculate BMI 10prior . When the participant was < 50 years old, it was also used to calculate BMI age25 . But when the participant was ≥ 50 years of age, height at age 25 was used to calculate BMI age25 .
Referring to several recent studies 36,37 , according to BMI age25 and BMI 10prior , we defined five patterns of weight change which are shown in Supplementary Table S1, including stable normal (both BMI < 25), non-obese to obese (BMI age25 < 30 and BMI 10prior ≥ 30), obese to non-obese (BMI age25 ≥ 30 and BMI 10prior < 30), maximum overweight (25 ≤ BMI ≤ 29.9 at age 25 or ten years before baseline, but BMI not ≥ 30 at the other time), and stable obese (both BMI ≥ 30). Age 10 prior ranged from 30 to 75 years old. Those whose weight change patterns were from age 25 to age 30 accounted for 2.7% of the total participants. To compare with previous studies 14,25 , we also further analyzed the association between absolute weight change during this period and incident asthma. We calculated the absolute weight change by subtracting weight age25 from weight 10prior. Then, we divided the absolute weight changes into five groups: weight loss > 2.5 kg, weight change within 2.5 kg (the absolute weight loss or weight gain ≤ 2.5 kg), 2.5 kg < weight gain ≤ 9.9 kg, 10 kg ≤ weight gain ≤ 19.9 kg, and weight gain ≥ 20.0 kg, weight change within 2.5 kg was used as the reference group. We chose these categories for comparison with the previous studies 15, [36][37][38] .
Covariate data were obtained from demographic and questionnaire surveys, including gender, baseline age, ethnicity (Mexican American, non-Hispanic White, non-Hispanic Black, and other race), baseline smoking status, baseline education level (less than high school, high school or equivalent, college or above) and baseline www.nature.com/scientificreports/ family income. Poverty income ratio (0-1.0, 1.1-3.0, > 3.0) was used to reflect family income. Smoking status included never smoking and ever smoke defined as "smoked at least 100 cigarettes in lifetime".
Statistical analysis. According to NHANES analytic guidelines, we analyzed the data using a weight suggested by the CDC to account for the complex survey design (including oversampling). The categorical variables are presented as percentages and the continuous variables are expressed as means ± SD or medians [interquartile range (IQR)]. All the variables were compared among different groups by the weighted chi-square tests when they were categorical variables, and by the weighted linear regression models when they were continuous variables. We assessed BMI at each time point, the weight change pattern and the absolute weight change during adulthood and risk of asthma respectively. First, we used Cox Proportional Hazards Regression Analyses to evaluate the association of BMI at each time point (BMI age25 and BMI 10prior ) and the risk of asthma. We divided the participants into six groups: underweight, normal weight, overweight, class 1 obesity, class 2 obesity, class 3 obesity and we used normal weight as a reference.
Second, we evaluated the association of weight change patterns and incident asthma. To assess the increased risk of incident asthma, the stable normal category was used as a reference to compare with all other categories. On the contrary, the stable obesity category was used as a reference to compare with all other categories to assess the decreased risk. In this study, we used three sequential models to control for potential confounders. Model 1 is a non-adjusted model. Model 2 was adjusted for age, gender and race. Model 3 was adjusted for age, gender, race, education, family income and smoking status. Because 2298 participants missed information regarding family income, which was more than 10% of total participants, we set these mission values as another level of the categorical covariate to be adjusted in Model 2. Twenty-three and 15 participants missed the data of baseline education levels and smoking status which represented less than 1% of total participants. We continued our analysis without making any further adjustments. We used the survival curves of weight change patterns to test the proportional hazards assumption and no significant deviation from proportionality in hazards over time was detected. Then, we performed the subgroup analysis stratified by baseline age, gender, and smoking status to determine whether there were differences among each subgroup. We divided the participants into two groups by baseline age at 50 years old. We also performed the interaction tests between age, gender, smoking status and weight change patterns for the risk of asthma. Gender, smoking status and weight change patterns were treated as nominal variables. Age group was treated as an ordinal variable and the median of each category was used for analysis. The adjusted log likelihood ratio test was used to analyze the interaction.
Third, we evaluated the association of absolute weight change and incident asthma adjusted for the same covariates. Firstly, we performed the analysis by taking absolute weight change as a continuous variable. Secondly, we took the absolute weight change as a categorical covariate and compared with all other categories. We used body weight change within the 2.5 kg category as a reference. Furthermore, we used generalized linear models with a logit link to evaluate the nonlinearity between them.
The Cox Proportional Hazards Regression Analyses was used to conduct the following sensitivity analyses. First, we excluded participants with missing variables. Second, to reduce the effect of underweight, we excluded underweight participants. Third, considering the heterogeneity of the maximum overweight group, which includes participants with weight change from normal to overweight, from overweight to normal, and stable overweight, we divide those participants into three groups according to BMI age25 and BMI 10prior. Finally, we analyzed the association between the weight change in underweight group at age 25 and the risk of asthma.
We used the EmpowerStats software (www. empow ersta ts. com version R.3.4.3) and statistical software package R to process and analyze all our data. A two-tailed P value < 0.05 was considered statistically significant.

Data availability
The data are available on the National Health and Nutrition Examination Survey website (https:// www. cdc. gov/ nchs/ nhanes/). www.nature.com/scientificreports/ Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.