Abstract
Asthma is a common chronic respiratory disease related to oxidative stress. Oxidative balance score (OBS) could assess systemic oxidative stress status. Thus, we tried to explore the prediction value of OBS in asthma and the disease course. The data were obtained from the National Health and Nutrition Examination Survey database. Asthma and the disease course were determined by the Patient Health Questionnaire. OBS was scored by 20 dietary and lifestyle components. The receiver operating characteristic and decision curve analysis were used to assess the prediction value of OBS. Logistic regression, XG Boost, and Random Forest methods were used to obtain an optimal OBS-based model and rank the importance of OBS components. Mediation analysis was used to explore the possible interplay of OBS components on the disease course of asthma. From 2011 to 2018, 7348 participants including 6597 participants without asthma and 751 participants with asthma were enrolled. Receiver operating characteristic and decision curve analysis curves exhibited that the OBS-based model showed an improved prediction value than the OBS for the disease course of asthma. Machine learning techniques results showed that the body mass index, niacin, and selenium were the key components of OBS. Besides, niacin had a direct relation with the disease course and could also regulate the course of asthma by regulating body mass index. OBS could predict the disease course of asthma, and niacin may be the most important component of OBS in the development of asthma.
Keywords: asthma, machine learning, oxidative balance score, the disease course
1. Introduction
Asthma, a prevalent chronic respiratory ailment,[1] afflicts an estimated 300 million individuals globally according to the World Health Organization Statistics.[2] As a multifaceted respiratory disease, the occurrence and development of asthma are intricately influenced by numerous factors. Among these, environmental factors have a significant impact on asthma.[3] Air pollution, pernicious indoor agents, pollen, and dust mites are known to incite or exacerbate asthma attacks.[4] Furthermore, genetic determinants play a pivotal role in the evolution of asthma. There is a correlation between a family history of the ailment, genetic variations, and an individual’s susceptibility to asthma.[5,6] Additionally, lifestyle choices, including dietary habits, physical activity, and tobacco use have been closely associated with the risk of developing asthma.[7,8] Previous investigations have identified certain factors linked to asthma but with controversial results. Thus, comprehending the risk factors associated with asthma holds paramount significance in its prevention, diagnosis, and treatment.
Oxidative stress emerges as a pivotal factor in asthma pathogenesis.[9] Oxidative stress is a state of cellular oxidative damage due to the excessive accumulation of reactive oxygen species (ROS) within the intracellular and extracellular environments.[10] Remarkably, patients afflicted by asthma tend to have abnormal oxidative stress responses, which may result in excessive production of ROS and induce an imbalance between oxidant production and elimination. This, in turn, sets the stage for inflammatory cascades, airway remodeling, and an array of pathophysiological alterations for asthma.[11] Conversely, certain investigations hypothesize that the onset of asthma is not directly related to oxidative stress, but is the result of complex interactions with other factors, such as allergies and inflammation.[12]
In recent times, scholars have proposed the concept of an oxidative balance score (OBS), a composite index encompassing various markers of oxidative stress that quantitatively assesses the oxidative balance within the human body.[13] Research on OBS in clinical diseases has received considerable attention. Notably, in the investigation of chronic ailments, OBS has emerged as an important biomarker to evaluate oxidative stress levels and predict disease susceptibility.[14–16] Nevertheless, the precise association between OBS and asthma remains elucidated. Hence, the primary objective of this study is to delve deeper into the intricate correlation between OBS and asthma, while assessing its potential utility in disease diagnosis and prevention. Therefore, we used a comprehensive research design involving a large-scale and multicenter approach to collect and analyze relevant clinical data and biological specimens related to OBS. By comparing OBS levels between patients with asthma and healthy controls, we aim to reveal the underlying mechanisms and influential factors governing OBS in asthma.
2. Methods
2.1. Data sources and study population
The National Health and Nutrition Examination Survey (NHANES, https://wwwn.cdc.gov/nchs/nhanes/Default.aspx) is a program of studies designed to assess the health and nutritional status of adults and children in the United States. Besides, the participants in the NHANES from 2011 to 2018 were included in this study, and all participants provided written informed consent and the protocol was approved by the NCHS Ethics Review Board[17] (Protocol #2011-17, Continuation of Protocol #2011-17, Continuation of Protocol #2011-17, Continuation of Protocol #2011-17, Protocol #2018-01). Thus, The Ethics Committee of Longyan First Affiliated Hospital of Fujian Medical University deemed that this research is based on open-source data, so the need for ethics approval was waived. The study flow chart is shown in Figure 1.
Figure 1.
Sample selection process flow chart.
2.2. Outcomes
Based on the answers to the questions “Ever been told you have asthma,” and “ Do you still have asthma,” we defined the participants who answered “yes” as “participates with asthma.” Besides, based on the answers to the questions “How old when you were first told you had asthma” and their age, we calculated the disease course for participants who still have asthma. The disease course greater or equal to 7 years was defined as a “long disease course “, while the disease course <7 years was defined as a “short disease course.”
2.3. OBS calculation
The OBS was calculated by combining information from both diet and lifestyle factors, which covered the levels of 4 lifestyle OBS components: physical activity (PA), body mass index (BMI), alcohol, and cotinine; and 16 dietary nutrient components: dietary fiber, carotene, riboflavin, niacin, vitamin B6, total folate, vitamin B12, vitamin C, vitamin E, calcium, magnesium, zinc, copper, selenium, total fat and iron. The PAQ survey within the NHANES database encompasses various categories of PA, including vigorous/moderate work-related activity (metablic equivalent [MET] = 8/4), walking or bicycling (MET = 4), and vigorous/moderate leisure-time physical activity (MET = 8/4). The value of PA was determined using the formula PA (MET-min/wk) = MET × weekly frequency × duration of each PA. Serum cotinine reflects smoking and the information on alcohol consumption was collected from the 24-hour dietary recall interviews. The dietary intake data were also obtained from the 24-hour dietary recall interviews.
The assignment scheme of OBS is shown in Table S1, Supplemental Digital Content, https://links.lww.com/MD/O776, the scores of each component were summed together to yield an overall OBS. For alcohol consumption, heavy drinkers (≥30 g/d for males and ≥ 15 g/d for females), non-heavy drinkers (0–30 g/d for males and 0–15 g/d for females), and nondrinkers received 0, 1, and 2 points, respectively. The other 19 components were categorized into 3 groups according to sex-specific tertiles. The antioxidants were assigned points from 0 to 2 for the tertile 1 group to the tertile 3 group. For pro-oxidants, the scoring was 2 for the tertile 1 group, 1 for the tertile 2 group, and 0 for the tertile group.
2.4. Covariates
According to previous studies, some potential confounders for asthma were selected for further analysis: relatives with asthma, household smokers, race, age, gender, infectious status, poverty income ratio (PIR), and sedentary time. The participant age chosen for this study was at the time of screening. Sex was dichotomized into male and female. Race was classified as Hispanic, Non-Hispanic White, Non-Hispanic Black, and other races. Based on the Department of Health and Human Services poverty guidelines, the PIR was calculated by dividing the median household income by the poverty line.
2.5. Statistical analysis
All statistical analyses were performed using SPSS 25 and R software (version 4.13). Continuous variables in baseline data were presented as median [interquartile range] due to non-normal distribution; categorical variables were presented as frequencies (percentages). Differences between continuous variables were assessed using the Mann–Whitney U test. Differences between categorical variables were analyzed using the chi-square test. P < .05 was considered statistically significant.
Receiver operating characteristic (ROC) and decision curve analysis (DCA) analyses were performed to evaluate the predictive value of OBS using Proc and ggDCA R packages. The DeLong test was adopted to analyze the significant differences in the area under the curve (AUC). To improve the clinical value of OBS, we constructed an OBS-based model using machine learning algorithms, including Random Forest, XG Boost, and logistic regression models to select the optimal model. Then, the importance of OBS components was ranked by using the above 3 algorithms. Venn analysis was used to select the common key components. Mediation analysis was performed in R software using the mediation R package.
3. Results
3.1. Characteristics of participants
The baseline features of participants are summarized in Table 1. There were 7348 participants included in the final analysis, including 6597 participants without asthma and 751 participants with asthma. There were significant differences in age, gender, race, infection status, PIR, sedentary time, OBS score, and OBS components except dietary fiber, calcium, alcohol, and carotene between the 2 subgroups. The participants with asthma tended to be elderly, male, non-Hispanic Black, had lower PIR, lower sedentary time, higher OBS scores, and higher content of OBS components, and were more likely to be infected and obese (all P < .05).
Table 1.
Baseline characteristics of participants with or without asthma.
| Variables | Group | Total (n = 7348) | Without asthma (n = 6597) | With asthma (n = 751) | Statistics | P |
|---|---|---|---|---|---|---|
| Relatives with asthma | Yes | 1414 (32.085) | 1043 (27.021) | 371 (67.824) | 366.068 | <.001 |
| No | 2993 (67.915) | 2817 (72.979) | 176 (32.176) | |||
| Race | Other Hispanic | 2266 (30.838) | 2089 (31.666) | 177 (23.569) | 94.183 | <.001 |
| Non-Hispanic White | 2052 (27.926) | 1878 (28.467) | 174 (23.169) | |||
| Non-Hispanic Black | 1842 (25.068) | 1545 (23.420) | 297 (39.547) | |||
| Other race | 1188 (16.168) | 1085 (16.447) | 103 (13.715) | |||
| Household smokers | Yes | 252 (11.650) | 218 (11.342) | 34 (14.108) | 1.591 | .207 |
| No | 1911 (88.350) | 1704 (88.658) | 207 (85.892) | |||
| Infection status | Yes | 381 (5.215) | 326 (4.970) | 55 (7.363) | 7.766 | .005 |
| No | 6925 (94.785) | 6233 (95.030) | 692 (92.637) | |||
| Gender | Male | 3699 (50.340) | 3262 (49.447) | 437 (58.189) | 20.614 | <.001 |
| Female | 3649 (49.660) | 3335 (50.553) | 314 (41.811) | |||
| Age | 7.000 [3.000, 10.000] | 7.000 [3.000, 10.000] | 8.000 [5.000, 11.000] | ‐7.758 | <.001 | |
| Poverty income ratio | 1.310 [0.720, 2.550] | 1.350 [0.730, 2.600] | 1.150 [0.650, 2.010] | 4.844 | <.001 | |
| Sedentary time | 7.000 [5.000, 7.000] | 7.000 [5.000, 7.000] | 7.000 [4.000, 7.000] | 2.466 | .007 | |
| Dietary fiber (g) | 12.350 [9.150, 16.300] | 12.300 [9.100, 16.200] | 12.550 [9.250, 16.950] | ‐1.587 | .113 | |
| Total fat (g) | 59.785 [45.035, 78.255] | 59.180 [44.590, 77.325] | 65.140 [49.890, 86.835] | ‐6.692 | <.001 | |
| Riboflavin (mg) | 1.717 [1.318, 2.190] | 1.710 [1.319, 2.182] | 1.788 [1.311, 2.268] | ‐1.97 | .049 | |
| Niacin (mg) | 17.873 [13.352, 23.419] | 17.655 [13.134, 23.118] | 19.973 [15.125, 25.643] | -7.834 | <.001 | |
| Vitamin B6 (mg) | 1.472 [1.104, 1.935] | 1.452 [1.097, 1.915] | 1.618 [1.184, 2.075] | -5.585 | <.001 | |
| Total folate (mcg) | 309.000 [223.500, 417.500] | 306.000 [221.500, 416.500] | 325.500 [240.500, 434.000] | -3.462 | <.001 | |
| Vitamin B12 (mcg) | 4.035 [2.850, 5.520] | 4.005 [2.835, 5.475] | 4.295 [2.970, 5.855] | -2.986 | .003 | |
| Vitamin C (mg) | 65.650 [36.400, 103.550] | 64.850 [35.700, 102.750] | 72.000 [42.000, 109.700] | -3.279 | .001 | |
| Vitamin E (ATE) (mg) | 5.700 [4.140, 7.810] | 5.640 [4.105, 7.740] | 6.190 [4.560, 8.630] | -5.354 | <.001 | |
| Calcium (mg) | 908.500 [671.500, 1196.000] | 907.500 [672.500, 1194.500] | 912.500 [652.000, 1218.000] | -0.016 | .988 | |
| Magnesium (mg) | 205.500 [162.000, 256.500] | 204.000 [162.000, 255.000] | 217.500 [164.000, 269.000] | -3.647 | <.001 | |
| Iron (mg) | 11.940 [8.740, 15.910] | 11.840 [8.645, 15.795] | 12.850 [9.775, 16.835] | -5.186 | <.001 | |
| Zinc (mg) | 8.345 [6.280, 10.960] | 8.285 [6.245, 10.865] | 9.245 [6.540, 11.750] | -4.885 | <.001 | |
| Copper (mg) | 0.782 [0.602, 1.005] | 0.777 [0.598, 0.996] | 0.823 [0.628, 1.074] | -4.192 | <.001 | |
| Selenium (mcg) | 80.800 [60.500, 104.250] | 80.050 [59.950, 103.000] | 87.550 [66.450, 115.900] | -6.044 | <.001 | |
| Alcohol (g) | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.016 | .862 | |
| Body mass index (kg/m2) | 17.100 [15.600, 20.100] | 17.000 [15.600, 19.900] | 18.100 [15.900, 22.500] | -6.492 | <.001 | |
| PA (MET-min/wk) | 1860.000 [840.000, 3600.000] | 1800.000 [800.000, 3600.000] | 1920.000 [880.000, 3480.000] | -0.217 | .829 | |
| Cotinine (ng/mL) | 0.030 [0.011, 0.164] | 0.028 [0.011, 0.148] | 0.062 [0.011, 0.362] | -5.974 | <.001 | |
| Carotene (RE) | 48.292 [22.375, 128.563] | 48.521 [22.625, 130.813] | 45.750 [20.542, 108.771] | 1.904 | .057 | |
| Oxidative balance score | 16.000 [12.000, 21.000] | 16.000 [12.000, 21.000] | 17.000 [12.000, 22.000] | -2.056 | .04 |
MET = metablic equivalent, PA = physical activity.
In addition, participants with longer disease courses tended to be older, had lower sedentary time, higher dietary fiber, total fat, niacin, vitamin B6, total folate, vitamin E, magnesium, iron, zinc, copper, selenium, OBS score, and were more likely to be obese (P < .05) (Table 2).
Table 2.
Baseline characteristics of participants between different disease course groups.
| Variables | Group | Total (n = 7348) | Short disease course (n = 506) | Long disease course (n = 245) | Statistics | P |
|---|---|---|---|---|---|---|
| Relatives with asthma | Yes | 371 (67.824) | 196 (64.474) | 175 (72.016) | 3.521 | 0.061 |
| No | 176 (32.176) | 108 (35.526) | 68 (27.984) | |||
| Race | Other Hispanic | 177 (23.569) | 120 (23.715) | 57 (23.265) | 1.438 | 0.697 |
| Non-Hispanic White | 174 (23.169) | 113 (22.332) | 61 (24.898) | |||
| Non-Hispanic Black | 297 (39.547) | 199 (39.328) | 98 (40.000) | |||
| Other race | 103 (13.715) | 74 (14.625) | 29 (11.837) | |||
| Household smokers | Yes | 34 (14.108) | 25 (14.286) | 9 (13.636) | 0.017 | 0.897 |
| No | 207 (85.892) | 150 (85.714) | 57 (86.364) | |||
| Infection status | Yes | 55 (7.363) | 40 (7.921) | 15 (6.198) | 0.712 | 0.399 |
| No | 692 (92.637) | 465 (92.079) | 227 (93.802) | |||
| Gender | Male | 437 (58.189) | 301 (59.486) | 136 (55.510) | 1.073 | 0.300 |
| Female | 314 (41.811) | 205 (40.514) | 109 (44.490) | |||
| Age | 8.000 [5.000, 11.000] | 6.000 [4.000, 9.000] | 11.000 [9.000, 12.000] | -15.719 | <0.001 | |
| Poverty income ratio | 1.150 [0.650, 2.010] | 1.080 [0.620, 2.010] | 1.180 [0.740, 1.990] | -1.500 | 0.134 | |
| Sedentary time | 7.000 [4.000, 7.000] | 7.000 [5.000, 7.000] | 5.000 [3.000, 7.000] | 2.543 | 0.008 | |
| Dietary fiber (g) | 12.550 [9.250, 16.950] | 12.350 [9.000, 16.500] | 13.550 [9.950, 17.850] | -2.766 | 0.006 | |
| Total fat (g) | 65.140 [49.890, 86.835] | 61.125 [48.085, 83.530] | 74.600 [55.465, 94.820] | -4.022 | <0.001 | |
| Riboflavin (mg) | 1.788 [1.311, 2.268] | 1.763 [1.303, 2.221] | 1.834 [1.362, 2.373] | -0.830 | 0.407 | |
| Niacin (mg) | 19.973 [15.125, 25.643] | 18.920 [14.676, 24.249] | 22.470 [16.685, 27.697] | -4.976 | <0.001 | |
| Vitamin B6 (mg) | 1.618 [1.184, 2.075] | 1.555 [1.152, 2.046] | 1.726 [1.292, 2.132] | -2.563 | 0.010 | |
| Total folate (mcg) | 325.500 [240.500, 434.000] | 318.000 [231.500, 415.000] | 349.000 [271.500, 463.000] | -3.306 | <0.001 | |
| Vitamin B12 (mcg) | 4.295 [2.970, 5.855] | 4.120 [2.970, 5.845] | 4.575 [2.960, 5.955] | -1.324 | 0.186 | |
| Vitamin C (mg) | 72.000 [42.000, 109.700] | 72.300 [43.100, 113.150] | 70.800 [39.300, 106.300] | 1.378 | 0.168 | |
| Vitamin E (ATE) (mg) | 6.190 [4.560, 8.630] | 6.025 [4.435, 8.335] | 6.575 [4.800, 9.400] | -2.638 | 0.008 | |
| Calcium (mg) | 912.500 [652.000, 1218.000] | 910.000 [637.000, 1208.000] | 912.500 [673.500, 1256.000] | -0.355 | 0.723 | |
| Magnesium (mg) | 217.500 [164.000, 269.000] | 211.000 [159.000, 264.000] | 225.500 [179.500, 286.000] | -2.814 | 0.005 | |
| Iron (mg) | 12.850 [9.775, 16.835] | 12.270 [9.425, 16.335] | 13.715 [10.520, 17.955] | -3.152 | 0.002 | |
| Zinc (mg) | 9.245 [6.540, 11.750] | 8.760 [6.410, 11.305] | 9.840 [6.905, 12.735] | -3.199 | 0.001 | |
| Copper (mg) | 0.823 [0.628, 1.074] | 0.802 [0.603, 1.049] | 0.862 [0.698, 1.134] | -3.183 | 0.001 | |
| Selenium (mcg) | 87.550 [66.450, 115.900] | 82.550 [63.500, 110.000] | 96.350 [72.850, 129.050] | -4.489 | <0.001 | |
| Alcohol (g) | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.000 [0.000, 0.000] | 0.088 | 0.326 | |
| BMI (kg m2) | 18.100 [15.900, 22.500] | 17.100 [15.600, 20.100] | 21.000 [17.400, 26.900] | -8.926 | <0.001 | |
| PA (MET-min/wk) | 1920.000 [880.000, 3480.000] | 2000.000 [900.000, 3600.000] | 1880.000 [800.000, 2760.000] | 1.076 | 0.284 | |
| Cotinine (ng/mL) | 0.062 [0.011, 0.362] | 0.063 [0.017, 0.337] | 0.057 [0.011, 0.380] | 0.365 | 0.713 | |
| Carotene (RE) | 45.750 [20.542, 108.771] | 44.813 [20.417, 103.792] | 46.667 [22.958, 120.875] | -0.765 | 0.445 | |
| Oxidative balance score | 17.000 [12.000, 22.000] | 16.000 [12.000, 21.000] | 18.000 [14.000, 23.000] | -3.303 | <0.001 |
MET = metablic equivalent, PA = physical activity.
3.2. OBS exhibited better value in predicting the disease course of asthma
To assess the clinical value of OBS, ROC, and DCA analyses were performed by setting the outcomes as with asthma and the disease course. As shown in Figure 2A, the ROC results exhibited that OBS could not effectively predict asthma (AUC = 0.523), and the DCA results exhibited that OBS provided fewer net benefits for asthma (Fig. 2B). However, the OBS showed a certain value in predicting disease course of asthma with AUC = 0.625 (Fig. 2C), and the DCA curves showed OBS provided better net benefits for disease course of asthma (Fig. 2D). Therefore, we further explored the association between OBS and disease course of asthma in the following analyses.
Figure 2.
The prediction performance of OBS. The ROC curve (A) and the DCA curve (B) of OBS for predicting asthma. The ROC curve (C) and the DCA curve (D) of OBS for predicting the disease course of asthma. AUC = area under the curve, DCA = decision curve analysis, OBS = oxidative balance score, ROC = receiver operating characteristic, 95% CI = 95% confidence interval.
3.3. OBS-based logistic regression model improved the efficacy in predicting disease course of asthma
To improve the prediction efficacy, we constructed an OBS-based model including factors with significant differences between 2 disease course groups. Besides, the OBS components were excluded to avoid collinearity. Thus, age, sedentary time, and OBS were included in the final model using machine learning techniques after the samples were divided into training and validation sets. From Tables 3 and 4, the AUCs of XG Boost in the training and validation sets were 0.996 and 0.801, respectively. Besides, the AUC of Random Forest was 0.781 in the validation set but 0.996 in the training set. It indicated that XG Boost and Random Forest probably had an over-fitting phenomenon because of the great difference between them. However, the AUC of the logistic model for the training and validation sets were respectively 0.841 and 0.821. Thus, the logistic model was chosen for further analysis due to its stability with relatively good performance.
Table 3.
The machine learning techniques for the training set.
| AUC (95% CI) | Cutoff (95% CI) | Accuracy (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | F1 score (95% CI) | |
|---|---|---|---|---|---|---|
| XG Boost | 0.996 (0.992–0.999) | 0.48 (0.323–0.637) | 0.961 (0.942–0.980) | 0.978 (0.969–0.988) | 0.953 (0.921–0.984) | 0.954 (0.932–0.976) |
| logistic | 0.821 (0.774–0.867) | 0.245 (0.231–0.259) | 0.759 (0.759–0.759) | 0.979 (0.939–1.020) | 0.618 (0.585–0.650) | 0.765 (0.753–0.777) |
| Random Forest | 0.996 (0.992–1.000) | 0.5 (0.402–0.598) | 0.971 (0.964–0.977) | 0.987 (0.979–0.995) | 0.969 (0.949–0.989) | 0.968 (0.955–0.981) |
AUC = area under the curve, 95% CI = 95% confidence interval.
Table 4.
The machine learning techniques for the validation set.
| AUC (95% CI) | Cutoff (95% CI) | Accuracy (95% CI) | Sensitivity (95% CI) | Specificity (95% CI) | F1 score (95% CI) | |
|---|---|---|---|---|---|---|
| XG Boost | 0.801 (0.701–0.901) | 0.48 (0.323–0.637) | 0.662 (0.611–0.713) | 0.969 (0.908–1.030) | 0.652 (0.494–0.811) | 0.734 (0.602–0.866) |
| Logistic | 0.841 (0.754–0.927) | 0.245 (0.231–0.259) | 0.708 (0.644–0.771) | 0.977 (0.933–1.022) | 0.636 (0.601–0.671) | 0.671 (0.587–0.756) |
| Random Forest | 0.781 (0.675–0.887) | 0.5 (0.402–0.598) | 0.721 (0.708–0.734) | 0.855 (0.711–0.999) | 0.672 (0.595–0.749) | 0.735 (0.718–0.753) |
AUC = area under the curve, 95% CI = 95% confidence interval.
The ROC curves showed that the AUC for the OBS and the OBS-based logistic regression model were respectively 0.625 and 0.915 (Fig. 3A), and the DeLong test P value < .001 (Table 5), which suggested that the OBS-based model had a significantly improved prediction value for disease course of asthma. Furthermore, DCA curves exhibited that the OBS-based model could provide a great net benefit for the disease course of asthma (Fig. 3B).
Figure 3.
The ROC and DCA curves for comparing the predictive value of OBS and OBS-based models for asthma course. (A) The comparison between OBS and OBS-based using ROC curve. (B) The comparison between OBS and OBS-based using DCA curve. The OBS-based model was constructed by OBS, age, and sedentary time. AUC = area under the curve, DCA = decision curve analysis, OBS = oxidative balance score, ROC = receiver operating characteristic, 95% CI = 95% confidence interval.
Table 5.
The comparison of the AUC values between OBS and OBS-based model.
| Name | Model | OBS |
|---|---|---|
| OBS-based | <0.001 | |
| OBS | <0.001 |
AUC = area under the curve, OBS = oxidative balance score.
3.4. Identification of the key components of OBS
Due to the essential value of OBS in predicting the disease course of asthma, we next assessed the key OBS components using machine learning methods. The most important predictor in the logistic, Random Forest, and XG Boost model was BMI (Figs. 4A–C). Besides, the Venn plot showed the common factors in the top 5 important predictors of the 3 models were BMI, niacin, and selenium (Fig. 4D).
Figure 4.
The importance of the OBS component ranked by different models. The importance of the OBS components were ranked by the logistic regression model (A), Random Forest model (B), and the XG Boost model (C). (D) Venn plot showed the common factors in the top 5 important predictors of the 3 models. OBS = oxidative balance score.
The contents of niacin and selenium are ingested by diet, and diet might result in a change in BMI. Thus, we tried to determine the potential function of BMI in the association of niacin and selenium with the disease course of asthma using mediation analysis and evaluated the interplay among the factors. The results indicated that Niacin was directly related to the asthma course and also related to the course by regulating BMI (Table 6). However, BMI was not a mediator in the association of Selenium with the disease course of asthma (Table 7).
Table 6.
Mediating effect of BMI on the association between niacin and disease course of asthma.
| Niacin | Path | Coef | SE | P | CI [97.5%]] | Sig |
|---|---|---|---|---|---|---|
| X: BMI (kg/m2) | 0.102 | 0.023 | <.001 | 0.147 | Yes | |
| Y: BMI (kg/m2) | 0.031 | 0.004 | <.001 | 0.039 | Yes | |
| Total | 0.008 | 0.002 | <.001 | 0.013 | Yes | |
| Direct | 0.005 | 0.002 | .014 | 0.01 | Yes | |
| Indirect | 0.003 | 0.001 | <.001 | 0.005 | Yes |
X stands for OBS, and Y stands for disease course of asthma course.
BMI = body mass index.
Table 7.
BMI was not a mediator in the association of selenium with the disease course of asthma.
| Selenium | Path | Coef | SE | P | CI [97.5%] | Sig |
|---|---|---|---|---|---|---|
| X: BMI (kg/m2) | 0.005 | 0.003 | .090 | 0.01 | No | |
| Y: BMI (kg/m2) | 0.031 | 0.004 | <.001 | 0.039 | Yes | |
| Total | 0.000 | 0 | .179 | 0.001 | No | |
| Direct | 0.000 | 0 | .393 | 0.001 | No | |
| Indirect | 0.000 | 0 | .020 | 0.001 | Yes |
X stands for OBS, and Y stands for disease course of asthma course.
BMI = body mass index.
4. Discussion
In this study, we embarked upon an exploration of the intricate interplay between OBS and asthma employing a substantial cohort of children sourced from the NHANES database. Our findings unveiled a noteworthy correlation suggesting that individuals with elevated OBS levels exhibited prolonged durations of asthma affliction. Furthermore, in comparison with the prediction value of asthma, OBS had a better prediction value in the disease course of asthma.
We found that the OBS level may promote the development of asthma. The sign of asthma is airway inflammation. Oxidative stress may trigger and aggravate inflammation, or it may be caused by inflammation.[18] Decades of studies have yielded a substantial body of evidence showcasing an augmented burden of oxidative stress in the context of asthma with the increasing content of oxidants and decreasing content of antioxidants in asthma patients compared with healthy people,[19–21] On the one hand, inflammation will lead to the recruitment of a variety of inflammatory cells, such as macrophages, neutrophils, and eosinophils, which will produce a large number of ROS and aggravate oxidative stress.[22] On the other hand, excessive ROS production increases the infiltration of inflammatory cells, and at the same time, it intensifies the level of inflammation by inducing the production of pro-inflammatory cytokines, including TNF-α, IL-6, and IL-1β.[23] Therefore, some studies indicated that the severity of the asthma showcased a direct correlation with heightened oxidative stress levels.[24,25] OBS, a comprehensive gauge amalgamating information pertaining to dietary and lifestyle factors, acts as an evaluative marker for oxidative stress levels. It is known to increase in response to antioxidant exposure and has been demonstrated to be intricately linked to the development of various ailments, including type 2 diabetes, cardiovascular disease, and diverse forms of cancer.[26] In our study, we noted a discernible elevation in OBS levels within the asthma group, particularly among those with longer disease courses, when compared to their healthy counterparts and individuals with shorter disease durations. This observation suggested a positive association between OBS and asthma development. However, the results of the ROC analysis and DCA for OBS were not deemed satisfactory. These outcomes underscore the need to incorporate additional factors beyond OBS for more effective prediction of asthma and disease progression.
The further analysis indicated that a model incorporating age, sedentary time, and OBS proved more effective in predicting the disease course of asthma. Extensive evidence has confirmed the significant impact of age and sedentary behavior on asthma. Arbes et al highlighted that older children were more prone to asthma within a specific age range, suggesting that age played a crucial role in asthma susceptibility.[27] Furthermore, the interaction between age and eosinophils has shown promise in predicting childhood asthma with greater accuracy.[27] These findings suggested that older children may be more vulnerable to various factors that contribute to asthma development. On a related note, Roncada et al demonstrated a high prevalence of sedentary behavior among children with asthma.[28] Moreover, several studies have found a negative association between increased sedentary time and asthma outcomes, indicating that prolonged periods of inactivity were linked to worsened asthma symptoms.[29] However, a retrospective study demonstrated that while self-reported sedentary time was significantly longer in youth with asthma compared to those without asthma, no notable difference was observed when sedentary time was objectively measured using accelerometry data.[30] These results implied that the influence of age and sedentary behavior on asthma development was influenced by various factors, including levels of physical activity. This suggested that the interaction among these factors held potential implications for both asthma development and treatment strategies.
It is worth noting that certain components of OBS are acquired through dietary intake. Thus, we can infer that certain dietary substances may play a role in asthma development by influencing BMI. Through mediation analysis examining the relationship between OBS components and the disease course of asthma, we identified a significant positive association between niacin and the disease course of asthma. Moreover, we found that Niacin was directly related to the asthma course and also related to the course by regulating BMI. A previous study suggested that dietary intake could influence antioxidant status,[31] with a notable impact on BMI.[32,33] Furthermore, Lu and Forno proposed that the consumption of a “Western diet” may exacerbate asthma outcomes by modulating inflammatory processes.[34] A further study found that the high-fat meal promoted neutrophilic airway inflammation and inhibited bronchodilator recovery, which aggravated asthma.[35] In addition, high-fat meal usually increases BMI. Some studies demonstrated a progressive impairment of antioxidant status with an increasing BMI.[36] Moreover, based on the in vivo results, MDA content was significantly increased in asthma mice, while GSH activity was significantly decreased, which indicated the oxidative stress was induced in asthma mice.[37] Moreover, a positive correlation between MDA and NF-κB activation was observed in the obesity-asthma mice, which suggested that the function of oxidative stress in asthma may be regulated by NF-κB signal pathway.[37]
However, it is important to acknowledge the limitations of our study. Firstly, our data was derived from NHANES and primarily represents the American population, thereby potentially introducing geographic and racial biases. Secondly, although we have adjusted for numerous confounding factors, there may still be other unaccounted variables that could influence our results. Additionally, the definition of asthma is based on questionnaire data with potential recall bias, and hence the results should be verified in a cohort with clinically diagnosed asthma. Moreover, future studies are warranted to further investigate the potential impact of interventions targeting oxidative balance on asthma outcomes.
5. Conclusion
To summarize, our study reveals that an OBS-based logistic regression model integrating OBS, in conjunction with age and sedentary time, can serve as a predictive tool for the disease course of asthma. Furthermore, our findings suggest that niacin may be a key component within OBS that is related to asthma duration. To mitigate the course of asthma, it is crucial to implement dietary control measures to maintain oxidative balance and minimize sedentary behaviors among children.
Author contributions
Conceptualization: Yan Lin.
Data curation: Yan Lin, Bin-Wei Qiu, Jin-Liang Lin.
Formal analysis: Yan Lin.
Methodology: Kai-Li Xu.
Supervision: Jin-Liang Lin.
Writing – original draft: Yan Lin, Bin-Wei Qiu, Kai-Li Xu, Jin-Liang Lin.
Writing – review & editing: Jin-Liang Lin.
Supplementary Material
Abbreviations:
- AUC
- area under the curve
- BMI
- body mass index
- DCA
- decision curve analysis
- NHANES
- National Health and Nutrition Examination Survey
- OBS
- oxidative balance score
- PA
- physical activity
- PIR
- poverty income ratio
- ROC
- receiver operating characteristic
- ROS
- reactive oxygen species
The authors have no funding and conflicts of interest to disclose.
The Ethics Committee of Longyan First Affiliated Hospital of Fujian Medical University deemed that this research is based on open-source data, so the need for ethics approval was waived.
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.
Supplemental Digital Content is available for this article.
How to cite this article: Lin Y, Qiu B-W, Xu K-L, Lin J-L. Association between oxidative balance score and asthma course in the American children: A cross-sectional analysis of the NHANES 2011–2018. Medicine 2025;104:17(e42262).
Contributor Information
Yan Lin, Email: Danva01@163.com.
Bin-Wei Qiu, Email: 35727910@qq.com.
Kai-Li Xu, Email: 615201604@qq.com.
References
- [1].Ma R, Liang L, Kong Y, et al. Spatiotemporal variations of asthma admission rates and their relationship with environmental factors in Guangxi, China. BMJ Open. 2020;10:e038117. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [2].Chen H, Xu X, Cheng S, et al. Small interfering RNA directed against microRNA-155 delivered by a lentiviral vector attenuates asthmatic features in a mouse model of allergic asthma. Exp Ther Med. 2017;14:4391–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Kim JS, Shin IS, Shin NR, Nam JY, Kim C. Genome-wide analysis of DNA methylation and gene expression changes in an ovalbumin-induced asthma mouse model. Mol Med Rep. 2020;22:1709–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Hu RH, Wu CT, Wu TS, et al. Systematic characterization of the group 2 house dust mite allergen in dermatophagoides microceras. Front Cell Infect Microbiol. 2021;11:793559. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Liu H, Min J, Sun H, Zhang CL. Correlation study between gene and respiratory disease in children. Exp Ther Med. 2017;14:1463–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Amoah AS, Forson AG, Boakye DA. A review of epidemiological studies of asthma in Ghana. Ghana Med J. 2012;46(2 Suppl):23–8. [PMC free article] [PubMed] [Google Scholar]
- [7].Nkosi V, Rathogwa-Takalani F, Voyi K. The frequency of fast food consumption in relation to wheeze and asthma among adolescents in gauteng and North West Provinces, South Africa. Int J Environ Res Public Health. 2020;17:1994. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Prossegger J, Huber D, Grafetstatter C, et al. Winter exercise reduces allergic airway inflammation: a randomized controlled study. Int J Environ Res Public Health. 2019;16:2040. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Khreis H, Nieuwenhuijsen MJ. Traffic-related air pollution and childhood asthma: recent advances and remaining gaps in the exposure assessment methods. Int J Environ Res Public Health. 2017;14:312. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Youssef MS, Mira MM, Renault S, Hill RD, Stasolla C. Phytoglobin expression alters the Na(+)/K(+) balance and antioxidant responses in soybean plants exposed to Na(2)SO(4). Int J Mol Sci . 2022;23:4072. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [11].Vincenzo SD, Ferrante G, Ferraro M, et al. Oxidative stress, environmental pollution, and lifestyle as determinants of asthma in children. Biology (Basel). 2023;12:133. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Liou CJ, Cheng CY, Yeh KW, Wu YH, Huang WC. Protective effects of casticin from vitex trifolia alleviate eosinophilic airway inflammation and oxidative stress in a murine asthma model. Front Pharmacol. 2018;9:635. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [13].Qu H. The association between oxidative balance score and periodontitis in adults: a population-based study. Front Nutr. 2023;10:1138488. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Xu Z, Xue Y, Wen H, Chen C. Association of oxidative balance score and lung health from the National Health and Nutrition Examination Survey 2007–2012. Front Nutr. 2022;9:961950. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Suvakov S, Jerotic D, Damjanovic T, et al. Markers of oxidative stress and endothelial dysfunction predict haemodialysis patients survival. Am J Nephrol. 2019;50:115–25. [DOI] [PubMed] [Google Scholar]
- [16].Moon ME, Jung DH, Heo SJ, Park B, Lee YJ. Oxidative balance score and new-onset Type 2 diabetes mellitus in korean adults without non-alcoholic fatty liver disease: Korean Genome and Epidemiology Study-Health Examinees (KoGES-HEXA) Cohort. Antioxidants (Basel). 2024;13:107. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Wang X, Yang S, He G, Xie L. The association between weight-adjusted-waist index and total bone mineral density in adolescents: NHANES 2011-2018. Front Endocrinol (Lausanne). 2023;14:1191501. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [18].Dozor AJ. The role of oxidative stress in the pathogenesis and treatment of asthma. Ann N Y Acad Sci. 2010;1203:133–7. [DOI] [PubMed] [Google Scholar]
- [19].Karadogan B, Beyaz S, Gelincik A, Buyukozturk S, Arda N. Evaluation of oxidative stress biomarkers and antioxidant parameters in allergic asthma patients with different level of asthma control. J Asthma. 2022;59:663–72. [DOI] [PubMed] [Google Scholar]
- [20].Ahmad A, Shameem M, Husain Q. Relation of oxidant–antioxidant imbalance with disease progression in patients with asthma. Ann Thorac Med. 2012;7:226–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [21].Ozaras R, Tahan V, Turkmen S, et al. Changes in malondialdehyde levels in bronchoalveolar fluid and serum by the treatment of asthma with inhaled steroid and beta2-agonist. Respirology. 2000;5:289–92. [DOI] [PubMed] [Google Scholar]
- [22].Kirkham P, Rahman I. Oxidative stress in asthma and COPD: antioxidants as a therapeutic strategy. Pharmacol Ther. 2006;111:476–94. [DOI] [PubMed] [Google Scholar]
- [23].Qu J, Li Y, Zhong W, Gao P, Hu C. Recent developments in the role of reactive oxygen species in allergic asthma. J Thorac Dis. 2017;9:E32–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Ercan H, Birben E, Dizdar EA, et al. Oxidative stress and genetic and epidemiologic determinants of oxidant injury in childhood asthma. J Allergy Clin Immunol. 2006;118:1097–104. [DOI] [PubMed] [Google Scholar]
- [25].Dut R, Dizdar EA, Birben E, et al. Oxidative stress and its determinants in the airways of children with asthma. Allergy. 2008;63:1605–9. [DOI] [PubMed] [Google Scholar]
- [26].Song L, Zhou H, Yang Q, et al. Association between the oxidative balance score and thyroid function: Results from the NHANES 2007–2012 and Mendelian randomization study. PLoS One. 2024;19:e0298860. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [27].Arbes SJ, Jr., Calatroni A, Mitchell HE, Gergen PJ. Age-dependent interaction between atopy and eosinophils in asthma cases: results from NHANES 2005–2006. Clin Exp Allergy. 2013;43:544–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [28].Roncada C, de Oliveira SG, Cidade SF, et al. Burden of asthma among inner-city children from Southern Brazil. J Asthma. 2016;53:498–504. [DOI] [PubMed] [Google Scholar]
- [29].Lu KD, Forno E, Radom-Aizik S, Cooper DM. Low fitness and increased sedentary time are associated with worse asthma-The National Youth Fitness Survey. Pediatr Pulmonol. 2020;55:1116–23. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Hartz J, Bucholz E, Phipatanakul W, de Ferranti S, Powell-Wiley T. The association of objectively measured sedentary time with asthma in US youth: a glimpse into the connection between obesity and asthma. Pediatr Pulmonol. 2023;58:1582–91. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [31].Misso NL, Thompson PJ. Oxidative stress and antioxidant deficiencies in asthma: potential modification by diet. Redox Rep. 2005;10:247–55. [DOI] [PubMed] [Google Scholar]
- [32].Thielman J, Copes R, Rosella LC, Chiu M, Manson H. Is neighbourhood walkability related to body mass index among different age groups? A cross-sectional study of Canadian urban areas. BMJ Open. 2019;9:e032475. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Kummer K, Jensen PN, Kratz M, et al. Full-fat dairy food intake is associated with a lower risk of incident diabetes among american indians with low total dairy food intake. J Nutr. 2019;149:1238–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Lu KD, Forno E. Exercise and lifestyle changes in pediatric asthma. Curr Opin Pulm Med. 2020;26:103–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [35].Wood LG, Garg ML, Gibson PG. A high-fat challenge increases airway inflammation and impairs bronchodilator recovery in asthma. J Allergy Clin Immunol. 2011;127:1133–40. [DOI] [PubMed] [Google Scholar]
- [36].Lei X, Xu Z, Chen W. Association of oxidative balance score with sleep quality: NHANES 2007–2014. J Affect Disord. 2023;339:435–42. [DOI] [PubMed] [Google Scholar]
- [37].Liu X, Yi M, Jin R, et al. Correlation between oxidative stress and NF-kappaB signaling pathway in the obesity-asthma mice. Mol Biol Rep. 2020;47:3735–44. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.