Data was extracted from National Health and Nutrition Examination Survey (NHANES). NHANES is a cross-sectional probability survey based on the U.S. population aged 0 to 85 to collect information on the health and nutrition of the U.S. family population . The survey is divided into two parts: family interview and health examination. NHANES is a free public dataset approved by the Research Ethics Review Board (ERB). The document approving the NHANES 2005–2006 data review was Protocol #2005-06 .
Selected from 10,348 subjects from NHANES 2005–2006, after screening for allergy, DM, and medical conditions questionnaires and excluding data for no demographic variables, BMI, serum cotinine, alcohol use, physical activity and high blood pressure (HBP), a total of 1687 subjecs were obtained as the final study population. Figure 1 shows the screening process for study subjects.
2.2. Allergic symptoms
Allergy questionnaires and medical condition questionnaires were integrated and selected to obtain allergy symptom data. Screening the questions on the Allergy Questionnaire for "Doctors diagnosed hay fever, allergy, eczema or dymptoms in the past 12 months" (AGQ010, AGQ030, AGQ040, AGQ060, AGQ100, AGQ140, AGQ180)  and the questions on the Medical Conditions Questionnaire for "Doctor diagnosed asthma and asthma attacks in the past 12 months" (MCQ010, MCQ040) . Allergy symptoms were considered to be present if the subject answered yes to some question , and no allergy symptoms were considered to be present if the subject answered no to all questions.
2.3. Allergen sensitization
The lower limit of detection (LLOD) for the detection of specific immunoglobulin E (sIgE) antibodies against 19 allergens, D. Farinae, D. Pteronyssinus, cat, dog, cockroach, Alternaria, peanut, egg, milk, ragweed, rye grass, Bermuda grass, oak, birch, shrimp, aspergillus, thistle, mouse and rat, was the same at 0.35 kU/L, and there was no data exceeding the upper limit of detection . Allergen sensitization was considered to be present if a subject's test values were higher than the LLOD (sIgE ≥ 0.35 kU/L), and no specific sensitization was considered to be present if all subjects' test values were lower than the LLOD (sIgE < 0.35 kU/L) .
2.4. Allergen allergy
A subject was considered to have allergen allergy if the subject had allergic symptoms and allergen sensitization, otherwise it was not considered to have allergen allergy.
All subjects aged ≥ 12 years underwent morning FPG testing to assess DM after a 9-hour fast . For the complete sample, testing for A1c was also performed .
Screening and integration of the questions on the Diabetes Questionnaire for "Doctors diagnosed Diabetes and taking diabetes medications" (DIQ010, DIQ070) . For the missing questions in 103 prediabetics (DIQ010 = 3), estimates were refined based on the use of oral antidiabetic drugs, Acarbose, Cholopropamide, Glimepiride, Glipizide, Glucagon, Glyburide, Metformin, Miglitol, Nateglinide, Pioglitazone, Repagilinide, Rosiglitazone, Tolazamide, Troglitazone, Glipizide and Metformin, Glyburide and Metformin, Metformin and Rosiglitazone, reported in the prescription drug section of the household questionnaire . The missing response code for DID070 was “1” (with DM) if the prediabetics reported any oral antidiabetic drug for all medications used in the month prior to the interview, and if the prediabetics did not report the use of any such oral hypoglycemic agents, the code was “2” (without DM) .
DM was considered to be present if FPG ≥ 126mg/dL, or A1c ≥ 6.5% , or the subject answered yes to DM in the diabetes questionnaire interview, and no DM was considered to be present if all the above three conditions are negative.
2.6. Confounding factors
Some known confounding factors affecting the development of allergic disease and DM, such as age , sex, race or ethnicity, country of birth, total number of households , family poverty income ratio (PIR) , education level of household reference , body mass index (BMI) , cotinine , alcohol consumption , physical activity  and HBP  were sorted out.
Age was divided into three confounding factors: young people (< 44 years old), middle-aged people (45–59 years old), and elderly people (> 60 years old), according to the age classification criteria of the World Health Organization (WHO) . Sex was divided into two confounding factors: males and females. Race or ethnicity was divided into five confounding factors: Mexican American, other Hispanic, non-Hispanic White, non-Hispanic Black, and other race. Country of birth was divided into three confounding factors: U.S., Mexico and elsewhere. Total number of households was divided into five confounding factors: 1, 2, 3, 4, ≥ 5 . PIR is a measure of family economic well-beings and is the ratio of income to household poverty. When PIR < 1.85, the family is eligible for federal food assistance . PIR was divided into two confounding factors: <1.85 and ≥ 1.85. The household reference person is the first household member aged ≥ 18 listed on the Screener Household Member Roster to own or lease the dwelling in which the household member lives . Use information about the reference person to describe the household situation in which the survey subjects lived. Reference person education level which is the highest grade or level of school that the household reference person has received was divided into three confounding factors: <12th Grade, 12th Grade/GED, and > 12th Grade . BMI is a statistical index that estimates body fat in males and females of any age group, obtained by dividing a person's weight in kilograms by the square of their height in meters . BMI was divided into underweight (< 18.5 kg/m2), normal weight (18.5–24.9 kg/m²), overweight (25-29.9 kg/m²) and obese (≥ 30 kg/m²), according to the BMI classification of WHO and the National Institutes of Health . Cotinine, which was used to describe the subjects’ smoking status, was divided into three confounding factors: < 0.015 ng/mL, < 10 ng/mL, ≥ 10 ng/mL . Alcohol consumption screened the alcohol use questionnaire was divided into two confounding factors: yes (drinking) and no (non-drinking) . Physical activity screened the physical activity questionnaire was divided into two confounding factors: yes (moderate activity) and no (no moderate activity) . HBP screened by the blood pressure & cholesterol questionnaire was divided into two confounding factors: yes (HBP) and no (no HBP) .
2.7. Statistical analysis
Independent logistic regression models were established, with allergic symptoms, allergen sensitization and allergy as the three outcome variables, DM as the independent variable, after adjusting for the above confounding factors. To illustrate the complex sample survey design of NHANES 2005–2006, the design variables [56, 57] were applied to all data analyses in this study.
In order to observe whether allergy symptoms and allergen sensitization are related to DM under specific conditions, stratified analysis was performed on the basis of the above independent models. Each time a confounding factor was selected as a layer variable, a logistic regression model was established under the condition that all the remaining confounding factors were adjusted, and the dependent and independent variables remained unchanged, and correlation analysis was performed.
Logistic regression analysis with complex sampling was performed using SPSS (version 23.0.0, IBM, USA). Forest plots were drawn using Rstudio (version 1.3.1073.0). The Wald test was used to check whether the constraints in the regression hold. Results were considered statistically significant when the probability (P) < 0.05.