Multivariable logistic regression

Socio-demographic core profile

We first examined the effect of socio-demographic characteristics- these served as the “core model” to which other groups (e.g. biochemical and hematological marker groups) were added separately. Among these factors, an increase in age was associated with a modest but statistically significant increase in risk, yielding a 2 % increase in the odds of taste dysfunction per additional year at screening (aOR=1.02, 95 % CI: [1.004, 1.04]; p=0.034). Race and Hispanic origin demonstrated robust independent significance within the model (p=0.018). Using Asians and Other Races as the reference cohort, the adjusted odds of reporting taste dysfunction were substantially higher among Non-Hispanic White (aOR=2.54, 95 % CI: [1.15, 5.56]) and Non-Hispanic Black (aOR=2.45, 95 % CI: [1.29, 4.68]) participants.

Educational attainment proved to be highly significant overall (p<0.001), driven primarily by individuals with less than a high school education, who displayed a 59 % increase in the odds of taste dysfunction compared to college graduates (aOR=1.59, 95 % CI: [1.20, 2.10]). Similarly, annual household income was an independent predictor (p=0.002), with lower economic strata (<$45,000/yr) exhibiting a trend toward elevated risk (aOR=1.73, 95 % CI: [0.88, 3.39]) relative to high earners (≥$75,000/yr). Conversely, gender (p=0.380) and marital status (p=0.720) did not significantly influence taste outcomes in the fully adjusted model. Figure 2 illustrates the results of multivariate regression in the form of a Forrest plot.

Figure 2:: Forest plot of socio-demographic and clinical predictors for self-reported taste dysfunction: The plot displays the adjusted odds ratios (aORs) and corresponding 95 % confidence intervals (CIs) derived from multivariate logistic regression analysis. The vertical dashed line represents the null effect (aOR) = 1.00). Data points to the right of the line indicate increased odds of taste dysfunction, while points to the left indicate decreased odds. The x-axis is presented on a logarithmic scale.Red markers and error bars denote statistically significant predictors (where the 95 % CI does not cross 1.00); blue markers and error bars indicate non-significant predictors. Gray square markers located exactly on the null line represent the designated reference categories (ref) for categorical variables (race/hispanic origin, marital status, educational attainment, and annual household income). Fatty acids (model 1) and fatty acids (model 2) represent separate multivariate models adjusted for the core socio-demographic profile and respective biochemical parameters. Abbreviations: aOR, adjusted odds ratio; ALT, alanine aminotransferase; CBC, complete blood count; CI, confidence interval; NHANES, N ational Health and Nutrition Examination Survey; RBC, red blood cell; WBC, white blood cell.

Figure description: Forest plot presenting adjusted odds ratios (aORs) with 95% confidence intervals for predictors of 12-month self-reported taste dysfunction derived from multivariable logistic regression models. The x-axis is a logarithmic scale of odds ratios with a vertical reference line at aOR = 1.0 indicating no association. Data points to the right of the line indicate increased odds of taste dysfunction, whereas points to the left indicate decreased odds. Categories evaluated include socio-demographic characteristics, fatty acid measures, liver and kidney biomarkers, nutritional markers, vitamins, and hematologic indices. Significant predictors include older age, non-Hispanic White race, non-Hispanic Black race, lower educational attainment, lower percentage of saturated fatty acids, higher percentage of monounsaturated fatty acids, lower albumin concentration, and higher serum copper concentration. Most remaining biochemical and hematologic variables show confidence intervals crossing the null value and are not statistically significant. Gray squares denote reference categories for categorical variables.

Circulating fatty acid profiles (models 1 & 2)

The independent contributions of circulating fatty acid speciation metrics were analyzed using two separate multivariable models to account for collinearity between the saturated and unsaturated fatty acids. In the first model that included the relative percentage of saturated and omega-3 fatty acids, the former demonstrated an inverse association with taste problems; each percentage unit increase in SFAs was associated with a 12 % reduction in the odds of experiencing a taste problem (aOR=0.88, 95 % CI: [0.80, 0.96]; p=0.006).

In the second model, which included the percentage of monounsaturated and omega-3 fatty acids as covariates, the former exhibited a positive association with taste problems- the chances of self-reported taste problems increased by 14 % per unit increase in the percentage of MUFAs (aOR=1.14, 95 % CI: [1.04, 1.25]; p=0.006). The relative percentage of omega-3 fatty acids was not associated with taste problems (Table 3 and Figure 2).

Variable name | Adjusted odds ratio (aOR) | 95 % confidence interval (CI) | p-Value
Socio-demographic core profile |  |  | 
Age | 1.02 | [1.004, 1.04] | 0.034
Male gender | 1.21 | [0.70, 2.20] | 0.380
Race/Hispanic origin | – | – | 0.018
Mexican Americans and Hispanics | 1.67 | [0.78, 3.62] | –
Non-Hispanic White | 2.54 | [1.15, 5.56] | –
Non-Hispanic Black | 2.45 | [1.29, 4.68] | –
Asians and other races | 1.00 (ref) | – | –
Marital status | – | – | 0.720
Married or living with partner | 1.06 | [0.74, 1.53] | –
Widowed, divorced or separated | 1.00 (ref) | – | –
Educational attainment | – | – | <0.001
Less than high school | 1.59 | [1.20, 2.10] | –
High school graduate | 0.73 | [0.38, 1.41] | –
College graduate or more | 1.00 (ref) | – | –
Annual household income | – | – | 0.002
Less than $45,000/yr | 1.73 | [0.88, 3.39] | –
$45,000 – Less than $75,000/yr | 0.57 | [0.24, 1.36] | –
More than $75,000/yr | 1.00 (ref) | – | –
Biochemical and hematologic parameters |  |  | 
Fatty acids (model 1) |  |  | 
Percentage of saturated fatty acids | 0.88 | [0.80, 0.96] | 0.006
Percentage of Omega-3 fatty acids | 1.13 | [0.89, 1.44] | 0.290
Fatty acids (model 2) |  |  | 
Percentage of monounsaturated fatty acids | 1.14 | [1.04, 1.25] | 0.006
Percentage of Omega-3 fatty acids | 1.29 | [0.98, 1.70] | 0.064
Liver & kidney biomarkers |  |  | 
Albumin | 0.41 | [0.18, 0.98] | 0.044
ALT | 1.005 | [0.99, 1.01] | 0.190
Alkaline phosphatase | 1.001 | [0.99, 1.01] | 0.390
Chloride | 0.94 | [0.86, 1.02] | 0.100
Potassium | 1.36 | [0.60, 3.05] | 0.440
Creatinine | 1.25 | [0.84, 1.85] | 0.250
Glucose | 0.99 | [0.99, 1.01] | 0.640
Nutritional markers & vitamins |  |  | 
RBC folate | 1.001 | [0.99, 1.004] | 0.280
5-Methyl tetrahydrofolate | 0.97 | [0.93, 1.004] | 0.081
Serum selenium | 1.01 | [0.99, 1.02] | 0.440
Serum copper | 1.02 | [1.01, 1.03] | 0.006
Serum iron | 1.006 | [0.99, 1.02] | 0.270
Blood mercury | 0.90 | [0.68, 1.19] | 0.430
Total cholesterol | 0.99 | [0.98, 1.003] | 0.180
Hematologic CBC indices |  |  | 
WBC count | 1.08 | [0.96, 1.21] | 0.220
Monocyte count | 0.92 | [0.47, 1.77] | 0.780

Biochemical, hepatic, and renal biomarkers

Serum albumin was the only variable in this group that had a significant inverse association with taste problems (p=0.044), i.e., higher albumin concentrations were associated with lower odds of taste dysfunction. A 1 g/dL increase in albumin reduced the odds of taste problems by 59 % (aOR=0.41, 95 % CI: [0.18, 0.98]). Other markers such as ALT (p=0.190), alkaline phosphatase (p=0.390), potassium (p=0.440), chloride (p=0.100), creatinine (p=0.250), and glucose (p=0.640), failed to reach statistical significance.

Nutritional markers, micronutrients, and hematology

Of the heavy metals and nutritional factors tracked, serum copper remained associated with taste impairment (p=0.006). For each 1 µg/dL increase in copper concentration, the adjusted odds of taste dysfunction rose by 2 % (aOR=1.02, 95 % CI: [1.01, 1.03]).

The folate intermediate 5-methyl tetrahydrofolate showed a marginally significant association with taste dysfunction (aOR=0.97, 95 % CI: [0.93, 1.004]; p=0.081). No associations were observed between RBC folate (p=0.280), selenium (p=0.440), iron (p=0.270), mercury (p=0.430), total cholesterol (p=0.180), or complete blood count indices, including white blood cell (p=0.220) and monocyte (p=0.780) counts and taste problems.