Factors Influencing Glycohemoglobin (HbA1c) - Essay On Study Design And Statistical Analysis.

Background and Goals

Blood sugar, also known as glucose is the main sugar found in the blood, it is sourced from the food consumed by and plays a critical role in supplying energy to the body.  However, too much blood sugar may lead to disorders, such as diabetes. Glycohemoglobin (HbA1c) is a medical test applicable in the diagnosis of diabetes. Notably, HbA1c of 6.5% is recommended as the cut point for diagnosing diabetes; however, a value of less than 6.5% does not exclude diabetes diagnosed using glucose tests. The test reflects average plasma glucose over the previous 8 to 12 weeks; besides, it is performed at any time of the day and does not require any special preparation such as fasting. Therefore, the following study seeks to expose the Factors Influencing Glycohemoglobin (HbA1c).

As revealed, the test is used to identify the average of blood glucose levels over the last 2 to 3 months. Moreover, it could be used to determine whether a person has diabetes or is susceptible to developing it over time (at risk). Therefore, the NHANES dataset from the year 2009-2010 revolving around Glycohemoglobin levels (HbA1C) was used to assess the factors influencing the HbAIC. The dataset incorporated 6795 observations and 20 variables. Consequently, the dataset was divided into 2 categories, with Diabetes and without Diabetes. The division was done through filtering the DX variable, where 1= with Diabetes had 890 observations whereas 0= without Diabetes had 5881 observations. In both data sets, the variables of interest or linked to have an influence on the Glycohemoglobin levels include BMI levels, Age, sex (gender), income and race.

Loading the data into R-markdown and sub-setting into two datasets

load(file="nhgh.rda")
## Filtering the data
diabetes <- subset(nhgh, dx==1)
nodiabetes <- subset(nhgh, dx==0)

People without Diabetes

GH Vs BMI

The plot above exhibits a positive relationship between GH and BMI

GH Vs BMI

The plot above exhibits a positive relationship between GH and Age

GH Vs Sex

The plot above exhibits a that both male and female tend to have similar levels of GH

GH Vs Income

The plot above exhibits income may not influence the level of GH of an individual

GH Vs Race

The plot above exhibits race may not influence the level of GH of an individual

The plot above exhibits a positive relationship between GH and BMI

The plot above exhibits no relationship between GH and Age

The plot above exhibits that male have higher GH levels compared to the female

The plot above exhibits income may influence the level of GH of an individual

The plot above exhibits race may influence the level of GH of an individual

As revealed, the plots above exhibited the relationship between GH and factors thus to assess whether the relationship is significant, 2 multiple regression analysis were performed to evaluate the type and significance of the relationship. In both data sets, the response variable is Glycohemoglobin (GH) whereas the explanatory variables include BMI, AGE, SEX, INCOME, RACE. Therefore, each regression model would be assessed to evaluate the relationship between independent and dependent variables.

Study Design and Data

Model 1: People without Diabetes

mod1=lm(gh~bmi+age+sex+income+re, data=nodiabetes)
summary(mod1)
## Call:
## lm(formula = gh ~ bmi + age + sex + income + re, data = nodiabetes)
## Residuals:
##     Min      1Q  Median      3Q     Max
## -1.7967 -0.2462 -0.0238  0.1961  8.9641
## Coefficients:
##                                       Estimate Std. Error t value Pr(>|t|)   
## (Intercept)                          4.8049126  0.0482141  99.658  < 2e-16
## bmi                                  0.0133834  0.0010845  12.341  < 2e-16
## age                                  0.0093688  0.0003595  26.061  < 2e-16
## sexfemale                           -0.0415093  0.0139616  -2.973  0.00296
## income[5000,10000)                   0.0041296  0.0475172   0.087  0.93075   
## income[10000,15000)                  0.0509317  0.0433311   1.175  0.23988   
## income[15000,20000)                  0.0096436  0.0447335   0.216  0.82932   
## income[20000,25000)                  0.0561812  0.0426723   1.317  0.18804   
## income[25000,35000)                  0.0152656  0.0405916   0.376  0.70687   
## income[35000,45000)                  0.0583567  0.0422923   1.380  0.16769   
## income[45000,55000)                 -0.0339233  0.0432499  -0.784  0.43287   
## income[55000,65000)                 -0.0145950  0.0458058  -0.319  0.75002   
## income[65000,75000)                 -0.0334274  0.0493590  -0.677  0.49829   
## income> 20000                        0.0389431  0.0522840   0.745  0.45640   
## income< 20000                        0.1733670  0.0752782   2.303  0.02131
## income[75000,100000)                -0.0257531  0.0425787  -0.605  0.54531   
## income>= 100000                     -0.0138706  0.0402677  -0.344  0.73051   
## reOther Hispanic                    -0.0503793  0.0275240  -1.830  0.06725
## reNon-Hispanic White                -0.1585478  0.0194146  -8.166  3.9e-16
## reNon-Hispanic Black                -0.0207042  0.0230383  -0.899  0.36886   
## reOther Race Including Multi-Racial -0.0244872  0.0335387  -0.730  0.46535   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.5197 on 5564 degrees of freedom
##   (296 observations deleted due to missingness)
## Multiple R-squared:  0.1649, Adjusted R-squared:  0.1619
## F-statistic: 54.95 on 20 and 5564 DF,  p-value: < 2.2e-16

There is a significant positive relationship between BMI levels affect the levels of Glycohemoglobin (β1= 0.0134, p <0.0001). Therefore, an increase in BMI levels affect the levels of Glycohemoglobin in people who do not have diabetes.

There is a significant positive relationship between BMI levels affect the levels of Glycohemoglobin (β2= 0.0094, p <0.0001). Therefore, an increase in age affect the levels of Glycohemoglobin in people who do not have diabetes.

Sex has a significant influence on the levels of Glycohemoglobin (β3= -0.0042, p =0.0029). Therefore, the GH levels of female is 0.0042 lower than that of male in people who do not have diabetes.

Income (<20,000) has a significant influence on the levels of Glycohemoglobin (β14= 0.1734, p =0.0213). Therefore, the GH levels of a person earning less than 20,000 is 0.1734 higher than that of above 20,000 in people who do not have diabetes.

Race has a significant influence on the levels of Glycohemoglobin (β18= -0.1585, p <0.0001). Therefore, the GH levels of Non-Hispanic white is 0.1585 lower than that of other races in people who do not have diabetes.

mod2=lm(gh~bmi+age+sex+income+re, data=diabetes)
summary(mod2)

##
## Call:
## lm(formula = gh ~ bmi + age + sex + income + re, data = diabetes)
##
## Residuals:
##     Min      1Q  Median      3Q     Max
## -3.2694 -1.0250 -0.3766  0.4604  9.2554
##
## Coefficients:
##                                      Estimate Std. Error t value Pr(>|t|)   
## (Intercept)                          6.363797   0.506010  12.576  < 2e-16
## bmi                                  0.006930   0.007459   0.929 0.353074   
## age                                  0.001425   0.003808   0.374 0.708340   
## sexfemale                           -0.402533   0.114421  -3.518 0.000457
## income[5000,10000)                   1.257823   0.451250   2.787 0.005429
## income[10000,15000)                  0.874046   0.413894   2.112 0.034992
## income[15000,20000)                  0.822784   0.414020   1.987 0.047202
## income[20000,25000)                  0.563283   0.419271   1.343 0.179466   
## income[25000,35000)                  1.095861   0.399654   2.742 0.006232
## income[35000,45000)                  1.094427   0.411725   2.658 0.008002
## income[45000,55000)                  0.687672   0.420766   1.634 0.102551   
## income[55000,65000)                  0.918352   0.441506   2.080 0.037814
## income[65000,75000)                  0.656110   0.452602   1.450 0.147520   
## income> 20000                        1.120662   0.443747   2.525 0.011732
## income< 20000                        0.570630   0.593597   0.961 0.336665   
## income[75000,100000)                 0.790378   0.428535   1.844 0.065469
## income>= 100000                      0.431785   0.413836   1.043 0.297066   
## reOther Hispanic                    -0.089702   0.217801  -0.412 0.680547   
## reNon-Hispanic White                -0.711512   0.156495  -4.547 6.23e-06
## reNon-Hispanic Black                -0.207027   0.175810  -1.178 0.239295    
## reOther Race Including Multi-Racial -0.015121   0.264778  -0.057 0.954471   
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.657 on 869 degrees of freedom
##   (24 observations deleted due to missingness)
## Multiple R-squared:  0.06816,    Adjusted R-squared:  0.04671
## F-statistic: 3.178 on 20 and 869 DF,  p-value: 3.512e-06

There is no significant relationship between BMI levels affect the levels of Glycohemoglobin (β1= 0.0069, p =0.353).

There is no significant relationship between BMI levels affect the levels of Glycohemoglobin (β2= 0.0014, p =0.7083).

Sex has a significant influence on the levels of Glycohemoglobin (β3= -0.4025, p =0.0029). Therefore, the GH levels of female is 0.4025 lower than that of male in people with diabetes.

Income (>20,000) has a significant influence on the levels of Glycohemoglobin (β13= 1.121, p =0.0117). Therefore, the GH levels of a person earning greater than 20,000 is 1.121 higher than that of less 20,000 in people with diabetes.

Race has a significant influence on the levels of Glycohemoglobin (β18= -0.7115, p <0.0001). Therefore, the GH levels of Non-Hispanic white is 0.1585 lower than that of other races in people with diabetes.