Does Total Cholesterol and Cigarette Smoking Act as Suitable Predictors of MI: An analysis of Framingham Trial Subset Data

0 / 5. 0

Does Total Cholesterol and Cigarette Smoking Act as Suitable Predictors of MI: An analysis of Framingham Trial Subset Data

Category: Research Paper

Subcategory: Statistics

Level: Masters

Pages: 3

Words: 825

Does Total Cholesterol and Cigarette Smoking Act as Suitable Predictors of MI: An analysis of Framingham Trial Subset Data
Introduction
The Framingham Heart Trial is one of the most elaborative trials, which was conducted to evaluate the cardiovascular risk factors as all-cause morbidity and mortality. Various biochemical, cardiovascular and physical parameters were collected. These factors were correlated and regressed in various forms to understand the potential risk for cardiovascular ailments. These included stroke, myocardial infarction and coronary heart diseases.
The review of literature suggests that smoking leads to vasoconstriction and hence it may be speculated to cause constriction of coronary blood vessels leading to ischemia and myocardial infarction. On the other hand, cholesterol may form atherosclerosis in the coronary vasculature, which could narrow the lumen and may lead to decreased oxygen supply to myocardium, leading to infarction. The present study was done to evaluate a subset data of the Framingham Study. This was done to find out whether total cholesterol levels are significantly different in individuals who have a prevalent myocardial infarction compared to individuals who do not have a prevalent myocardial infarction. The study also aimed to evaluate whether prevalent myocardial infarction can be predicted from episodes of smoking and the total cholesterol levels.
Methodology
Sampling
Subset data were collected from the Framingham trial based on the total cholesterol levels, some cigarettes smoked and presence of prevalent/non-prevalent myocardial infarction (Refer xls.). Data was sorted out from the excel sheet on total cholesterol and total cigarettes smoked per day. Data was randomly collected.
Hypothesis Testing & Statistical Tests
For the first question “whether total cholesterol levels are significantly different in individuals who have a prevalent myocardial infarction compared to individuals who do not have a prevalent myocardial infarction?”, an independent samples t-test was used. The null hypothesis contends that there is no significant difference between mean total cholesterol levels in prevalent MI group compared to No prevalent MI group (p>0.05). On the other hand, alternate hypothesis contends that there is a significant difference between mean total cholesterol levels in prevalent MI group compared to No prevalent MI group (p<0.05) (Cox, 2006).
For the second question “whether prevalent myocardial infarction can be predicted from some cigarettes smoked and the total cholesterol levels”, a logistic regression equation was constructed. The regression equation was framed to predict the chances of prevalent MI from episodes of smoking and the total cholesterol levels. Therefore, the dependent variable was myocardial infarction and the independent variables were some cigarettes smoked and the total cholesterol levels. The null hypothesis contends that prevalent myocardial infarction cannot be significantly predicted from episodes of smoking and the total cholesterol levels (p>0.05). On the other hand, alternate hypothesis contends that prevalent myocardial infarction can be significantly predicted from some cigarettes smoked and the total cholesterol levels (p<0.05) (Cox, 2006).
Software Used
All data were analyzed through MedCalc software.
Results
Representation of Independent Samples t test
Variable NoMI
Sample 2
Variable PREVMI
  Sample 1 Sample 2
Sample size 2481 94
Arithmetic mean 232.2120 122.7553
95% CI for the mean 230.0598 to 234.3642 97.6080 to 147.9026
Variance 2988.6478 15074.3588
Standard deviation 54.6685 122.7777
Standard error of the mean 1.0975 12.6636
F-test for equal variances P < 0.001
Welch-test (assuming unequal variances)
Difference -109.4567
Standard Error 12.7110
95% CI of difference -134.6947 to -84.2187
Test statistic t(d) -8.611
Degrees of Freedom (DF) 94.4
Two-tailed probability P < 0.0001

Regression Statistics
PREVMI
Method Enter
Sample size 2464
Positive cases a 46 (1.87%)
Negative cases b 2418 (98.13%)
a PREVMI = 1b PREVMI = 0
Overall Model Fit
Null model -2 Log Likelihood 457.379
Full model -2 Log Likelihood 455.343
Chi-squared 2.035
DF 2
Significance level P = 0.3615
Cox & Snell R2 0.0008256
Nagelkerke R2 0.004873
Coefficients and Standard Errors
Variable Coefficient Std. Error Wald P
CIGPDAY 0.014230 0.011181 1.6198 0.2031
TOTCHOL 0.0024593 0.0032098 0.5870 0.4436
Constant -4.69131 0.80151 34.2585 <0.0001
Odds Ratios and 95% Confidence Intervals
Variable Odds ratio 95% CI
CIGPDAY 1.0143 0.9923 to 1.0368
TOTCHOL 1.0025 0.9962 to 1.0088
Hosmer & Lemeshow test
Chi-squared 12.7058
DF 8
Significance level P = 0.1224
Contingency table for Hosmer & Lemeshow test [Show]Contingency table for Hosmer & Lemeshow test [Hide]Group Y=0 Y=1 Total
Observed Expected Observed Expected 1 240 238.590 2 3.410 242
2 243 245.237 6 3.763 249
3 248 246.040 2 3.960 250
4 235 236.042 5 3.958 240
5 245 243.731 3 4.269 248
6 245 242.545 2 4.455 247
7 242 242.286 5 4.714 247
8 235 240.953 11 5.047 246
9 242 240.468 4 5.532 246
10 243 242.108 6 6.892 249
Classification table (cut-off value p=0.5)
Actual group Predicted group Percent correct
0 1 Y = 0         2418 0 100.00%
Y = 1         46 0 0.00%
Percent of cases correctly classified 98.13%
ROC curve analysis
Area under the ROC curve (AUC)  0.570
Standard Error 0.0424
95% Confidence interval 0.550 to 0.590

Discussion and Conclusion
The results of the t-test indicated that total cholesterol levels significantly differed between prevalent MI group and No-MI group (p< 0.05). Therefore, it is suggested that total cholesterol significantly differs in prevalent MI group and No-MI group. However, in the No-Mi group, cholesterol levels were higher. This is a contradictory result, as cholesterol is anticipated to increase MI. Thus, the study should be reframed with LDL cholesterol levels as a risk factor for prevalent MI. This is because total cholesterol contains both LDL and HDL cholesterol, and it might be possible that due to high HDL levels, the MI was not found in the No-MI group. This was reciprocated by the regression equation too. The regression equation elucidated that prevalent myocardial infarction cannot be significantly predicted from episodes of smoking and the total cholesterol levels (p>0.05). The study reflected that prediction of MI should be studied from other variables, which constitutes cardiovascular risk, but in such model number of cigarettes smoked and the total cholesterol levels should not be incorporated.
References
Cox, D.R. (2006). Principles of Statistical Inference, Cambridge University Press