Broms Et Al. (2010) Found That Evening

¶ … Broms et al. (2010) found that evening type people were more likely to become smokers, and less likely to cease their nicotine addiction.

The study was a cohort design that used random-effects model regressions over a longitudinal period of time in order to study the comparison of attraction to smoking and tendency to remain addicted between diurnal and non-diurnal participants.

A sample of 23289 same-sex adult twin individuals from Finland was used. Nicotine dependence was studied on a sub-sample of 676 individual twins. The subjects were split into four groups: morning type, somewhat morning type, somewhat evening type, and evening type. Past and present smoking patterns were assessed. The DSM-IV and Fagerstrom Test for Nicotine Dependence (FTND) were both used to assess nicotine dependence.

Researchers found that of the four categories, it was diurnal types who were most likely to be past and current nicotine smokers and addicted to smoking, as well as less likely to relinquish the addiction. Even adjusting for potential confounders did not change these results.

The researchers concluded by suggesting that investigation should be conducted into discovering the causes for this association.

Analysis of statistics used and their significance

What statistical procedures are mentioned in the study?

The researchers gained their information from a longitudinal data set (gathered over a 15-year period), and, therefore, they used random-effects model regressions for repeated data. The 'random effects' model was used since the researchers randomly chose a subset of twins. They realized that they were dealing with a small subset of all possible values (i.e. twins as a category). The regressions model was used in order to assess association between inclination to smoking and diurnal type groups.

The researchers also conducted a cross-sectional analyses of the smoking statuses (amount of cigarettes smoked) in the three surveys that had been conducted. This was in order to extract possible differences between the three groups (participants of the three surveys)

What conclusions did the study reach? Are the conclusions appropriate? Why or why not?

Researchers found that of the four categories, it was diurnal types who were most likely to be past and current nicotine smokers and addicted to smoking, as well as less likely to relinquish the addiction. Even adjusting for potential confounders did not change these results.

The statistical tests accurately reflect the situation. Only a segment of twins was chosen therefore 'random effects' model was used.

The employed instruments were reliable; the only problem seems to be whether one can accept self-report of amount of cigarettes used. Either way, the statistical tests accurately reflected the situation with logistic regression being used in order to test for an association between diurnal type and nicotine dependence / smoking cessation. Diurnal type was the independent variable and nicotine dependence or smoking cessation the dependent variable since researchers were testing to see whether a certain part of the day prompted more tendencies to smoke. "Linear regression was used when the dependent variable was continuous FTND." (FTND: Fagerstrom Test for Nicotine Dependence). This was also the correct model to use since only one dependent variable is in question here (namely FTND) and researchers wanted to assess where possible association existed between diurnal type and FTND.

Because reporters were working with clustered rather than individual data (i.e. they were working with twins), they used the 'cluster' option in their statistical software which was Stata 9.0. This enabled them to be more correct in their standard errors and p-values.

Are the findings statistically significant? Why or why not? Describe the process you used to make this determination and provide the level of significance.

According to Tables 3 and 4, researchers found that evening types of both genders were much more likely to be current (OR= 2.91, 95% CI 2.50, 3.38) and lifetime smokers (OR=2.67, 95% CI 2.96, 4.07) than that of any other diurnal category. Nicotine dependence (as measured by the FTND test; linear regression Table 5) also showed that this was higher among evening types than any other diurnal categories. Evening types scored 0.59 (95% CI 0.01, 1.17) points higher than morning types on the FTND. The Odds are low in each case (2.91 / 2.67). CI (confidence interval) in each case is high (the statistical pack will show ." 05" as indicating a 5% chance of something not being true. This is higher which shows that there seems to be great significance here, in fact a 98% chance (1-.02=.98) of it being true. The various regressions were also close to one another.

Elaborations on my cognitive process are described below:

The descriptive results seem accurate (see Table 1: Proportions of current and ever smokers by diurnal type among all subjects, and men and women, in the three surveys). Logistic regression (Table 2) was then done of the descriptive data. Logistic regression odds ratios were performed with the intention of investigating the chance of the odds of one diurnal type being addicted to smoking more than another. This was the appropriate tool to use. 95% confidence intervals were chosen for the odds. Analyses were adjusted for age and gender which was also good since researchers recognized that age and gender may influence results. Other potential confounders that researchers included were life satisfaction and heavy drinking. The odds fell only slightly with these variables.

I fail to understand how heavy drinking can be a potential confounder in smoking. On the other hand, I see many other confounders that researchers ignored such as job, demands of job, marital status, and so forth. 'Life satisfaction' is also a vague eponym that is difficult to measure. These are two problems that I see with possibly effecting accuracy of results. On the other hand, to be even more thorough and to correct for this problem, researchers used two measures of mood: diagnosis of depression (the Beck Depression Inventory) in the Nicotine Addiction Genetics Finland study data and life satisfaction in the Finnish Twin Cohort