One purpose of the present study was to compare analyses of AmBP with our previously published parallel analysis of hypertension. Accordingly, we reexamined the effects of job strain on hypertension and left ventricular mass index (LVMI) now that we have added an eighth site to our study (see Table 2).
| Effect | Odds Ratio | Effect size (g/m2) | x2 | F | p value* |
| Case Status (N=264) | |||||
| Job strain | 2.7 | 5.90 | 0.015 | ||
| Left ventricular mass index (N=203) | |||||
| Job Strain | 9.7 | 3.37 | 0.001 | ||
| *Two tailored probability levels |
After controlling for the effects of age, BMI, BMI stratum,
Type A behavior, 24 hour sodium excretion, physical activity level
of the job, education level, smoking status, alcohol intake, and
work site, job strain is a significant predictor of case-control
status (estimated odds ratio, OR=2.7; p=0.015). After excluding
those subjects receiving anti-hypertensive medication, 203 subjects
had technically satisfactory echocardiograms. Using an ANCOVA
model, we showed that the relation of job strain to LVMI was 9.7
g/m2 (F=3.37; p=0.001) after controlling for the same variables
as above. This relation with LVMI was consistent across the three
10-year age groups.
Our job strain model predicts that blood pressure will be elevated
in the high-strain quadrant of the model. ANCOVA (high strain
versus the three other quadrants combined) supports the job strain
hypothesis. The effect of job strain on systolic AmBP at work
is 6.8mm Hg (F=5.0; p=0.03) is also statistically significant.
It is worth noting that the three non-high strain quadrants are
similar to each other in AmBP.
Job strain has the same magnitude of effect on systolic AmBP during
at home and asleep hours as it does for working hours, demonstrating
a strong effect of job strain on 24-hour AmBP. We also examined
the effect of job strain as well as other independent variables
on work minus home differences an AmBP. AmBP’s were about
3.5 mm Hg higher at work compared with home, and no variable was
found to be related to this difference.
As expected, age and BMI have a large and substantive effects
on all measures of AmBP. Regular alcohol consumption has an effect
on both systolic and diastolic AmBP at work of about 3.6 mm Hg
(p=0.06) and 2.8 mm Hg (p=0.02), respectively. Cigarette smoking
has a main effect on systolic AmBP of 4 mm Hg at work (NS), 5.2
mm Hg at home (F=5.5; p=0.02), and 3.9 mm Hg while asleep (NS).
The effects of smoking on diastolic AmBP are consistently positive
but small and nonsignificant. However, in our study population,
Type A behavior and education level were not associated with a
significant increase in any measure of blood pressure after controlling
for the other known risk factors.
We next turn to the issue of whether job strain may interact with
the other significant predictors of AmBP at work. The global test
for the set of job strain interactions was highly significant
for work systolic AmBP (p=0.009) and not significant for diastolic
AmBP. The most significant interaction term for systolic AmBP
was job strain with alcohol (F=7.4, p=0.007). Workers not in high
strain jobs exhibit no relation between alcohol consumption and
systolic AmBP at work, whereas those in high strain jobs exhibit
a very substantial relation. Viewed from the opposite perspective,
there is a weak, presumably insignificant, 4-mm Hg effect of job
strain on systolic AmBP for low alcohol consumption and a large
effect (17 mm Hg) for regular consumption of alcohol. It is important
to note that, although this interaction suggests that either job
strain moderates the effect of alcohol on systolic AmBP, there
is no evidence that the effect of job strain is mediated through
alcohol consumption, e.g., that increased alcohol use is a response
to job strain. In fact, the proportion of heavy alcohol consumes
is the same for those in high-strain jobs (23%) as those in other
jobs.
The global test of the remaining job strain interaction terms
is again significant, and the most significant term is the interaction
with age (F=3.7; p=0.026). With each successive age cohort, job
strain is associated with a greater increase in systolic AmBP
at work. The difference between those in high-strain jobs is 15
mm Hg greater than in the youngest age cohort. Just as those not
in high-strain jobs showed no relation between alcohol and systolic
AmBP at work, they also show virtually no relation between age
and systolic AmBP at work.
To investigate the possible influence of our stratified recruitment
scheme for case subjects and control subjects, we weighted the
sample to approximate what would have been achieved if case subjects
and control subjects had been sampled proportionately at each
site. The ANCOVA results for this re-weighted analysis did not
differ substantially from the results presented above, supporting
the conclusion that our sampling scheme has not biased the relation
between job strain and AmBP in our target population. (We reestimated
the "unweighted" ANCOVA model, adding a dummy variable
for whether the subject was a case subject in our case-control
study, and found that this reduced the effect size of job strain
on AmBP by 25%. The overall pattern of relations between job strain
and AmBP was unchanged, and the systolic AmBP findings were still
statistically significant. We decided to present the results for
the entire group regardless of hypertension status in the model
since we believe hypertension status to be an intervening variable
between job strain and AmBP and therefore inappropriate to control
when testing the overall effects of job strain and the other significant
predictors (including the two-way interaction effects of job strain
with alcohol and age) were the same for case subjects and control
subjects was not significant (p>0.10) for both systolic and
diastolic AmBP at work.