Therefore, we report results on models including these two covariates. Models for Shorter TTFC (within 5 min) In a model with HS ARQ197 and MD alone (not shown), the HR for heavy versus nonheavy smokers independent of MD was 2.6 (95% CI: 2.1�C3.3; p < .001), while the adjusted HR for MD was 1.9 (95% CI: 1.3�C2.7; p < .001) compared with the unadjusted HR of 3.7 (95% CI: 2.6�C5.3). Similarly, in a model with ��smoke more under stress�� and MD alone (not shown), the adjusted HR for ��smoke more under stress�� was 7.5 (95% CI: 5.5�C10.2; p < .001), while the HR for MD was reduced to 2.3 (95% CI: 1.5�C3.6; p < .001). In a model with all three variables, the effect of MD independent of HS and ��smoke more under stress�� was further reduced to 1.7 (95% CI: 1.1�C2.5) but remained statistically significant (p = .
02; see Table 3). Table 3. Crude and Adjusted Models for Risk of Short TTFC Defined With Different Cutoffs Among Smokers and Nonsmokers at Baseline (n = 11,705) Models for Longer TTFC (within 30 and 60 min) In a model predicting TTFC within 30 min with HS and MD alone (not shown), the HR for heavy versus nonheavy smokers was 2.1 (95% CI: 1.6�C2.7; p < .001), but the adjusted HR for MD was reduced to 1.0 (95% CI: 0.6�C1.5) and was no longer significant (p = .9) compared with unadjusted HR for MD of 2.1 (95% CI: 1.4�C3.1; p < .001). When controlling for both HS and ��smoke more under stress�� simultaneously, the HR for MD did not change in value and did not predict TTFC (HR = 1.0, 95% CI: 0.5�C1.7; p = .9). Additionally, both variables remained significant predictors of TTFC.
Similar results were obtained for MD (HR = .7, 95% CI: 0.4�C1.3; p = .6) when modeling TTFC within 60 min with the exception that after adjustment for HS status (HR = 1.7, 95% CI: 1.2�C2.8; p = .04), the HR for ��smoke more under stress�� was greatly reduced to a value of 1.1 (95% CI: 0.8�C1.6). This variable was no longer a significant predictor of TTFC (p = .4) and was removed from the model shown in Table 3. Similar results were obtained when number of CPD was modeled as continuous variable rather than dichotomous HS status, when CPD was modeled as nonlagged variable relative to TTFC ascertainment, and when distress was added to these models. These results are available upon request.
To assess the effects of MD on risk of shorter TTFC as a function of baseline smoking status, an interaction term between MD and smoking status was added to a model (not shown) with main effects of MD, HS, and smoking status. There was no evidence of effect modification (p = .8). Batimastat We also reran separate analyses restricted to baseline current smokers at risk for shorter TTFC and baseline never- or former smokers (see Supplementary Table). Our main results were found to be robust when the analysis was approached in these differing ways.