Table 3. Factors Associated With Mortality After Septal Myectomya.
| Variable | Comparisonb | Adjusted Hazard Ratio (95% CI)c | P Value |
|---|---|---|---|
| Age, y | 65.6:44.5 | 3.09 (2.12-4.52) | <.001 |
| Year of surgery | 2013:2004 | 0.82 (0.55-1.22) | .001 |
| BMI | 34:26 | 1.22 (0.90-1.66) | .001 |
| Diabetes mellitus | Yes:no | 1.57 (1.10-2.24) | .01 |
| New York Heart Association class | IV:I | 2.31 (1.03-5.15) | .04 |
| Amiodarone | Yes:no | 1.59 (1.00-2.54) | .05 |
| Pacemaker | Yes:no | 1.38 (0.98-1.95) | .07 |
| Nonsustained ventricular tachycardia | Yes:no | 0.70 (0.42-1.18) | .18 |
| Hypertension | Yes:no | 1.20 (0.91-1.57) | .19 |
| Disopyramide | Yes:no | 0.79 (0.52-1.18) | .25 |
| Use of angiotensin-converting enzyme inhibitors or angiotensin receptor blockers | Yes:no | 0.83 (0.59-1.17) | .29 |
| Presyncope | Yes:no | 1.12 (0.88-1.44) | .35 |
| Dyslipidemia | Yes:no | 1.12 (0.86-1.46) | .41 |
| Prior septal reduction | Yes:no | 1.27 (0.71-2.27) | .42 |
| Syncope | Yes:no | 1.11 (0.83-1.51) | .48 |
| Mitral valve regurgitation grade | Moderate:mild | 1.04 (0.93-1.16) | .49 |
| Race | Nonwhite:white | 0.68 (0.22-2.17) | .69 |
| Unknown:white | 1.17 (0.66-2.10) | ||
| β-Blocker | Yes:no | 0.95 (0.71-1.27) | .72 |
| Calcium-channel blocker | Yes:no | 1.04 (0.81-1.34) | .74 |
| Family history of hypertrophic cardiomyopathy | Yes:no | 0.94 (0.64-1.37) | .75 |
| Ethnicity | Hispanic:non-Hispanic | 1.17 (0.35-3.88) | .80 |
| Family history of sudden cardiac death | Yes:no | 1.06 (0.68-1.65) | .81 |
| Anteroseptal wall thickness, mm | 23:17 | 1.06 (0.80-1.40) | .84 |
| Sex | Female:male | 0.98 (0.76-1.26) | .86 |
| Implantable cardioverter defibrillator | Yes:no | 1.01 (0.56-1.81) | .97 |
Abbreviation: BMI, body mass index (calculated as weight in kilograms divided by height in meters squared).
A multivariable Cox regression model was fitted for all-cause mortality on 2328 participants using the variables shown; we further adjusted for 2 nonmodeled variables (atrial fibrillation and left ventricular ejection fraction) by stratifying the model on levels of each (with left ventricular ejection fraction grouped into quartiles), owing to failure of the proportional hazards assumption.
Continuous variables were expanded into multiple terms using restricted cubic splines (using 4 knots) to allow for nonlinear effects. Owing to data discreteness, a linear association was assumed for ordinal scales of New York Heart Association and mitral regurgitation grade. The reference value is the righthand value in each pair.
Effects of variables are estimated with adjusted hazard ratios (and 95% CIs). For continuous variables, hazard ratios were calculated comparing the 75th percentile with the 25th percentile. Since these variables are modeled nonlinearly, general tests of association can easily generate a significant P value, even when the confidence interval of a hazard ratio contains 1.0.