Table 2.
Effect of surgery on hazard ratios for overall survival.
| Models | Sample size, No. | HR (95% CI) | |||
|---|---|---|---|---|---|
| Surgery | Non-surgery | Events | HR (95% CI) | P value | |
| Unadjusted model | 618 | 1,628 | 1467 | 0.572 (0.506–0.647) | <0.001 |
| Multivariable-adjusted model a | 618 | 1,628 | 1467 | 0.642 (0.557–0.740) | <0.001 |
| Propensity score-adjusted model b | |||||
| Within-propensity score quintile | |||||
| Quintile 1 (Lowest propensity score) | 23 | 427 | 324 | 0.792 (0.457–1.372) | 0.405 |
| Quintile 2 | 42 | 407 | 302 | 0.495 (0.316–0.776) | 0.002 |
| Quintile 3 | 86 | 363 | 300 | 0.614 (0.446–0.845) | 0.002 |
| Quintile 4 | 160 | 290 | 293 | 0.782 (0.606–1.010) | 0.060 |
| Quintile 5 (Highest propensity score) | 307 | 141 | 248 | 0.525 (0.395–0.696) | <0.001 |
| Combined c | 618 | 1,628 | 1,467 | 0.635 (0.548–0.736) | <0.001 |
| Regression adjustment | 618 | 1,628 | 1,467 | 0.649 (0.561–0.749) | <0.001 |
| Weighting (IPTW) | 618 | 1,628 | 1,467 | 0.673 (0.597–0.759) | <0.001 |
| Matching 1:1 | 475 | 475 | 598 | 0.690 (0.585–0.814) | <0.001 |
HR, hazard ratio; CI, confidence interval; IPTW, inverse probability of treatment weighting.
Adjusted for age, sex, ethnicity, marriage status, anatomic sites, laterality, tumor grading, year of diagnosis, AJCC stage, tumor size, extent of tumor, and lymph node involvement.
The propensity of receiving surgery was estimated using a multivariable logistic regression model that included baseline age, sex, race, marital status, anatomic sites, laterality, tumor grading, year of diagnosis, AJCC stage, tumor size, extent of tumor and lymph node involvement.
Results were combined among strata using inverse variance weights under a fixed model, as there was no detectable heterogeneity among strata (all P for heterogeneity = 0.2).