Table 3.
Association between IHC4 score and risk of distant metastasis, overall, and by tumor subtype
| Tumor type | IHC4 category | Case subjects, No.* | Control subjects, No. | Odds ratio (95% confidence interval)† | |
|---|---|---|---|---|---|
| Univariate | Multivariable | ||||
| All subjects | ≤−7.81 | 49 | 86 | 1.00 (referent) | 1.00 (referent) |
| >−7.81 to 88.32 | 96 | 87 | 2.05 (1.27 to 3.32) | 1.50 (0.84 to 2.67) | |
| >88.32 | 113 | 85 | 2.53 (1.55 to 4.11) | 1.05 (0.55 to 2.01) | |
| P trend | <.001 | .93 | |||
| Continuous, per 10 units | 1.06 (1.03 to 1.09) | 1.01 (0.97 to 1.05) | |||
| Triple-negative‡ | Continuous, per 10 units | 0.91 (0.74 to 1.12) | 0.90 (0.66 to 1.22) | ||
| ER+/HER2− | ≤−7.81 | 49 | 71 | 1.00 (referent) | 1.00 (referent) |
| >−7.81 to 88.32 | 84 | 69 | 1.76 (1.10 to 2.84) | 1.62 (0.94 to 2.81) | |
| >88.32 | 14 | 8 | 2.54 (0.97 to 6.62) | 1.61 (0.48 to 5.47) | |
| P trend | .01 | .12 | |||
| Continuous, per 10 units | 1.09 (1.03 to 1.15) | 1.06 (1.00 to 1.13) | |||
| HER2+ | Continuous, per 10 units | 1.06 (0.94 to 1.20) | 1.07 (0.92 to 1.24) | ||
* The numbers of case subjects and control subjects shown in the triple-negative, estrogen receptor (ER)+/human epidermal growth factor receptor (HER2)−, and HER2+ subgroups are the number of unique individuals in the various categories, reflecting the fact that analyses in these subgroups were conducted using unconditional logistic regression. The 518 records in the study derive from 481 unique study subjects (because of matching with replacement), including 222 unique control subjects; of the 259 control subjects, 189 of them served as control subjects once, 29 served as control subjects twice, and 4 served as control subjects three times (189 + 58 + 12 = 259). Two subjects had no ER data and therefore cannot be classified with respect to triple-negative or ER+/HER2− status, and 11 subjects did not fall into one of these three subgroups. The 104 observations in the dataset that were classified as triple-negative derived from 98 unique subjects; the 321 ER+/HER2− observations in the dataset derived from 295 unique subjects; and the 78 HER2+ observations in the dataset derived from 75 unique subjects.
† Results for all subjects combined were obtained using conditional logistic regression models and results for breast cancer subtypes were obtained using unconditional logistic regression models. All multivariable analyses were adjusted for lymph node status, tumor size, tumor grade, and hormone therapy categorized as shown in Table 1; subgroup analyses additionally adjusted for age at diagnosis and duration of follow-up. All statistical tests were two-sided.
‡ Triple-negative breast cancer is breast cancer that is ER−, progesterone receptor (PR)−, and HER2−.