Table 2 Sensitivity analyses of the HER2DX risk score for RFS and OS in the overall cohort and in the node-negative subgroup

From: HER2DX in older patients with HER2-positive early breast cancer: extended follow-up from the RESPECT trial of trastuzumab ± chemotherapy

Outcome

Group

Time (yrs)

Low-risk estimate (95% CI)

High-risk estimate (95% CI)

HR (95% CI)

RFS

All patients

5

91.7 (86.7–97.1)

77.2 (65.1–91.5)

0.35 (0.1–0.9)

  

6

88.6 (82.7–94.9)

74.2 (61.6–89.4)

0.41 (0.2–0.9)

  

7

86.5 (80.1–93.4)

68.0 (54.4–85.0)

0.38 (0.2–0.8)

  

8

86.5 (80.1–93.4)

68.0 (54.4–85.0)

0.38 (0.2–0.8)

  

9

83.7 (76.6–91.5)

68.0 (54.4–85.0)

0.43 (0.2–0.9)

  

10

81.2 (72.9–90.4)

68.0 (54.4–85.0)

0.46 (0.2–1.0)

 

Node-negative

5

91.4 (86.2–96.9)

72.6 (54.6–96.4)

0.30 (0.1–0.9)

  

6

89.2 (83.4–95.2)

72.6 (54.6–96.4)

0.36 (0.1–1.0)

  

7

87.0 (80.7–93.9)

58.6 (38.9–88.4)

0.28 (0.1–0.7)

  

8

87.0 (80.7–93.9)

58.6 (38.9–88.4)

0.28 (0.1–0.7)

  

9

84.1 (76.8–92.0)

58.6 (38.9–88.4)

0.31 (0.1–0.8)

  

10

81.4 (72.8–90.9)

58.6 (38.9–88.4)

0.32 (0.1–0.8)

OS

All patients

5

97.3 (94.4–100)

83.9 (72.9–96.6)

0.17 (0.0–0.7)

  

6

95.4 (91.6–99.4)

80.8 (68.9–94.8)

0.23 (0.1–0.7)

  

7

93.5 (88.9–98.3)

74.3 (61.1–90.5)

0.23 (0.1–0.6)

  

8

93.5 (88.9–98.3)

74.3 (61.1–90.5)

0.23 (0.1–0.6)

  

9

90.9 (85.4–96.9)

74.3 (61.1–90.5)

0.29 (0.1–0.7)

  

10

88.7 (81.5–96.1)

69.7 (55.2–88.1)

0.30 (0.1–0.7)

 

Node-negative

5

97.2 (93.4–100)

72.2 (59.7–99.8)

0.12 (0.0–0.5)

  

6

95.2 (91.3–99.6)

70.8 (57.5–99.9)

0.19 (0.0–0.9)

  

7

94.2 (89.8–98.8)

63.7 (43.9–92.3)

0.14 (0.0–0.4)

  

8

94.2 (89.8–98.8)

63.7 (43.9–92.3)

0.14 (0.0–0.4)

  

9

91.6 (86.0–97.5)

63.7 (43.9–92.3)

0.18 (0.1–0.5)

  

10

89.2 (82.3–96.7)

47.7 (24.3–93.9)

0.16 (0.1–0.4)

  1. These analyses are exploratory and were conducted to descriptively assess the consistency of the prognostic effect of the HER2DX risk score over time. They are not intended for formal statistical inference. Proportional hazards assumptions were verified in the primary model. Estimates were also derived for the node-negative subgroup.