Table 4.

Performance of our final predictors in predicting breast cancer using Fine and Gray Competing Risk Regression in Nurses’ Health Study stratified by age.^a, ^b

Breast cancer risk factors	NHS Ages 55–74 (n=53,362)			NHS Ages 75+ (n=19,704)
	Beta	HR	p-value	Beta	HR	p-value
Age at study entryb			0.15			0.92
55–59	0.00	1.00	-	-	-	-
60–64	0.14	1.15	0.14	-	-	-
65–69	0.20	1.22	0.04	-	-	-
70–74	0.22	1.25	0.03	-	-	-
75–80	-	-	-	0.00	1.00	-
80+	-	-	-	−0.01	0.99	0.92
Postmenopausal hormone use			<0.0001			0.002
never user	0.00	1.00	-	0.00	1.00	-
past E-alone user, <5 years	−0.06	0.95	0.66	0.07	1.07	0.68
past E-alone user, 5+ years	−0.13	0.88	0.19	0.06	1.06	0.69
past E+P user, <5 years	−0.01	0.99	0.92	−0.04	0.96	0.82
past E+P user, 5+ years	0.12	1.13	0.14	0.10	1.10	0.57
current E-alone user	0.15	1.16	0.12	0.56	1.75	0.0006
current E+P user	0.80	2.22	<.0001	0.91	2.49	0.003
unknown type of use	0.01	1.01	0.91	−0.03	0.97	0.85
Number of 1st degree relatives with history of breast cancer and age at diagnosis			<0.0001			<0.0001
None	0.00	1.00	-	0.00	1.00	-
1, diagnosed age<50	0.38	1.46	0.0006	0.48	1.62	0.008
1, diagnosed age >50/unknown	0.29	1.34	<.0001	0.17	1.19	0.19
2+, at least 1 diagnosed age <50	0.92	2.52	<.0001	0.91	2.49	0.0004
2+, 2+ diagnosed age>50/unknown	0.53	1.69	0.02	0.78	2.18	0.001
History of breast biopsy			<0.0001			0.05
0	0.00	1.00	-	0.00	1.00	-
1	0.38	1.46	<.0001	0.17	1.18	0.11
2+	0.46	1.59	<.0001	0.48	1.62	0.04
Highest Body Mass Index (BMI) in past 10 years			0.0001			0.004
<20 kg/m2	−0.25	0.78	0.22	−1.05	0.35	0.01
20–24 kg/m2	0.00	1.00	-	0.00	1.00	-
25–29 kg/m2	0.17	1.18	0.009	0.24	1.28	0.02
30+ kg/m2	0.30	1.35	<.0001	0.24	1.27	0.07
Unknown	0.20	1.22	0.78	0.95	2.59	0.19
Age at menopause (years)			0.003			<0.0001
<45	−0.36	0.70	0.0007	−0.44	0.64	0.05
45–49	0.00	1.00	-	0.00	1.00	-
50–54	0.01	1.01	0.91	0.22	1.24	0.06
55+	0.08	1.08	0.39	0.48	1.61	0.006
Unknown	−0.19	0.83	0.45	−9.49	0	<.0001
Age at first birth (years) and parity			0.01			0.68
Nulliparous	0.20	1.23	0.09	−0.07	0.93	0.80
<25, 1–2 children/unknown	0.00	1.00	-	0.00	1.00	-
<25, 3+ children	−0.10	0.90	0.18	0.24	1.28	0.19
25–29, 1–2 children/unknown	0.01	1.01	0.92	0.24	1.27	0.25
25–29, 3+ children	−0.08	0.92	0.37	0.12	1.13	0.52
30+, 1–2 children/unknown	0.24	1.28	0.04	0.33	1.40	0.14
30+, 3+ children	0.20	1.23	0.26	0.21	1.24	0.40
unknown age at first birth	−0.12	0.89	0.57	0.23	1.26	0.55
Average alcohol use per day (highest use in past 10 years)			<0.0001			0.89
0 gm/day	0.00	1.00	-	0.00	1.00	-
1–4.9 gm/day	0.02	1.02	0.73	0.06	1.06	0.65
5–14.9 gm/day	0.18	1.20	0.02	0.06	1.06	0.66
15+ gm/day	0.34	1.41	<.0001	0.0002	1	1.00
Unknown	0.01	1.01	0.94	0.22	1.25	0.34
Cigarette use			0.027			0.90
Never	0.00	1.00	-	0.00	1.00	-
Past	−0.08	0.92	0.14	−0.05	0.95	0.61
Current	0.17	1.19	0.06	−0.12	0.89	0.63
Unknown	−0.93	0.40	0.36	−0.43	0.65	0.67
Mammogram in past 2 years‡			0.68			0.06
No	0.00	1.00	-	0.00	1.00	-
Yes	−0.01	0.99	0.92	0.26	1.30	0.07
Unknown	0.12	1.12	0.48	−0.23	0.79	0.48
Comorbidityb
Limited in moderate daily activity			0.02			0.001
not at all limited	0.00	1.00	-	0.00	1.00	-
limited	−0.10	0.91	0.10	−0.34	0.71	0.0002
Unknown	−0.40	0.67	0.01	−0.39	0.68	0.27
Diabetes	0.06	1.06	0.48	0.24	1.27	0.07
Myocardial infarction	−0.16	0.85	0.45	−0.50	0.61	0.09
Stroke	−0.25	0.78	0.31	−0.44	0.65	0.14
Emphysema	−0.10	0.90	0.34	−0.06	0.94	0.71
Congestive heart failure	0.11	1.12	0.55	−0.18	0.83	0.41
c-statistics (95% CI)	0.62 (0.60–0.63)			0.64 (0.62–0.66)
AIC	34383.02			9845.98

^aAkaike Information Criterion (AIC) -is a function of the log-likelihood that adds a penalty of 2 for each additional factor.

^bFine and Gray competing risk regression (CRR) allows for estimating the probability of breast cancer conditional on competing risk-free survival. Specifically, CRR assigns a weight less than one to participants that have experienced a competing risk event or are loss to follow up and these weights decline with increasing time from the competing event and the event of interest.