. 2017 Jun 3;165(1):193–200. doi: 10.1007/s10549-017-4325-2

Table 4.

Likelihood ratio test results comparing Cox-proportional hazards breast cancer risk models for different methods, and combinations of methods, of categorizing family history

Model	Goodness of fit		Likelihood ratio test^b ( $χ_{d f}^{2}$ ), P value
Model	df	χ ^2*	Likelihood ratio test^b ( $χ_{d f}^{2}$ ), P value
(a) Binary family history (Yes/No)	1	94.57
(b) No. of relatives with breast cancer (0, 1, ≥ 2)	2	99.29
(c) Type of relative with breast cancer (No, mother, sister, daughter, combination of relatives)	4	99.55
(d) Age of relative at breast cancer diagnosis (No, < 45, ≥ 45)	2	106.41
(e) Family History Score (FHS)^a	1	102.42
Combinations
(a + e) Binary family history added to FHS	2	106.51	$χ_{1}^{2}$ = 4.08, P = 0.04
(e + a) FHS added to Binary family history	2	106.51	$χ_{1}^{2}$ = 11.94, P = 0.0005
(b + e) No. of relatives with breast cancer added to FHS	3	110.14	$χ_{2}^{2}$ = 7.72, P = 0.02
(e + b) FHS added to No. of relatives with breast cancer	3	110.14	$χ_{1}^{2}$ = 10.86, P = 0.001
(c + e) Type of relative with breast cancer added to FHS	5	112.74	$χ_{4}^{2}$ = 10.31, P = 0.36
(e + c) FHS added to Type of relative with breast cancer	5	112.74	$χ_{1}^{2}$ = 13.18, P = 0.0003
(d + e) Age of relative at breast cancer diagnosis added to FHS	3	112.63	$χ_{2}^{2}$ = 10.20, P = 0.006
(e + d) FHS added to age of relative at breast cancer diagnosis	3	112.63	$χ_{1}^{2}$ = 6.22, P = 0.013

^*Variation accounted for by adding variables to model already including age at menarche, benign breast disease, oral contraceptive use, parity, age at first birth, breastfeeding, age at menopause, hormone replacement therapy use, physical activity, pre- and post-menopausal body mass index, alcohol intake, smoking status, and socioeconomic status

^aTrend across six groups

^bTest of improvement to the fit of the model by addition of alternative method for describing family history