Automated Quantitative Measures of Terminal Duct Lobular Unit Involution and Breast Cancer Risk

Kevin H. Kensler; Emily Z.F. Liu; Suzanne C. Wetstein; Allison M. Onken; Christina I. Luffman; Gabrielle M. Baker; Laura C. Collins; Stuart J. Schnitt; Vanessa C. Bret-Mounet; Mitko Veta; Josien P.W. Pluim; Ying Liu; Graham A. Colditz; A. Heather Eliassen; Susan E. Hankinson; Rulla M. Tamimi; Yujing J. Heng

doi:10.1158/1055-9965.EPI-20-0723

DOI: 10.1158/1055-9965.EPI-20-0723 Published November 2020

Article Figures & Data

Figures

Tables
Additional Files

Download figure
Open in new tab
Download powerpoint
Figure 1.
A panel of a region of a WSI describing our method and how the eight quantitative TDLU measures are calculated. A, A region of a WSI. B, Our computational pathology method segments TDLUs (purple areas), detects acini (blue dots), and segments adipose tissue (yellow areas). Quantitative TDLU involution measures investigated in this study consisted of the three standardized measures [median TDLU span (C), TDLU counts per nonadipose tissue area (D), and median acini counts per TDLU (E)], and five novel measures [median TDLU area (F), total TDLU area as a percentage of tissue area and nonadipose tissue area (G), total number of acini (detected in TDLUs) per nonadipose tissue area (H), and median acini density (I)].
Download figure
Open in new tab
Download powerpoint
Figure 2.
TDLU involution was evaluated among 827 controls using qualitative categories: no type 1 lobules (n = 114), mixed lobule types (n = 409), and predominant type 1 and no type 3 lobules (n = 334). TDLU involution was significantly correlated with age at BBD biopsy (P_trend < 0.001; A) and significantly inversely correlated with the eight quantitative measures derived from our automated method (P_trend < 0.001; B–I).

Table 1.

Participants' characteristics in this study.

	Cases, n (%)	Controls, n (%)
N	287	1,083
Age at BBD biopsy
<40 years	76 (26.5)	244 (22.5)
40–49 years	131 (45.6)	431 (39.8)
50–59 years	56 (19.5)	272 (25.1)
≥60 years	24 (8.4)	136 (12.6)
Year of BBD biopsy
Before 1970	30 (10.4)	55 (5.1)
1970–1979	78 (27.2)	224 (20.7)
1980–1989	128 (44.6)	475 (43.9)
After 1989	51 (17.8)	329 (30.4)
Age at breast cancer diagnosis/index date
<45 years	41 (14.3)	197 (18.2)
45–54 years	111 (38.7)	361 (33.3)
≥55 years	135 (47.0)	525 (48.5)
Years between BBD biopsy and breast cancer diagnosis/index date
0.5–4.9 years	85 (29.6)	501 (46.3)
5.0–9.9 years	101 (35.2)	274 (25.3)
10.0–14.9 years	54 (18.8)	169 (15.6)
≥15.0 years	47 (16.4)	139 (12.8)
BBD histologic subtype
Nonproliferative	59 (20.6)	303 (28.0)
Proliferative without atypia	149 (51.9)	625 (57.7)
Atypical hyperplasia	79 (27.5)	155 (14.3)

Abbreviation: BBD, benign breast disease.

Table 2.

Quantitative TDLU measures and breast cancer risk factors among 1,083 controls.

	n	Median TDLU span (μm)	TDLU counts/mm²	Median acini counts/TDLU	Median TDLU area (mm²)
Age at BBD biopsy
<40 years	244	0.56 (0.55–0.58)	0.48 (0.44–0.52)	7.56 (7.11–8.04)	0.11 (0.10–0.11)
40–49 years	431	0.52 (0.51–0.53)	0.49 (0.46–0.51)	7.52 (7.18–7.87)	0.09 (0.09–0.10)
50–59 years	272	0.47 (0.45–0.48)	0.43 (0.40–0.46)	5.32 (5.02–5.64)	0.07 (0.07–0.07)
≥60 years	136	0.46 (0.44–0.47)	0.38 (0.34–0.42)	4.33 (3.99–4.71)	0.07 (0.06–0.07)
P		<0.001	<0.001	<0.001	<0.001
BBD histologic subtype
Nonproliferative	303	0.49 (0.48–0.51)	0.43 (0.40–0.46)	6.40 (6.05–6.78)	0.08 (0.08–0.09)
Proliferative without atypia	625	0.51 (0.50–0.52)	0.46 (0.44–0.48)	6.36 (6.12–6.62)	0.09 (0.08–0.09)
Atypical hyperplasia	155	0.51 (0.50–0.53)	0.48 (0.44–0.53)	6.86 (6.33–7.43)	0.09 (0.08–0.10)
P		0.06	0.14	0.25	0.04
Body size at ages 5–10 years
Level 1	308	0.51 (0.50–0.53)	0.46 (0.43–0.49)	6.58 (6.22–6.97)	0.09 (0.08–0.09)
Level 1.5 to 2	276	0.51 (0.50–0.52)	0.45 (0.42–0.48)	6.66 (6.27–7.07)	0.09 (0.08–0.09)
Level ≥2.5	353	0.50 (0.49–0.51)	0.46 (0.44–0.49)	6.12 (5.80–6.45)	0.08 (0.08–0.09)
P		0.44	0.84	0.07	0.10
BMI (kg/m²)
<25	612	0.51 (0.50–0.52)	0.45 (0.43–0.47)	6.52 (6.26–6.78)	0.09 (0.08–0.09)
25 to <30	288	0.50 (0.49–0.52)	0.45 (0.42–0.48)	6.62 (6.24–7.01)	0.08 (0.08–0.09)
≥30	169	0.51 (0.49–0.52)	0.48 (0.44–0.52)	5.90 (5.47–6.36)	0.08 (0.08–0.09)
P		0.88	0.44	0.04	0.34
Age of menarche
≤12 years	515	0.50 (0.49–0.51)	0.46 (0.44–0.49)	6.46 (6.19–6.75)	0.08 (0.08–0.09)
13 years	317	0.51 (0.50–0.52)	0.45 (0.42–0.48)	6.51 (6.16–6.88)	0.09 (0.08–0.09)
≥14 years	246	0.51 (0.50–0.53)	0.45 (0.42–0.49)	6.29 (5.91–6.70)	0.09 (0.08–0.09)
P		0.55	0.79	0.70	0.84
Parity
Nulliparous	101	0.48 (0.46–0.51)	0.33 (0.30–0.38)	5.20 (4.71–5.73)	0.08 (0.07–0.08)
Parous	978	0.51 (0.50–0.52)	0.47 (0.45–0.49)	6.58 (6.38–6.79)	0.09 (0.08–0.09)
P		0.03	<0.001	<0.001	0.03
Age at first birth among parous women
<25 years	535	0.50 (0.49–0.51)	0.46 (0.44–0.49)	6.55 (6.28–6.83)	0.08 (0.08–0.09)
25–29 years	347	0.51 (0.50–0.52)	0.49 (0.46–0.52)	6.42 (6.10–6.76)	0.09 (0.08–0.09)
≥30 years	99	0.52 (0.50–0.54)	0.44 (0.39–0.50)	6.77 (6.14–7.46)	0.09 (0.08–0.10)
P		0.22	0.28	0.63	0.20
Birth index among parous women
≤30	226	0.52 (0.50–0.53)	0.47 (0.43–0.51)	6.43 (5.99–6.91)	0.09 (0.08–0.10)
31–59	275	0.52 (0.50–0.53)	0.48 (0.45–0.52)	7.29 (6.86–7.74)	0.09 (0.09–0.10)
≥60	218	0.52 (0.50–0.53)	0.51 (0.47–0.56)	7.05 (6.56–7.58)	0.09 (0.08–0.09)
P		0.97	0.35	0.04	0.82
Breastfeeding among parous women
Never	390	0.50 (0.49–0.51)	0.46 (0.43–0.49)	6.39 (6.09–6.70)	0.08 (0.08–0.09)
<6 months	203	0.50 (0.49–0.52)	0.48 (0.44–0.52)	6.85 (6.41–7.32)	0.09 (0.08–0.09)
≥6 months	292	0.51 (0.50–0.53)	0.47 (0.44–0.51)	6.50 (6.15–6.88)	0.09 (0.08–0.09)
P		0.43	0.70	0.25	0.65
Menopausal status
Pre	667	0.52 (0.51–0.53)	0.49 (0.46–0.52)	7.35 (7.03–7.68)	0.09 (0.09–0.09)
Post	332	0.49 (0.47–0.50)	0.39 (0.36–0.43)	4.93 (4.59–5.29)	0.08 (0.07–0.08)
P		<0.01	<0.01	<0.001	<0.001

	% TDLU area (total)	% TDLU area (nonadipose)	Acini counts/mm²	Median acini density
Age at BBD biopsy
<40 years	11.15 (9.95–12.50)	13.08 (11.69–14.64)	9.27 (8.21–10.46)	80.88 (76.82–85.16)
40–49 years	7.87 (7.22–8.57)	9.64 (8.86–10.49)	7.64 (6.97–8.37)	90.92 (87.46–94.51)
50–59 years	4.90 (4.39–5.45)	6.19 (5.56–6.88)	4.41 (3.93–4.95)	83.00 (79.05–87.15)
≥60 years	3.39 (2.91–3.95)	4.45 (3.83–5.17)	2.80 (2.38–3.29)	73.01 (68.14–78.23)
P	<0.001	<0.001	<0.001	<0.001
BBD histologic subtype
Nonproliferative	4.99 (4.52–5.52)	6.26 (5.67–6.92)	5.01 (4.49–5.58)	87.47 (83.44–91.70)
Proliferative without atypia	7.59 (7.09–8.14)	9.31 (8.69–9.97)	6.52 (6.04–7.03)	82.51 (79.86–85.25)
Atypical hyperplasia	7.94 (6.90–9.15)	9.71 (8.45–11.17)	7.10 (6.09–8.28)	84.80 (79.34–90.64)
P	<0.001	<0.001	<0.001	0.13
Body size at ages 5–10 years
Level 1	7.45 (6.73–8.25)	9.16 (8.29–10.13)	6.66 (5.97–7.43)	84.80 (80.90–88.88)
Level 1.5–2	6.97 (6.26–7.75)	8.49 (7.64–9.44)	6.15 (5.48–6.91)	83.78 (79.72–88.03)
Level ≥2.5	6.41 (5.83–7.05)	7.98 (7.27–8.76)	5.82 (5.26–6.45)	83.00 (79.44–86.72)
P	0.10	0.14	0.21	0.81
BMI (kg/m²)
<25	6.84 (6.36–7.35)	8.34 (7.77–8.96)	6.10 (5.64–6.59)	83.86 (81.12–86.69)
25 to <30	6.85 (6.17–7.61)	8.46 (7.62–9.39)	6.36 (5.68–7.13)	87.02 (82.89–91.35)
≥30	6.49 (5.66–7.43)	8.28 (7.24–9.48)	5.73 (4.95–6.64)	80.77 (75.84–86.02)
P	0.78	0.96	0.54	0.18
Age of menarche
≤12 years	6.74 (6.24–7.29)	8.33 (7.71–9.00)	6.13 (5.64–6.67)	85.23 (82.21–88.36)
13 years	6.90 (6.25–7.63)	8.53 (7.73–9.41)	6.24 (5.61–6.95)	84.30 (80.51–88.26)
≥14 years	6.78 (6.06–7.59)	8.31 (7.43–9.29)	5.94 (5.26–6.71)	81.53 (77.39–85.90)
P	0.94	0.92	0.83	0.39
Parity
Nulliparous	4.41 (3.70–5.26)	5.44 (4.57–6.47)	3.52 (2.92–4.25)	73.92 (68.11–80.21)
Parous	7.10 (6.71–7.51)	8.75 (8.28–9.25)	6.48 (6.10–6.88)	85.33 (83.14–87.58)
P	<0.001	<0.001	<0.001	<0.01
Age at first birth among parous women
<25 years	6.64 (6.15–7.16)	8.21 (7.62–8.85)	6.15 (5.67–6.66)	86.92 (83.92–90.04)
25–29 years	7.51 (6.83–8.24)	9.31 (8.48–10.21)	6.73 (6.09–7.45)	84.03 (80.44–87.79)
≥30 years	6.89 (5.78–8.22)	8.36 (7.03–9.94)	6.27 (5.19–7.56)	81.34 (74.95–88.28)
P	0.13	0.11	0.38	0.24
Birth index among parous women
≤30	7.48 (6.60–8.48)	9.20 (8.14–10.41)	6.64 (5.79–7.61)	80.05 (75.45–84.92)
31–59	7.77 (7.00–8.63)	9.57 (8.63–10.61)	7.24 (6.46–8.12)	88.67 (84.38–93.18)
≥60	7.95 (7.01–9.02)	9.74 (8.61–11.03)	7.70 (6.71–8.84)	90.27 (85.04–95.81)
P	0.82	0.83	0.37	0.01
Breastfeeding among parous women
Never	6.71 (6.14–7.33)	8.33 (7.64–9.09)	6.06 (5.51–6.66)	84.89 (81.52–88.41)
<6 months	6.88 (6.09–7.78)	8.48 (7.51–9.57)	6.54 (5.74–7.46)	89.52 (84.62–94.69)
≥6 months	7.38 (6.67–8.18)	9.11 (8.23–10.07)	6.70 (6.01–7.48)	84.58 (80.70–88.64)
P	0.37	0.41	0.35	0.25
Menopausal Status
Pre	7.76 (7.16–8.42)	9.52 (8.78–10.31)	7.49 (6.86–8.17)	89.94 (86.64–93.35)
Post	5.40 (4.73–6.15)	6.70 (5.89–7.63)	4.13 (3.59–4.75)	72.15 (67.93–76.63)
P	<0.001	<0.001	<0.001	<0.001

Note: Data presented for age are means (95% CI). Data for other variables are presented as age-adjusted means (95% CI); age was adjusted as a continuous variable. Bold text represents values of statistical significance (P < 0.05).
Abbreviation: BBD, benign breast disease.

Table 3.

The association between automated TDLU measures and breast cancer risk was evaluated using unconditional logistic regression models to estimate ORs and 95% CIs.

	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P_trend
Median TDLU span
Cases/Controls, n	65/271	72/270	73/271	77/271
Model 1	Ref	1.07 (0.73–1.57)	0.98 (0.67–1.45)	0.96 (0.65–1.43)	0.75
Model 2	Ref	0.94 (0.64–1.39)	0.92 (0.62–1.37)	0.89 (0.59–1.33)	0.56
TDLU counts/mm²
Cases/Controls, n	67/271	73/270	71/271	76/271
Model 1	Ref	1.10 (0.75–1.60)	1.06 (0.73–1.55)	1.17 (0.80–1.71)	0.45
Model 2	Ref	1.04 (0.71–1.53)	0.96 (0.65–1.41)	1.15 (0.79–1.69)	0.49
Median acini counts/TDLU
Cases/Controls, n	26/121	79/348	89/311	93/303
Model 1	Ref	1.02 (0.63–1.71)	1.21 (0.75–2.02)	1.19 (0.73–1.99)	0.40
Model 2	Ref	0.94 (0.57–1.57)	1.00 (0.61–1.69)	1.05 (0.64–1.77)	0.59
Median TDLU area
Cases/Controls, n	58/271	78/270	66/271	85/271
Model 1	Ref	1.29 (0.88–1.91)	0.99 (0.66–1.49)	1.19 (0.79–1.78)	0.72
Model 2	Ref	1.15 (0.78–1.71)	0.87 (0.57–1.31)	1.10 (0.73–1.66)	0.90
% TDLU area (total)
Cases/Controls, n	58/271	82/270	63/271	84/271
Model 1	Ref	1.34 (0.92–1.98)	0.99 (0.66–1.48)	1.18 (0.79–1.78)	0.82
Model 2	Ref	1.15 (0.78–1.71)	0.86 (0.57–1.30)	1.04 (0.69–1.58)	0.90
% TDLU area (nonadipose)
Cases/Controls, n	58/271	87/270	57/271	85/271
Model 1	Ref	1.42 (0.97–2.08)	0.89 (0.59–1.34)	1.22 (0.82–1.82)	0.81
Model 2	Ref	1.23 (0.84–1.82)	0.80 (0.52–1.21)	1.10 (0.73–1.66)	0.98
Acini counts/mm²
Cases/Controls, n	64/271	71/270	69/271	83/271
Model 1	Ref	1.09 (0.74–1.61)	0.99 (0.67–1.47)	1.10 (0.75–1.63)	0.72
Model 2	Ref	0.98 (0.67–1.45)	0.86 (0.58–1.29)	1.03 (0.69–1.53)	0.83
Median acini density
Cases/Controls, n	57/271	89/270	61/271	80/271
Model 1	Ref	1.59 (1.09–2.33)	1.08 (0.72–1.62)	1.33 (0.91–1.96)	0.49
Model 2	Ref	1.54 (1.05–2.27)	1.05 (0.70–1.58)	1.36 (0.92–2.01)	0.41

Note: Each quantitative TDLU measure was categorized into quartiles as defined by the distribution among the controls. Model 1 adjusted for matching factors. Model 2 adjusted for matching factors and BBD histologic subtypes. The median value for each quartile was included as a continuous variable in the unconditional logistic regression for Model 1 and 2 to obtain the P_trend value (Wald test).

Table 4.

The association between manual qualitative TDLU categories and breast cancer risk was evaluated using unconditional logistic regression models to estimate ORs and 95% CIs.

	Cases/Controls n	Model 1	Model 2
Among women with both quantitative and qualitative data
No type 1 lobules	23/114	Ref	Ref
Mixed lobule types	93/409	1.25 (0.76–2.14)	1.15 (0.69–1.98)
Predominant type 1 and no type 3 lobules	61/334	0.95 (0.54–1.70)	0.95 (0.54–1.71)

No type 1 lobules or mixed lobule types	116/523	Ref	Ref
Predominant type 1 and no type 3 lobules	61/334	0.79 (0.54–1.13)	0.84 (0.58–1.22)

Women in full BBD nested case–control study with qualitative data
No type 1 lobules	41/192	Ref	Ref
Mixed lobule types	151/640	1.17 (0.79–1.74)	1.09 (0.74–1.65)
Predominant type 1 and no type 3 lobules	96/542	0.82 (0.53–1.28)	0.86 (0.56–1.35)

No type 1 lobules or mixed lobule types	192/832	Ref	Ref
Predominant type 1 and no type 3 lobules	96/542	0.72 (0.54–0.96)	0.80 (0.59–1.07)

Note: Model 1 adjusted for matching factors. Model 2 adjusted for matching factors and BBD histologic subtypes.
Abbreviation: BBD, benign breast disease.

Additional Files

Figures
Tables

Supplementary Data

Supplementary Figure S1 - Correlation matrix of quantitative terminal duct lobular unit involution measures with age at benign breast disease diagnosis among 1083 controls. The majority of the Spearman's correlations were significant at p<0.05 (black text). Correlations that did not achieve significance (p>0.05) are represented by grey text.
Supplementary Table 1-8 - Supplementary Table S1. The eight terminal duct lobular unit (TDLLU) involution measures in this study were calculated by combining data from multiple images for each participant and the following values extracted from the computational method: tissue areas (total, adipose, non-adipose; mm2), TDLU area (mm2), TDLU span (Âµm), and number of acini per TDLU. Supplementary Table S2. The association between quantitative terminal duct lobular units (TDLUs) and parity, with parous women further subdivided by number of births, and time between last birth and benign breast disease (BBD) biopsy. Data are presented as age-adjusted means (95% confidence interval); age was adjusted as a continuous variable. Supplementary Table S3. Quantitative terminal duct lobular unit (TDLU) measures, age at menopause, and elapsed time between menarch and menopause among 332 post-menopausal controls. Data presented for age are means (95% confidence interval). Data for other variables are presented as age-adjusted means (95% confidence interval); age was adjusted as a continuous variable. Supplementary Table S4. The association between automated terminal duct lobular unit (TDLU) measures and breast cancer risk was evaluated using unconditional logistic regression models to estimate odd ratios (ORs) and 95% confidence intervals (CI). Supplementary Table S5. Automated terminal duct lobular unit (TDLU) measures and breast cancer risk stratified by parity. Data were analyzed using unconditional logistic regression models adjusting for the matching factors and BBD histological subtypes to estimate odd ratios (ORs) and 95% confidence intervals (CI). Supplementary Table S6. Automated terminal duct lobular unit (TDLU) measures and breast cancer risk stratified by menopausal status. Data were analyzed using unconditional logistic regression models adjusting for the matching factors and BBD histological subtypes to estimate odd ratios (ORs) and 95% confidence intervals (CI). Supplementary Table S7. Automated terminal duct lobular unit (TDLU) measures and breast cancer risk stratified by benign breast disease (BBD) histological subtypes. Data were analyzed using unconditional logistic regression models adjusting for the matching factors to estimate odd ratios (ORs) and 95% confidence intervals (CI). Supplementary Table S8. Automated terminal duct lobular unit (TDLU) measures and breast cancer risk defined by tumor estrogen receptor (ER) expression. Data were analyzed using polytomous logistic regression models to estimate odd ratios (ORs) and 95% confidence intervals (CI).