Method of tumor detection and clinicopathological features
In the main study series (N = 1,884), 408 patients (22%) had screen-detected breast cancer and 1,476 (78%) had non-screen-detected breast cancer. Mammography screening is organized mainly for women 50 to 69 years of age. In the current series, 484 patients (26%) were ages 50 to 59 years and 365 (19%) were ages 60 to 69 years. Of these patients, 254 (52%) and 93 (25%) were patients with screen-detected cancer, respectively, and 230 (48%) and 272 (75%) had non-screen-detected cancer, respectively.
The median tumor size for the entire cohort was 20 mm. The median tumor size was 13 mm among screen-detected tumors, and it was 20 mm in non-screen-detected tumors. Eighty-three percent of screen-detected tumors were 20 mm or less (T1), compared to only 53% of non-screen-detected tumors.
Nodal involvement was less frequent in patients with screen-detected breast cancer than in non-screen-detected breast cancer (21% vs 37%; P < 0.001), which was also true among women 50 to 69 years of age (20% vs 35%; P < 0.001). Among the screen-detected cancers, 66% of tumors were 20 mm or less and had no nodal involvement (T1N0), and only 36% of non-screen-detected cancers belonged to this category.
The majority of screen-detected tumors were ductal carcinomas (70%). However, the proportion of ductal carcinomas was significantly higher (75%; P < 0.04) in non-screen-detected cancers. The proportion of lobular carcinomas (16%) was equal in both diagnostic groups. Screen-detected tumors were more often of a special histological type (14% vs 9%; P < 0.04). Similar differences in histological profile were seen in the 50- to 69-year-old age cohort (ductal: 70% vs 73%; lobular: 16% vs 19%; special type: 14% vs 8%; P = 0.02).
Tumors detected by mammography screening were more often of a lower grade of differentiation. There were 129 grade 1 breast cancers (32%) in the screen-detected group and 241 (16%) among tumors detected by other methods (P
< 0.001). Correspondingly, screen-detected cancers were less often grade 2 or 3 tumors (n
= 202, 49%) than were non-screen-detected cancers (n
= 841, 57%) (P
< 0.001). However, this association weakened with increasing tumor size (Additional file 2
, Table S2). In T1N0 tumors, 36% of screen-detected tumors were well-differentiated, but only 22% in the non-screen-detected group were. In patients with larger tumors, such as T2 tumors (21 to 50 mm), the proportion of G1 tumors was equal in both diagnostic groups (11%) and the proportion of G2-3 tumors was 75% in screen-detected tumors compared to 64% in non-screen-detected tumors (P
Screen detection was associated with hormone receptor status (ER and PR), although in analyses consisting of all age groups, the association was weak (P = 0.038 and P = 0.094, respectively). In women at screening age (50 to 69 years), there were only 9% ER- tumors in screen-detected tumors compared to 19% in non-screen-detected tumors (P = 0.002). Similar differences were seen according to PR status (P < 0.001).
No statistically significant association between screen detection and HER2 gene amplification or p53 expression was seen in any of the analyzed age groups. Somewhat fewer Ki-67-positive tumors were seen in screen-detected breast cancers compared to those detected outside screening (18% vs 26%), but the association was not statistically significant (P = 0.08).
Adjuvant systemic therapy was given less frequently to patients with screen-detected breast cancer than to women with non-screen-detected breast cancer (22% vs 41%; P < 0.001). The use of adjuvant systemic therapy was not known in 5 patients (1%) with screen-detected cancers and in 40 patients (3%) with non-screen-detected cancers.
In women 50 to 69 years of age, screen detection was statistically significantly associated with smaller primary tumor size, less frequent lymph node metastasis, special histological type, lower grade of differentiation and positive hormone receptor status (ER and PR) compared with non-screen-detected cancers. The distribution of clinicopathological features in women with breast cancer diagnosed by screening mammography and by other methods is shown in Table .
Descriptive statistics for main analyses (N = 1,884)
Method of detection and breast cancer survival
Women with screen-detected breast cancer had more favorable BCSS than patients with non-screen-detected cancer. The 15-year BCSS rate for patients with screen-detected breast cancer was 86%, and for patients with non-screen-detected breast cancer it was 66% (P < 0.0001, RR = 2.91, RR = risk ratio). The survival advantage was seen in tumor size categories 1 to 10 mm, 11 to 20 mm and 21 to 50 mm (Figure ). Only two patients available for analysis in the screen-detected group had tumors larger than 50 mm, which did not allow a statistical comparison. Women with screen-detected T2 category tumors (21 to 50 mm) had prognoses similar to those of patients with non-screen-detected T1 category tumors (1 to 20 mm) (15-year BCSS 76.6% vs 77%; P = 0.863 and RR = 0.95) (Table ).
Breast cancer-specific survival by primary tumor size and mode of detection. Solid lines represent cancers detected by mammography screening, dashed lines represent cancers detected outside screening.
Breast cancer-specific survival according to primary tumor diameter
Influence of screening on outcome according to nodal status
This study included 1,203 (62%) node-negative and 654 (34%) node-positive breast cancer patients. Among the node-negative breast cancer patients, breast cancer was screen-detected in 317 patients (78%) and non-screen-detected in 862 patients (58%) (P < 0.001). In patients with screen-detected node-negative breast cancer, 15-year BCSS was 91% compared to 77% in the non-screen-detected group (P < 0.0001 and RR = 3). Patients with screen-detected, node-positive breast cancer had a survival advantage compared to patients diagnosed by other methods (71% vs 51%, respectively; P = 0.0011 and RR = 1.95).
Influence of age at diagnosis on outcome
In this study, there were 873 patients (45%) ages 50 to 69 years, of whom 347 (40%) were diagnosed by screening and 502 (58%) were diagnosed on the basis of other methods. There was a statistically significant difference in 15-year BCSS between the two subgroups (15-year BCSS 86% for patients with screen-detected tumors and 68% for patients diagnosed by other methods; P < 0.0001 and RR = 2.61). Similar differences in survival were observed in all other age cohorts (Table ).
Breast cancer-specific survival by method of detection and age at diagnosis
Multivariate survival analyses
Screening mammography is commonly offered to women 50 to 69 years of age. Because the proportion of screen-detected cancers was different according to age groups, and since screen-detected tumors were smaller, less often node-positive, less often of the ductal type, more differentiated, and associated with a more favorable hormonal status, we performed a multivariate analysis to adjust for these factors. As a result, the method of tumor detection was an independent prognostic factor, with a hazard ratio (HR) of 1.69 (95% CI = 1.06 to 2.70) between patients whose tumors were detected outside screening and those whose tumors were screen-detected. Tumor size, nodal status, tumor grade and HER2 amplification OK were also independent of other variables (Table ). The HR for the method of detection was 1.48 (95% CI = 0.97 to 2.25) in the first sensitivity analysis and 1.45 (95% CI = 0.97 to 2.17) in the second sensitivity analysis.
Contralateral breast cancer
Differences in the probability of having future contralateral breast cancer between screen-detected and non-screen-detected breast cancer patients during the 15-year follow-up time were assessed. For this analysis, we used the same exclusion criteria that we applied in the main analysis, with the exception of later contralateral breast cancers. Without any adjustments according to other variables, there were no significant differences if the primary breast cancer was diagnosed by screening mammography or by other methods (7.3% vs 6.8%, respectively; P = 0.74 and OR = 0.96). When patients 50 to 69 years of age were analyzed, the results were similar (7.0% vs 7.9%, respectively; P = 0.60, respectively). After adjustments according to the same covariates as those used in the Cox multivariate survival analyses, the risk of having later contralateral breast cancer was slightly increased in patients with breast cancer detected outside screening compared to patients with screen-detected breast cancer (95% CI = 0.55 to 2.57 and OR = 1.18), but the difference was not statistically significant.