Mammographic screening reduces risk of breast carcinoma recurrence

Abstract

BACKGROUND

The current report is a long-term evaluation of breast carcinoma recurrence, factors predicting recurrence, and postrecurrence prognosis in relation to patients' use of service screening, which has been provided in Turku, Finland, since 1987 for women ages 40–74 years.

METHODS

The current study included 527 invasive breast carcinomas: 418 in the screening group (which included screen-detected and interval malignancies) and 109 in the nonscreening group (which included breast carcinomas detected before initial screening and those detected in patients who chose not to undergo screening). These breast carcinomas were diagnosed among women ages 40–74 years between 1987 and 1993, with follow-up extending until the end of 2001.

RESULTS

In the screening group, the risk of recurrence was only approximately half of the corresponding risk in the nonscreening group (hazard ratio [HR], 0.57; 95% confidence interval [CI], 0.39–0.83; P = 0.003). Five years after the primary diagnosis, 16% of patients in the screening group and 28% of patients in the nonscreening group (P = 0.001) had experienced recurrence; 10 years after diagnosis, the corresponding rates were 21% and 34%, respectively (P = 0.001). Postrecurrence prognosis was comparable for both detection groups (HR, 1.17; 95% CI, 0.70–1.94; P = 0.551), with approximately half of all patients dying of disease 5 years after recurrence. Detection of breast carcinoma via a method other than mammographic screening was associated with a high risk of recurrence on univariate analysis. On Cox multivariate analysis, risk factors for recurrence included lobular histologic type (HR, 2.23; 95% CI, 1.44–3.48; P < 0.001), poor histologic grade (HR, 2.02; 95% CI, 1.20–3.39; P = 0.008), and large tumor size (HR, 1.60; 95% CI, 1.07–2.37; P = 0.021).

CONCLUSIONS

Long-term data from a population-based program demonstrated that mammographic screening reduced patients' risk of breast carcinoma recurrence. Specifically, the risk for patients with screen-detected disease was only approximately half of the risk for patients with non-screen-detected disease. Nonetheless, postrecurrence prognosis was comparable for patients in both detection groups. Cancer 2005. © 2004 American Cancer Society.

As a consequence of the excellent results regarding mortality reduction in early randomized studies, population-based breast carcinoma screening programs have been launched in more than 20 countries since the late 1980s/early 1990s.1 In Finland, a national population-based mammographic screening program for women ages 50–59 years was started in 1987.2 The city of Turku, however, offered mammographic screening to all women ages 40–74 years. In our report on the results from the first 11 years of the Turku screening program, we demonstrated that primary invasive breast carcinomas detected during screening or during the interval between screenings were more often small, localized, and better differentiated than were malignancies detected without screening. In addition, survival rates were higher in the screening group than in the nonscreening group for women of all ages (40–49 years: hazard ratio [HR], 2.47; 95% confidence interval [CI], 1.21–5.05; P = 0.011; 50–59 years: HR, 2.00; 95% CI, 0.93–4.29; P = 0.069; 60–69 years: HR, 2.14; 95% CI, 1.05–4.35; P = 0.032; 70–74 years: HR, 3.94; 95% CI, 1.48–10.48; P = 0.003). In that study, increased rates of axillary lymph node negativity (P < 0.0001), histologic Grade I disease (P = 0.0005), and small tumor size (P = 0.0118) accounted for the beneficial effects of screening on survival.3 These three factors are significant predictors of both overall and recurrence-free survival for patients with breast carcinoma.3–6 In another report, we showed the beneficial effect of screening on breast carcinoma treatment costs. Over a 5-year follow-up period, the mean treatment cost per patient diagnosed with breast carcinoma between 1987 and 1993 was 1.3 times greater for unscreened women compared with screened women. The estimated rate of savings resulting from early treatment was 26–30%, measured as a proportion of screening costs during this period.7

It also has been shown that screening reduces breast carcinoma–related mortality.3, 8–11 Despite the overall favorable prognosis for women with this malignancy, unpredictable breast carcinoma recurrences and mortalities are observed even among those who have T1N0M0 disease; we found that high Ki-67 immunopositivity was the strongest predictor of recurrence for this subset of patients.12 Very little is known about the value of breast carcinoma screening in the prevention of recurrent disease.13, 14 In the current study, we report the results of a long-term evaluation of breast carcinoma recurrence, factors predicting recurrence, and postrecurrence prognosis in relation to service screening use.

MATERIALS AND METHODS

The current study was an observational longitudinal analysis assessing the value of mammographic screening in the reduction of breast carcinoma recurrence and postrecurrence mortality.

In Finland, public healthcare services offer screening mammography and treatment for breast diseases. Nationwide population-based breast carcinoma screening was introduced in Finland in 1987, and this national program covers women ages 50–59 years. In the city of Turku in southwest Finland, population-based mammographic screening has been offered free of charge since 1987 for all women ages 40–74 years, as has been described in detail elsewhere.15, 16 This screening, which involved two viewers and independent double-reading of results, was performed at the Cancer Society of Southwest Finland Breast Examination Center in cooperation with the city of Turku. No other mass mammographic screening took place concomitantly with this public screening program. Mammography was available at public health clinics for all women with symptoms of breast disease, whereas private clinics offered mammography only for patients referred by a physician.

During the study period, the population in Turku was 160,000, approximately 36,000 of whom were women ages 40–74 years (40–49 years, n = 12,600; 50–59 years, n = 9250; 60–69 years, n = 9950; 70–74 years, n = 4200). Women ages 40–49 years who were born in even years were invited to screening yearly, and those born in uneven years were invited every third year. Women ages 50–74 years were invited for screening every second year. Between 1987 and 1993, the screening attendance rate was 88%. For women ages 40–49 years, 50–59 years, and 60–69 years, attendance rates varied between 85.9% and 90.3% but remained stable throughout the study period. Women ages 70–74 years were first invited for screening in 1988, and their participation rates were slightly lower (mean, 81%) but improved toward the end of the study period.

Between 1987 and 1993, there were 562 women diagnosed with primary invasive breast carcinoma and 41 diagnosed with carcinoma in situ at age 40–74 years. Malignancies that were present at baseline (i.e., those detected before the screening period) were not included in the study, nor were carcinomas in situ. Primary treatment data were not available for 6 invasive breast carcinoma cases, and these cases were subsequently excluded from the analysis, leaving 427 cases in the screening group and 129 cases in the nonscreening group. In the former group, 9 of 427 primary malignancies (2%) had distant metastases at detection (2%), compared with 20 of 129 primary malignancies in the latter group (16%) (P < 0.001). These malignancies also were excluded from our analysis of recurrence, leaving a total of 527 breast carcinoma cases in the current study: 418 (79%) in the screening group and 109 (21%) in the nonscreening group. All of these 527 cases were followed through the end of 2001.

Patients were classified into two groups according to their screening status: screened women had breast carcinoma detected via screening or detected during the interval between two screening mammograms, whereas unscreened women were invited for screening but chose not to participate or else had breast carcinoma detected before the first screening mammogram. Malignancies that were not detected by screening were diagnosed either at public or private clinics.

Malignancies were classified postoperatively according to the International Union Against Cancer TNM classification and reclassified histologically as described elsewhere.16, 17 Dates and causes of death were obtained from Statistics Finland and were reevaluated and confirmed by two specialists (an oncologist and a forensic pathologist) who were blinded to screening status. Data on sociodemographic status at the time of diagnosis was obtained from Statistics Finland for 80% of the study cohort. Permission to access patient records was granted by the Finnish Ministry of Social Affairs and Health and by Statistics Finland.

Turku has two hospitals that, with very few exceptions, treat all patients with breast carcinoma. Both of these hospitals have inpatient wards and outpatient clinics. Surgery is performed either at Turku City Hospital or at Turku University Central Hospital, with the latter providing oncologic treatment. In Finland, treatment is standardized according to the recommendations of the Finnish Breast Cancer Group.18 For the current study, data on primary treatment after diagnosis and data on recurrence between the time of primary treatment and the end of 2001 were obtained from individual patient records and confirmed via clinical and/or pathologic assessment. Data on treatment and recurrence were comprehensive, covering 99% of all eligible breast carcinoma cases. The proportion of patients with distant metastases at detection was 5.2%, which was 0.5% higher than the rate calculated using official statistics from Cancer Finland. Locoregional recurrence was defined as the reappearance of breast carcinoma in the treated breast or in the regional lymph nodes after primary treatment.

Recurrence-free rates were calculated from the date of the primary diagnosis to the date of first recurrence (locoregional or distant) or, for patients without recurrent disease, to the date of most recent follow-up. Postrecurrence survival rates were calculated from the date of first recurrence to the date of disease-related death or, for patients who were still alive, to the date of most recent follow-up. Time to locoregional progression was defined as the time from primary diagnosis to locoregional recurrence, and the time to distant progression was defined as the time from primary diagnosis to the appearance of metastases (either as a first recurrence or after locoregional recurrence). The median follow-up was 10.0 years for screened women and 12.4 years for unscreened women.

Differences between the two detection groups in terms of patient characteristics, clinical and histopathologic variables, and treatment were analyzed via cross-tabulation, and significance was assessed using the Fisher exact test. Differences in recurrence-free rates, recurrence rates, and survival rates were evaluated using survival analysis methods. Survival curves were generated using the Kaplan–Meier method, and statistical testing was performed using the log-rank test or (in Cox proportional hazards analysis) the Wald test. Multivariate analyses and HR calculations with 95% CIs were performed using the Cox proportional hazards model. All computations were executed using SAS software (Release 8.2.2001; SAS Institute, Cary, NC).

RESULTS

Data regarding patient age, sociodemographic characteristics, treatment, and clinicopathologic characteristics are shown in Table 1. The two detection groups did not differ significantly in terms of age or sociodemographic characteristics. In the screening group, the histologic type of the primary tumor was invasive ductal carcinoma (IDC) in 296 cases (71%), invasive lobular carcinoma (ILC) in 70 cases (17%), and other in 52 cases (12%); in the nonscreening group, there were 93 IDCs (85%), 7 ILCs (6%), and 9 other breast carcinomas (8%; Table 1) (P = 0.005). In the former group, 74% of all IDCs and 74% of all ILCs measured < 20 mm, compared with 49% of all IDCs and 71% of all ILCs in the latter group. Seventy-six percent of all IDCs and 84% of all ILCs in the screening group had negative lymph node status, whereas 60% of all IDCs and 71% of all ILCs in the nonscreening group exhibited lymph node negativity.

Table 1. Summary of Patient Age and Sociodemographic Characteristics at the Time of Primary Diagnosis and of Clinical and Histopathologic Variables and Primary Tumor Treatment by Detection Group

Variable	Screened (%)	Unscreened (%)	P value
a Data from Statistics Finland for 80% of cohort. b Not necessarily performed/administered in all cases.
Age (n = 527)			0.167
40–49 yrs	91 (21.8)	32 (29.4)
50–59 yrs	128 (30.6)	23 (21.1)
60–69 yrs	145 (34.7)	40 (36.7)
70–74 yrs	54 (12.9)	14 (12.8)
Marital status (n = 426)a			0.104
Unmarried	36 (10.6)	16 (18.8)
Married	195 (57.2)	38 (44.7)
Widowed	51 (14.9)	14 (16.5)
Divorced	59 (17.3)	17 (20.0)
Level of education (n = 426)a			0.323
Basic	205 (60.1)	57 (67.1)
Intermediate	87 (25.5)	15 (17.6)
High	49 (14.4)	13 (15.3)
Socioeconomic status (n = 414)a			0.071
Manual laborer	43 (12.7)	5 (6.8)
Employee	98 (28.8)	14 (18.9)
Pensioner	176 (51.8)	47 (63.5)
Other	23 (6.8)	8 (10.8)
Income level (for those in labor force; n = 187)a			0.247
Low (< € 16,819/yr)	91 (59.5)	23 (67.7)
Moderate or high	62 (40.5)	11 (32.3)
Axillary lymph node metastases (n = 527)			< 0.001
N0 (negative)	328 (78.5)	68 (62.4)
N1–3 (positive)	90 (21.5)	41 (37.6)
Tumor size (n = 516)			< 0.001
1–20 mm	306 (75.0)	57 (52.8)
> 20 mm	102 (25.0)	51 (47.2)
Histologic type (n = 527)			0.005
Invasive ductal carcinoma	296 (70.8)	93 (85.3)
Invasive lobular carcinoma	70 (16.8)	7 (6.4)
Other	52 (12.4)	9 (8.3)
Histologic grade (n = 513)			< 0.001
I (well-differentiated)	139 (34.3)	14 (13.0)
II–III (moderately well or poorly differentiated)	266 (65.7)	94 (87.0)
Treatment (n = 527)
Surgery	236 (56.5)	43 (39.5)
Modified radical mastectomy, axillary evacuation, radiotherapy	44 (10.5)	32 (29.4)	< 0.001
Breast conserving surgery, axillary evacuationb, radiotherapy	51 (12.2)	9 (8.3)
Surgery, radiotherapyb, hormone therapy	51 (12.2)	20 (18.4)
Surgery, radiotherapy, chemotherapy	34 (8.1)	4 (3.7)
Other	2 (0.5)	1 (0.9)

In the screening group, 57% of all patients underwent surgery alone, and the remaining 43% required additional treatment; in contrast, 40% of all patients in the screening group received surgery alone, whereas 60% required additional treatment (Table 1). The rate of breast-conserving surgery was 15% among screened patients and 9% among unscreened patients (P = 0.148).

During follow-up, 93 screened patients (22%) and 38 unscreened patients (35%) developed recurrent disease. Survival analysis revealed a significantly higher recurrence-free survival rate after the diagnosis of the primary malignancy for women in the screening group. The risk of first recurrence (locoregional or distant) in the screening group was only 57% as large as the corresponding risk in the nonscreening group (HR, 0.57; 95% CI, 0.39–0.83; P = 0.003). The difference in risk between the two groups was already statistically significant after 2 years of follow-up. At that point, 4% (95% CI, 2–5%) of all screened women and 15% (95% CI, 8–22%) of all unscreened women had recurrent disease (Fig. 1). After 5 years of follow-up, 16% of screened women and 28% of unscreened women had experienced disease recurrence (P = 0.001), and after 10 years, the corresponding rates were 21% and 34%, respectively (P = 0.001) (Table 2).

Figure 1. Kaplan–Meier recurrence-free survival rates after primary breast carcinoma treatment for screened (solid line) and unscreened (dashed line) women. P = 0.003; hazard ratio (screened vs. unscreened) = 0.57 (95% confidence interval, 0.39–0.83).

Table 2. Recurrence Rates for 527 Invasive Breast Carcinomas Detected between 1987 and 1993, by Age and Detection Group

Type of recurrence	% of patients
	5 yr follow-up			10 yr follow-up			Entire follow-up
	Screened	Unscreened	Log-rank P value	Screened	Unscreened	Log-rank P value	Screened	Unscreened	Log-rank P value
a Only as the first type of treatment failure. b Only when occurring as the first type of treatment failure or after a locoregional recurrence. c Only as the first type of treatment failure.
Locoregionala	4.8	11.9	0.003	6.7	13.8	0.006	7.4	13.8	0.018
40–49 yrs	7.7	12.5	0.309	8.8	15.6	0.191	11.0	15.6	0.325
50–59 yrs	3.1	4.4	0.730	7.8	4.4	0.610	8.6	4.4	0.468
60–69 yrs	3.5	15.0	0.005	3.4	17.5	0.001	3.5	17.5	0.001
70–74 yrs	7.4	14.3	0.297	9.3	14.3	0.418	9.3	14.3	0.429
Distantb	12.0	17.4	0.078	15.3	23.9	0.020	16.3	25.7	0.024
40–49 yrs	16.5	28.1	0.094	17.6	31.3	0.060	18.7	34.4	0.047
50–59 yrs	13.3	17.4	0.555	18.8	21.7	0.707	20.3	21.7	0.939
60–69 yrs	9.0	12.5	0.495	13.1	22.5	0.163	13.8	22.5	0.231
70–74 yrs	9.3	7.1	0.965	9.3	14.3	0.432	9.3	21.4	0.250
Locoregional or distantc	15.8	28.4	0.001	20.8	33.9	0.001	22.3	34.9	0.003
40–49 yrs	22.0	40.6	0.015	24.2	43.8	0.012	26.4	43.8	0.027
50–59 yrs	16.4	21.7	0.467	25.8	26.1	0.893	28.1	26.1	0.814
60–69 yrs	11.0	25.0	0.021	15.2	32.5	0.010	15.9	32.5	0.018
70–74 yrs	16.7	21.4	0.457	18.5	28.6	0.243	18.5	35.7	0.142

The majority of recurrences (74%; 71% in the screening group and 82% in the nonscreening group) were observed during the first 5 years after primary tumor detection. The median interval from diagnosis to recurrence was 3.2 years for screened women and 2.1 years for unscreened women.

At detection, there were 31 locoregional recurrences (33%) and 62 distant recurrences (67%) in the screening group, compared with 15 locoregional recurrences (39%) and 23 distant recurrences (61%) in the nonscreening group. In the former group, 6 patients with locoregional recurrence after primary diagnosis developed distant metastases during follow-up, compared with 5 patients in the latter group; thus, by the end of follow-up, there were 68 distant recurrences (73%) documented in the screening group and 28 distant recurrences (74%) documented in the nonscreening group. According to our survival analyses, locoregional recurrence rates were significantly different between the two detection groups after both 5 and 10 years of follow-up, and the distant recurrence rate was significantly different between these two groups after 10 years of follow-up (Table 2).

In our age group analysis, the rate of first recurrence (locoregional or distant) was significantly lower for screened women compared with unscreened women in the 40–49-year-old and 60–69-year-old age groups after 5 years of follow-up, and these differences remained significant after 10 years of follow-up and also for the follow-up period as a whole (Table 2).

The recurrence rate for IDC cases was 21% (63 of 296) in the screening group and 35% (33 of 93) in the nonscreening group. The corresponding rates were 39% (27 of 70) and 43% (3 of 7), respectively, for ILC cases and 6% (3 of 52) and 22% (2 of 9), respectively, for other histologic types. On survival analysis, the difference between detection groups was statistically significant for patients with IDC (P = 0.002), but not for patients with ILC (P = 0.892) or patients with other histologic breast carcinoma types (P = 0.088).

Factors that were predictive of first recurrence (locoregional or distant) were investigated using the Cox proportional hazards model. Univariate analysis showed that detection of the primary malignancy via a method other than screening (P = 0.003), receipt of multimodality treatment (P < 0.001), positive axillary lymph node status (P < 0.001), ILC (P < 0.001), size > 20 mm (P = 0.002), and poor histologic differentiation of the primary tumor (P = 0.004) were associated with a high risk of recurrence. On Cox multivariate analysis, the hazard ratio for first recurrence was significantly higher for patients who had ILC compared with patients who had IDC (HR, 2.23; 95% CI, 1.44–3.48; P < 0.001), and histologic type was the most significant predictor of first recurrence. Other predictors of recurrence included poor histologic grade (HR, 2.02; 95% CI, 1.20–3.39; P = 0.008) and large tumor size (HR, 1.60; 95% CI, 1.07–2.37; P = 0.021) (Table 3). Positive axillary lymph node status was a near-significant predictor of first recurrence, and it also appeared to be a significant predictor of distant recurrence (HR, 2.49; 95% CI, 1.33–4.68; P = 0.004), as did lobular histologic type (HR, 2.64; 95% CI, 1.58–4.41; P < 0.001) and poor histologic grade (HR, 2.60; 95% CI, 1.32–5.11; P = 0.006).

Table 3. Multivariate Analysis of Predictors of Recurrence in 503 Patients with Invasive Breast Carcinomaa

Variable	P value	HR	95% CI
HR: hazard ratio; CI: confidence interval. a All variables in the table were included in a Cox proportional hazards model. b Not necessarily performed/administered in all cases.
Age (yrs)	0.133
40–49		1.00
50–59		0.98	0.60–1.57
60–69		0.61	0.37–1.01
70–74		1.02	0.53–1.93
Detection group	0.129
Unscreened		1.00
Screened		0.72	0.47–1.10
Treatment	0.342
Surgery		1.00
Modified radical mastectomy, axillary evacuation, radiotherapy		1.02	0.58–1.80
Breast-conserving surgery, axillary evacuationb, radiotherapy		0.44	0.18–1.12
Surgery, radiotherapyb, hormone therapy		1.38	0.72–2.62
Surgery, radiotherapy, chemotherapy		1.04	0.46–2.37
Axillary lymph node metastases	0.054
N0 (negative)		1.00
N1–3 (positive)		1.70	0.99–2.93
Histologic type	< 0.001
Invasive ductal carcinoma		1.00
Invasive lobular carcinoma		2.23	1.44–3.48
Other		0.43	0.16–1.19
Tumor size (mm)	0.021
1–20		1.00
> 20		1.60	1.07–2.37
Histologic grade	0.008
I (well-differentiated)		1.00
II–III (moderately well or poorly differentiated)		2.02	1.20–3.39

During follow-up, 74 of 527 patients (14%; 52 of 418 [12%] in the screening group and 22 of 109 [20%] in the nonscreening group) died of breast carcinoma, and another 70 (13%; 49 of 418 [12%] in the screening group and 21 of 109 [19%] in the nonscreening group) died of other causes. Among women who died of other causes, 31% in the screening group died of other malignancies, and 45% died of cardiovascular conditions; the corresponding rates in the nonscreening group were 24% and 52%, respectively, and the difference between the two detection groups was not significant. Women who died of causes other than breast carcinoma did not differ in terms of age or socioeconomic status. All breast carcinoma–related deaths were preceded by a distant recurrence, which either represented the first recurrence or followed a locoregional recurrence. Survival analysis revealed no statistically significant difference between the two detection groups (HR, 1.17; 95% CI, 0.70–1.94; P = 0.551). Approximately half of all patients died of breast carcinoma within 5 years of the detection of recurrent disease (Fig. 2).

Figure 2. Kaplan–Meier overall survival rates after first recurrence for screened (solid line) and unscreened (dashed line) women. (Deaths due to other causes have been censored.) P = 0.551; hazard ratio (screened vs. unscreened) = 1.17 (95% confidence interval, 0.70–1.94).

DISCUSSION

The current study demonstrated that long-term breast carcinoma recurrence rates were significantly lower among screened patients compared with unscreened patients after 5 years and 10 years of follow-up, with this trend holding true for the follow-up period as a whole. Olivotto et al.13 from British Columbia, Canada, and Magee et al.14 from the United Kingdom have also reported significant differences in 5-year recurrence rates between screened and unscreened women. The results of the current study are consistent with those findings, although the observed recurrence rates cannot be directly compared. The study conducted by Olivotto et al.13 included patients with distant metastases at detection, whereas such patients were excluded from the current study. In the current study, tumors that were metastatic at diagnosis accounted for 16% of all malignancies encountered in the nonscreening group and only 2% of all malignancies encountered in the screening group. If these malignancies had been included, the observed difference in recurrence-free survival rates would have been even larger. In contrast to the current study, the study conducted by Magee et al.14 was confined to patients treated with breast-conserving surgery and radiotherapy.

Several studies have evaluated prognostic factors for local or locoregional breast carcinoma recurrence after mastectomy or breast-conserving surgery and radiotherapy. Factors such as tumor size, lymph node involvement, and histologic grade have been identified as significant predictors of both recurrence-free and overall survival.3–6 Adding to these findings, the current study demonstrates that detection of breast carcinoma via a method other than screening is an independent predictor of recurrence. On multivariate analysis, statistically significant factors underlying this risk included lobular histologic type, poor histologic grade, and large tumor size.

The prevalence of ILC cases was higher among screened women compared with unscreened women. In general, however, the absolute prevalence of ILC was low, a finding that is consistent with previous findings made in this cohort.3 There is evidence that screening mammography may fail to reveal tumors with lobular or mucinous histology, as well as some rapidly proliferating, high-grade tumors.19 Therefore, it is surprising that ILC was more common among screened women in the current study. In general, the proportion of ILC cases is increasing, a finding that has also been reported in other studies.3, 20 In a study involving postmenopausal women, long-term use of continuous combined hormone replacement therapy (HRT) was associated with a higher risk of ILC compared with sequential combined HRT.21 Overall, tumors found in HRT users have been shown to be smaller, to have better histologic differentiation, and to exhibit lower proliferation rates (as measured by S-phase fraction) compared with tumors found in nonusers.22 Furthermore, survival rates after breast carcinoma detection have been shown to be higher in postmenopausal users of HRT than in nonusers.23, 24 To some extent, these findings may reflect a surveillance bias, because HRT users receive more active mammographic screening; however, the observation of lower S-phase fractions, among other findings, suggests that tumors found in HRT users may intrinsically tend to be more benign. Nonetheless, the overall effect of HRT on breast carcinoma mortality in the general population remains uncertain.25

The use of HRT in Finland increased steadily until 2003. In 2002, approximately 22% of all women age > 45 years were users, with a peak in usage rates at ages 55–56 years.26 Comparisons between screened and unscreened women may be susceptible to a selection bias if women receiving HRT are more likely to receive mammographic screening.22, 25 The two detection groups examined in the current study did not differ in terms of age or sociodemographic status. During the study period (1987–1993), screening was offered free of charge, and attendance rates were high; thus, the presence of a selection bias with regard to economic factors is relatively unlikely.15 The two detection groups also did not differ in terms of breast carcinoma history, because only primary malignancies were included in the study. Mortality due to causes other than breast carcinoma was higher among unscreened women, and according to our previous study, this difference was already present during the first 5-year follow-up period7; however, overall, the number of deaths due to other causes was too small to allow any statistically significant difference to be found. Finally, Aro et al.27 found that women who did not attend screening were less likely to comply with health recommendations or receive health services; more likely to be socially isolated, depressed, or anxious; and more likely to smoke.27

Mammographic screening detects breast carcinomas in a preclinical stage, thereby providing a ‘lead time’ and, consequently, a survival advantage to women with screen-detected disease.10 It is possible that recurrence-free survival data were biased by this lead time effect. Although differences in follow-up duration between the two detection groups did not affect our analyses covering 5-year and 10-year follow-up periods, these differences may have influenced the results of our analysis covering patients' entire follow-up periods. In both groups, however, most recurrences were documented during the first 5 years after primary treatment. After 15 years of follow-up, unscreened patients continued to have lower recurrence-free survival rates compared with screened patients. Thus, screened patients can also be assumed to have had superior survival over this length of time. Furthermore, the concept that mammographic screening is more likely to detect less aggressive, slow-growing tumors that do not have life-threatening potential is not plausible, according to our previous study.3

The current study, which is based on long-term data from a population-based program, demonstrates that mammographic screening is associated with a reduced risk of breast carcinoma recurrence. Specifically, the risk of recurrence for patients with screen-detected disease was only approximately half of the corresponding risk for patients with non-screen-detected disease. These results are in accordance with the findings of our previous study, which showed that women with screen-detected breast carcinoma had higher survival rates compared with women who had non-screen-detected breast carcinoma. In general, our work suggests that it is possible to translate the benefits seen in randomized breast screening trials to the general population, as has been shown in Scandinavia and North America.

Acknowledgements

The authors thank Dr. Antti Jääskeläinen and Kari Ojala (Chief Executive Officer), Cancer Society of Southwest Finland, Turku, Finland.