This study estimated the excess incidence (overdiagnosis) of breast cancer associated with starting mammographic screening at an earlier age, by using data from the Dalarna County component of the Swedish Two-County Trial of breast cancer screening.
In Dalarna County, Sweden, 38,589 women aged 40 to 74 years were randomized to invitation to regular mammographic screening (active study population [ASP]) and 18,582 women to usual care (passive study population [PSP]). After 3 screening rounds (6-8 years after randomization), the PSP was invited to screening. The cumulative incidence of breast cancer was calculated in the ASP and PSP from randomization to 29 years later. In addition, cumulative incidence was calculated for invasive cancers, advanced invasive cancers (≥2 cm in maximum diameter or node-positive), and nonadvanced cancers (<2 cm and node negative).
There was no excess of cancers in the ASP at 29 year follow-up (relative risk, 1.00; 95% confidence interval, 0.92-1.08). Cumulative incidence in the 2 arms approximately equalized at the conclusion of the first round of screening of the PSP. There was an excess of nonadvanced cancers and a deficit of advanced cancers in the ASP, both of which persisted to 29 years.
There was no additional breast cancer incidence associated with 100,000 additional screens in the ASP. Results suggest that overdiagnosis is small and largely confined to the prevalence screen. Cancer 2012. © 2012 American Cancer Society.
In recent years, considerable attention has been given to the issue of overdiagnosis in breast cancer screening.1-4 In this context, overdiagnosis is defined as the diagnosis of breast cancer as a result of screening that would not have arisen in the lifetime of the host, and thus would have remained unknown had the screening not taken place. Overdiagnosis is often estimated by comparison of observed incidence in a population that is offered screening with that expected in the absence of screening. Probably due to the lack of consensus on how to estimate the latter, estimates of overdiagnosis vary widely.1, 2 Those estimates which more fully take into account complexities such as lead time and changes in underlying risk tend to report estimates of the order of 10% or less,1, 3 whereas those which do not adjust for these known influences on incidence report much larger rates.2, 4
The setting and context in which the estimate of overdiagnosis should be most straightforward is from long-term follow-up of randomized trials of screening; however, this often is not possible due to the control group being offered screening at the close of the trial5 or to the unavailability of long-term follow-up for breast cancer incidence.6 Estimates of overdiagnosis are available from the Canadian National Breast Screening Studies (NBSS-1 and NBSS-2), because these were randomized trials of screening that did not screen the control group at the close of the trial. The estimated excess incidence was available 8 years after the end of the trials, which was a total of 13 years follow-up. The investigators estimated 14% overdiagnosis in the NBSS-1, in which women were aged 40 to 49 years at recruitment, and 11% in the NBSS-2, in which women were aged 50 to 59 years at recruitment.6 It should be noted, however, that at 13 years follow-up, there remained considerable expected life years in which the control group incidence might catch up further, thus reducing the ratio of observed versus expected cases.
The Malmö Trial was another randomized trial in which the control group was not screened at closure. At 15 years after the end of the Malmö Trial (an average of 20 years total follow-up), a 10% excess incidence was observed.7 It should be noted, however, that there were likely members of the study group who continued to be screened after the upper age limit in the trial, which would increase observed versus expected incidence.8
Although it was not possible to estimate overdiagnosis for screening versus no screening, long-term follow-up of trials that offered screening to the control group at exit may provide estimates of the extent of overdiagnosis conferred by invitation to screening earlier in life. The Swedish Two-County Trial is a case in point.5 After an average of 3 screens offered to the study group, the control group was invited to screening. In August 2011, a workshop took place in Falun Central Hospital, Falun, Sweden, the headquarters of the Dalarna County component of the trial, to carry out downstream research on the Dalarna County data. Here, we report on 29-year breast cancer incidence data from this component of the Swedish Two-County Trial. In addition to overall incidence (in situ and invasive disease), we also report on incidence of invasive disease, and of advanced and nonadvanced invasive disease.
MATERIALS AND METHODS
The design of the Swedish Two-County Trial has been reported elsewhere.5, 9 Briefly, the female population aged 40 to 74 years in 2 Swedish counties, Dalarna (formerly known as Kopparberg) and Östergötland, were divided into 45 geographic clusters. These clusters were randomized into either the active study population (ASP), which was invited to one-view mammography screening, or into the control group, the passive study population (PSP), which received usual care. In Östergötland, the ASP:PSP randomization ratio was approximately 1:1, whereas in Dalarna it was approximately 2:1. After exclusion of women with previously diagnosed breast cancer, there were 77,080 women in the ASP and 55,985 in the PSP. The trial began in 1977 in Dalarna and 1978 in Östergötland. In Dalarna, there were 38,589 women in the ASP and 18,582 in the PSP.5 After an average of 3 screens in the ASP (6-8 years after randomization), the PSP was offered screening. As noted above, this article results from a workshop in Dalarna County on the long-term follow-up in that county only. Thus, we only had available data from the Dalarna component of the trial.
We had data on breast cancer incidence in the ASP and PSP in Dalarna to December 31, 2005, yielding a maximum of 29 years of follow-up. We calculated cumulative incidence of all breast cancers, invasive breast cancers, advanced invasive breast cancers (defined as either node-positive or ≥2 cm in maximum diameter or both), and nonadvanced invasive breast cancer (defined as node-negative and <2 cm in maximum diameter). Relative risks for the ASP compared to the PSP were calculated, assuming incidence to be Poisson distributed.10
Figure 1 shows the overall incidence of invasive and in situ breast cancer in the ASP and PSP by time since randomization. Table 1 shows the corresponding numbers of cases, person-years, and relative risks at 5-year intervals. There was an early excess of breast cancers in the ASP, which lasts to 8 years, when the PSP incidence rose following invitation of the PSP to screening. After 29 years, there was no excess incidence in the ASP (relative risk [RR], 1.00; 95% confidence interval [CI], 0.92-1.08).
|Active Study Population, Invited
|Passive Study Population, Control
|Relative Risk (95% Confidence Interval)
Figure 2 shows the cumulative incidence in the 4 age groups at randomization (40-49, 50-59, 60-69, and 70-74 years). There is no indication of excess incidence in the ASP except for a nonsignificant excess in the ASP in the oldest group, 70-74 years, at randomization (RR, 1.25; 95% CI, 0.97-1.61).
Figure 3 shows the incidence by invasive status. For invasive cases, the pattern was very similar to that for all cancers, with a slightly reduced early excess in the ASP and a 1% deficit at the end of follow-up (RR, 0.99; 95% CI, 0.92-1.07). The in situ cases showed a nonsignificant excess in the ASP (RR, 1.17; 95% CI, 0.88-1.55). In absolute terms, the final deficit in invasive disease was 22 cases, and the excess of in situ cases was very similar at 25 cases.
The incidence of advanced and nonadvanced invasive cancers shows very different patterns over time. From the start of the trial, a substantial excess of nonadvanced cancers was observed in the ASP (Fig. 4). This excess was slightly attenuated following the invitation of the PSP to screening, but a substantial difference remained up to the end of the 29-year follow-up period (RR, 1.13; 95% CI, 0.98-1.30). For advanced cancers, a deficit began to be observed in the ASP approximately 4 years after the start of the trial (Fig. 5). The cumulative incidence in the PSP became parallel to that of the ASP at approximately 8 years, and thereafter the deficit in the ASP remained to the end of the 29 years (RR, 0.81; 95% CI, 0.72-0.91).
After a maximum 29 years of follow-up within a randomized trial of mammographic screening, we found no excess of breast cancers in the group invited to screening. Because the noninvited group was offered screening at 6 to 8 years into the trial, this does not allow us to specifically estimate overdiagnosis from screening. It does, however, indicate 2 major points. First, there was no overdiagnosis associated with the additional 3 screens of the ASP in the first 8 years of observation. Thus, the earlier initiation of offering screening to the 38,589 women in the ASP, amounting to approximately 100,000 screening episodes, did not add any overdiagnosed cases to those from the screening in both groups thereafter. The “catch-up” of the PSP immediately after screening started in this group indicates that any overdiagnosis is largely incurred at the prevalence screen. This finding, combined with the lack of any overall excess incidence from the 100,000 additional screens in the ASP is consistent with the modest estimates of overdiagnosis from other randomized trials and from those studies that took full account of the complexities of lead time and changes in underlying risk.1, 3, 6, 7 Because an additional 100,000 screens added no excess incidence, it is unlikely that overdiagnosis rates associated with breast cancer screening of the order of 25% to 36% suggested by some can possibly apply.2, 11
The slight proportional deficit in invasive cancers in the ASP at 29 years may be due to detection and removal of carcinoma in situ in the ASP in the first 8 years. As we have previously noted, the excess of in situ disease in the ASP was largely balanced by a deficit in invasive cancer.12 However, the observed deficit is not significant, represents approximately 22 of 1903 tumors, and may be due to chance alone.
The absence of any excess incidence in the ASP was observed in all age groups except the oldest, ie, women aged 70 to 74 years at randomization. Although this was not significant, it is at least plausible, because in this age group, high competing risks might cause death in some screen-detected cases before the disease would have given rise to symptoms. Also, there was considerably less screening activity in the PSP after the trials closed in this age group than at younger ages.
Examination of Figures 4 and 5 indicates that the excess of cases apparent immediately after screening initiation is made up of nonadvanced cases, many of which would have occurred later as advanced cases. The long-term reduction in advanced cancers is consistent with the long-term reduction in breast cancer mortality observed in the trial.5
In conclusion, at 29 years follow-up, there was no excess incidence of breast cancer in the ASP in the Dalarna County component of the Swedish Two-County Trial, whether or not in situ disease is included. This, together with the approximate equalization of cumulative incidence at the first offer of screening to the PSP suggests that overdiagnosis is a minor phenomenon and is largely confined to the prevalence screen.
The Swedish Two-County Trial was initiated by the Swedish National Board of Health and Welfare and funded by the County Council of Dalarna. This analysis was supported by the American Cancer Society through a gift from the Longaberger Company's Horizon of Hope Campaign®. We owe a debt of gratitude to the women who took part in the trial.
CONFLICT OF INTEREST DISCLOSURE
The authors made no disclosure.