Differences in endpoints between the Swedish W-E (two county) trial of mammographic screening and the Swedish overview: methodological consequences.

OBJECTIVES: To characterize and quantify the differences in the number of cases and breast cancer deaths in the Swedish W-E Trial compared with the Swedish Overview Committee (OVC) summaries and to study methodological issues related to trials in secondary prevention. SETTING: The study population of the W-E Trial of mammography screening was included in the first (W and E county) and the second (E-county) OVC summary of all Swedish randomized mammography screening trials. The OVC and the W-E Trial used different criteria for case definition and causes of death determination. METHOD: A Review Committee compared the original data files from W and E county and the first and second OVC. The reason for a discrepancy was determined individually for all non-concordant cases or breast cancer deaths. RESULTS: Of the 2615 cases included by the W-E Trial or the OVC, there were 478 (18%) disagreements. Of the disagreements 82% were due to inclusion/exclusion criteria, and 18% to disagreement with respect to cause of death or vital status at ascertainment. For E-County, the OVC inclusion rules and register based determination of cause of death (second OVC) rather than individual case review (W-E Trial and 1st OVC) resulted in a reduction of the estimate of the effect of screening, but for W-County the difference between the original trial and the OVC was modest. CONCLUSIONS: The conclusion that invitation to mammography screening reduces breast cancer mortality remains robust. Disagreements were mainly due to study design issues, while disagreements about cause of death were a minority. When secondary research does not adhere to the protocols of the primary research projects, the consequences of such design differences should be investigated and reported. Register linkage of trials can add follow-up information. The precision of trials with modest size is enhanced by individual monitoring of case status and outcome status such as determination of cause of death.

Number needed to screen: lives saved over 20 years of follow-up in mammographic screening.

OBJECTIVE: To estimate the number needed to screen with mammography to save one life, based on a stated amount of screening activity and long-term follow-up for breast cancer death. SETTING: A randomised controlled trial of mammographic screening for breast cancer, with 77,080 women invited to screening and 55,985 not invited. The invited group was offered screening for seven years. Follow-up continued for a total of just over 20 years. METHODS: Number needed to screen for seven years to save one life over 20 years was calculated by dividing the number screened (not the number invited) by the total number of lives saved. Similarly, we calculated the number of mammographic examinations required to save one life. RESULTS: We estimate that the number of women needed to screen for seven years to save one life over 20 years is 465 (95% CI 324-819). The number of mammographic examinations needed to save one life was 1499 (95% CI 1046-2642). CONCLUSIONS: The number needed to screen to save one life is smaller than has been reported in the past. Mammographic screening is effective in absolute terms as well as relative. Long-term follow-up allowed us to estimate the absolute benefit with greater accuracy.

The Swedish Two-County Trial of mammographic screening: cluster randomisation and end point evaluation.

BACKGROUND: The Swedish Two-County Trial has been criticised on the grounds of the cluster randomisation and alleged bias in classification of cause of death. PATIENTS AND METHODS: In the Two-County Trial, 77 080 women were randomised to regular invitation to screening (active study population, ASP) and 55 985 to no invitation (passive study population, PSP), in 45 geographical clusters. After approximately 7 years, the PSP was invited to screening and the trial closed. We analysed data using hierarchical statistical models to take account of cluster randomisation, and performed a conservative analysis assuming a systematic difference between ASP and PSP in baseline breast cancer mortality in one of the counties. We also analysed deaths from causes other than breast cancer and from all causes among breast cancer cases diagnosed in the ASP and PSP. RESULTS: Taking account of the cluster randomisation there was a significant 30% reduction in breast cancer mortality in the ASP. Conservatively, assuming a systematic difference between ASP and PSP clusters in baseline breast cancer mortality, there was a significant 27% reduction in mortality in the ASP. Ignoring classification of cause of death, there was a significant 13% reduction in all-cause mortality in breast cancer cases in the ASP. CONCLUSIONS: Breast cancer mortality is a valid end point and mammographic screening does indeed reduce mortality from breast cancer. The criticisms of the Swedish Two-County Trial are unfounded.

Quantifying the potential problem of overdiagnosis of ductal carcinoma in situ in breast cancer screening.

The relevance of detection of ductal carcinoma in situ (DCIS) in a breast cancer screening programme, and the extent of overdiagnosis of non-progressive lesions, remains controversial. It was the purpose of this paper to estimate the incidence of non-progressive, 'overdiagnosed' DCIS. We defined non-progressive DCIS (DCIS(0)) as DCIS which could not have progressed to invasive disease if left untreated. Progressive DCIS (DCIS(1)) was defined as DCIS which has the propensity to progress to invasive disease. We fitted a Markov process model of the incidence of progressive and non-progressive DCIS, the transition of the former to preclinical invasive disease and the subsequent progression to clinical symptomatic cancer. We used data from the Swedish Two-County Trial and from service screening programmes in the UK, Netherlands, Australia and the USA to estimate the incidence of progressive and non-progressive DCIS, and the detection rates of each at the first and subsequent screening. Average incidence of non-progressive DCIS was 1.11 per 100000 per year. Average incidence of progressive DCIS was 2.1 per 1000 per year. At prevalence screen, 37% of DCIS cases were estimated to be non-progressive. A woman attending prevalence screen has a 19 times greater chance of having a progressive DCIS or an invasive tumour diagnosed than of having a non-progressive DCIS diagnosed. At incidence screen, only 4% of DCIS cases were estimated to be non-progressive. A woman attending an incidence screen has a 166 times higher probability of having a progressive DCIS or invasive lesion diagnosed than of having a non-progressive DCIS diagnosed. There is an element of overdiagnosis of DCIS in breast cancer screening, but the phenomenon is small in both relative and absolute terms.

The relative contributions of screen-detected in situ and invasive breast carcinomas in reducing mortality from the disease.

We aimed to quantify the benefits of detecting ductal carcinoma in situ (DCIS) and of downwards stage-shifting within invasive tumours in mammographic screening. Using data from the Swedish Two-County Trial of breast cancer screening, we examined the 20-year death rates from invasive tumours of stage II or worse, invasive tumours of stage I and DCIS. We then used these rates and their respective incidences in invited (active study population, ASP) and control (passive study population, PSP) arms of the trial, to estimate the numbers of deaths avoided by downward stage-shifting the larger stage II or worse tumours to stage I and the stage I cancers to DCIS. We also studied the association between the mortality reduction achieved and the proportion of DCIS cases detected in the randomised trials of breast cancer screening. In the Swedish Two County Trial, 141 breast cancer deaths were avoided in the ASP compared with the PSP at approximately 20 years of follow-up. Of these, 65% (91/141) were avoided as a result of stage-shifting from invasive stage II or worse to invasive stage I, and 5% (7/141) as a result of stage-shifting from invasive stage I to DCIS. If we assumed that 10% of stage II or worse tumours avoided were shifted not to stage I, but to DCIS, the estimated number of deaths prevented by shifting from invasive disease to in situ was 17, which is 12% of all deaths prevented. When the results of all the randomised trials of breast cancer screening were reviewed, there was no clear association between the percentage of DCIS cases diagnosed and the observed mortality reduction. We conclude that compared with downward stage-shifting of invasive tumours, detection of DCIS plays a small part in saving lives from breast cancer. Treatment decisions in DCIS, as in invasive carcinoma, should take full account of histopathological, clinical and radiological attributes of the tumour.

All-cause mortality among breast cancer patients in a screening trial: support for breast cancer mortality as an end point.

BACKGROUND: It has recently been suggested that all-cause mortality is a more appropriate end point than disease specific mortality in cancer screening trials, and that disease specific mortality is biased in favour of screening. This suggestion is based partly on supposed inconsistencies between all-cause mortality results and disease specific results in cancer screening trials, and alleged increases in deaths from causes other than breast cancer among breast cancer cases diagnosed among women invited to screening. METHODS: We used data from the Swedish Two-County Trial of mammographic screening for breast cancer, in which 77 080 women were randomised to an invitation to screening and 55 985 to no invitation. We estimated relative risks (RRs) (invited v control) of death from breast cancer, death from other causes within the breast cancer cases, and death from all causes within the breast cancer cases. RRs were adjusted for age and took account of the longer follow up of breast cancer cases in the invited group due to lead time. RESULTS: There was a significant 31% reduction in breast cancer mortality in the invited group (RR 0.69, 95% confidence interval (CI) 0.58-0.80; p<0.001). There was no significant increase in deaths from other causes among breast cancer cases in the invited group (RR 1.12, 95% CI 0.96-1.31; p=0.14). A significant 19% reduction in deaths from all causes was observed among breast cancer cases in the group invited to screening (RR 0.81, 95% CI 0.72-0.90; p<0.001). A more conservative estimation gave a significant 13% reduction (RR 0.87, 95% CI 0.78-0.97; p=0.01). These findings are consistent with the magnitude of the reduction in breast cancer mortality. CONCLUSIONS: Invitation to screening was associated with a reduction in deaths from all causes among breast cancer cases, consistent with high participation rates in screening. There is no significant evidence of bias in cause of death classification in the Two-County Trial, and as breast cancer mortality is the targeted clinical outcome in breast cancer screening, it is the appropriate end point in a breast cancer screening trial. All-cause mortality is a poor and inefficient surrogate for breast cancer mortality.

Interval cancers in the Norwegian breast cancer screening program: frequency, characteristics and use of HRT.

Breast cancers diagnosed between screening examinations among women who attend a breast cancer screening program are defined as interval cancers. The Norwegian Breast Cancer Screening Program started as a pilot project in 1996, and data from the first 2-year interval are available. Our study quantifies interval cancers in the pilot project and explores characteristics and factors that may be associated with interval cancer. Interval cancers in the screening population were identified through the Cancer Registry of Norway. The frequency of invasive interval cancer was calculated as cases per 10,000 screened and as observed/expected ratio. Characteristics of the interval cancers were compared to screening-detected and clinical cancers. Breast density was assessed in a blinded review of 3 categories of screening mammograms. Information on hormone replacement therapy (HRT) use was collected from a questionnaire. The frequency of invasive interval cancers was 18.2 (15.9-20.7) per 10,000 screened and the observed/expected ratio was 0.49 (0.43-0.56). The frequency in the second year of the interval was higher than reported from other programs. The median tumor size of the interval cancers was 19.5 mm and 44.0% of the patients had affected axillary lymph nodes. The interval cancer cases had higher proportions of dense breasts and reported use of HRT compared to screen normal and screening-detected cases. The reported frequency of interval cancers is similar to comparable programs. The interval cancers differed significantly from the cancers detected in the first screening round and were more similar to clinical cancers. Interval cancer was associated with dense breasts and use of HRT. Screening programs must keep these associations in focus.

Beyond randomized controlled trials: organized mammographic screening substantially reduces breast carcinoma mortality.

BACKGROUND: The efficacy of mammographic screening in the reduction of breast carcinoma mortality has been demonstrated in randomized controlled trials. However, the evaluation of organized screening outside of research settings (so-called "service screening") faces unique methodologic and conceptual challenges. The current study describes the evaluation of organized mammography screening in a clinical setting and demonstrates the benefit obtained from service screening in two Swedish counties. METHODS: In the group of subjects ages 20--69 years, there were 6807 women diagnosed with breast carcinoma over a 29-year period in 2 counties in Sweden and 1863 breast carcinoma deaths. All patients were classified from patient charts based on their screening status (i.e., whether they had been invited to undergo screening and whether they actually had undergone screening). The number of women who lived in the 2 counties during the 29-year study period was provided by the Central Bureau of Statistics. Breast carcinoma-specific mortality was compared across three time periods: 1) 1968--1977, when no screening was taking place because mammography had not been introduced; 2) 1978--1987, the approximate period of the Two-County randomized controlled trial of screening in women ages 40--74 years; and 3) 1988--1996, when all women in the 2 counties ages 40--69 years were invited to undergo screening (service screening). When comparing breast carcinoma mortality in screened women with that in women diagnosed before screening was introduced, a correction for self-selection bias was incorporated to prevent overestimation of the benefit of screening. RESULTS: The mortality from incident breast carcinoma diagnosed in women ages 40-69 years who actually were screened during the service screening period (1988--1996) declined significantly by 63% (relative risk [RR] = 0.37; 95% CI, 0.30--0.46) compared with breast carcinoma mortality during the time period when no screening was available (1968--1977). The mortality decline was 50% (RR = 0.50; 95% CI, 0.41--0.60) when breast carcinoma mortality among all women who were invited to undergo screening (nonattendees included) was compared with breast cancer mortality during the time period when no screening was available (1968--1977). The reduction in mortality observed during the service screening period, adjusted for selection bias, was 48% (RR = 0.52; 95% CI, 0.43--0.63). No significant change in breast carcinoma mortality was observed over the three time periods in women who did not undergo screening. This group included women ages 20--39 years because these individuals were never invited to undergo screening, and women ages 40--69 years who did not undergo screening (not invited during the randomized trial or invited during the second and third time periods but declined). CONCLUSIONS: Regular mammographic screening resulted in a 63% reduction in breast carcinoma death among women who actually underwent screening. The policy of invitation to organized screening with mammography appears to have reduced breast carcinoma mortality by 50% in these 2 counties.

The Swedish Two-County Trial twenty years later. Updated mortality results and new insights from long-term follow-up.

The benefit of invitation to mammographic screening observed in this trial is maintained as a highly significant 32% reduction in breast cancer mortality. Mammographic screening for breast cancer continues to save lives after up to 20 years. Screening derives this benefit by improving the distribution of tumors diagnosed with respect to prognostic categories based on node status, size, and histology of tumors. There is potential for modern screening programs with shorter interscreening intervals to achieve even greater improvements in prognostic category and greater reductions in breast cancer mortality. Mammography can discriminate a subpopulation of high-risk cases, those displaying casting-type calcifications on the mammogram, among very small tumors, with fundamental implications for diagnosis and treatment.

Effect on sensitivity and specificity of mammography screening with or without comparison of old mammograms.

PURPOSE: To evaluate the effect of old mammograms on the specificity and sensitivity of radiologists in mammography screening. MATERIAL AND METHODS: One hundred and fifty sets of screening mammograms were examined by 3 experienced screeners twice: once without and once in comparison with older mammograms. The films came from a population-based screening done during the first half of 1994 and comprised all 35 cancers detected during screening in 1994, 12/24 interval cancers, 14/34 cancers detected in the following screening and 89 normal mammograms. RESULTS: Without old mammograms, the screeners detected an average of 40.3 cancers (range 37-42), with a specificity of 87% (85-88%). With old mammograms, the screeners detected 37.7 cancers (range 34-42) with a specificity of 96% (94-99%). The change in detection rate was not significant. However, the increase in specificity was significant for each screener (p = 0.0002-0.03). CONCLUSION: Mammography screening with old mammograms available for comparison decreased the false-positive recall rate. The effect on sensitivity, however, was unclear.

Some random-effects models for the analysis of matched-cluster randomised trials: application to the Swedish two-county trial of breast-cancer screening.

BACKGROUND: In the Swedish two-county trial women aged 40-74 years from two counties in Sweden were randomised to invitation to mammographic screening for breast cancer. METHODS: This paper uses random effects logistic regression models to analyse recent data from the trial. The analysis accounts for the structure of the trial, where small geographical units are randomised within larger geographical strata (blocks of two or three small units that are socio-economically similar). RESULTS: Fixed effects and a variety of random effects models show a strong degree of agreement and yield a significant 29% or 30% reduction in breast-cancer mortality. DISCUSSION: Fixed effects and random effects models agree for this example, because heterogeneity both between strata and within strata between clusters is small and because the effect of treatment does not vary much in different strata.

Update of the Swedish Two-County Trial of breast cancer screening: histologic grade-specific and age-specific results.

UNLABELLED: The benefit of mammographic screening in reducing mortality from breast cancer is well established. Questions remain with respect to the magnitude of the long-term benefit of modern mammography screening, age specific benefits and the timing of these, and histology specific effects. METHODS: The Swedish Two-County Trial was set up in 1977, with 77,080 women aged 40-74 randomised to invitation to mammographic screening for breast cancer (active study population, ASP) and 55,985 women randomised to no invitation (passive study population, PSP). There is now follow-up for mortality to 31 December 1996, approximately 18 years average follow-up. We investigated the effect of invitation of screening on breast cancer mortality and incidence of advanced tumours by age group (40-49 and 50-74) and histologic type. In addition we estimated progression rates by histologic grade using markov chain models. RESULTS: A significant 29% reduction in breast cancer mortality was observed in association with invitation to screening (relative risk = 0.71, 95% confidence interval 0.60-0.83), maintaining the effect observed at previous stages of follow-up. Age-specific analyses show a smaller and later mortality benefit in women aged 40-49. This is related to the fact that there is a considerable benefit from early detection in terms of mortality from aggressive, poorly differentiated cancers in women aged 50-74, whereas the major effect in women aged 40-49 is on the less aggressive tumours of good or intermediate differentiation. Among women aged 50-74, the incidence of grade III tumours in the ASP is significantly lower than in the PSP, but this is not the case for women aged 40-49. This is related to the greater prevalence and rapidity of progression with respect to histologic grade, as evidenced by the results of markov chain models and the proportions of grade III tumours by time since last screen. CONCLUSIONS: The substantial and significant mortality benefit of invitation to mammographic screening in women aged 40-74 is maintained at 18 years of follow-up. To achieve a substantial mortality benefit at an early stage in the screening program in women aged under 50 years, an interscreening interval of 12-18 months would be required.

The natural history of breast carcinoma: what have we learned from screening?

BACKGROUND: The availability of breast carcinoma data from trials of mammographic screening provides an opportunity to study the natural history of breast carcinoma. METHODS: The Swedish Two-County study is a randomized, controlled trial of mammographic screening for breast carcinoma in which 77,080 women were randomized to receive an invitation to mammographic screening and 55,985 were randomized to receive no invitation. During the trial, a total of 2468 breast carcinoma cases were diagnosed. The authors examined the effect of screening on the pathologic attributes of the tumors diagnosed, mortality and survival from breast carcinoma, and the consequences of arresting tumor development by screening. RESULTS: Screening reduces mortality from breast carcinoma largely through its effect in detecting tumors at a smaller size, decreasing the probability of lymph node metastases, and reducing the opportunity for worsening of the grade of malignancy of the tumor. CONCLUSIONS: Breast carcinoma is not a systemic disease at its inception, but is a progressive disease and its development can be arrested by screening. The point at which the tumor's progression is arrested is crucial. Detection of small (<15 mm) and lymph node negative invasive tumors will save lives and confer an opportunity for less radical treatment. Tumor progression in the preclinical phase occurs more rapidly in women age <50 years, suggesting the need for a shorter screening interval for this group.

Tumour characteristics and survival in patients with invasive interval breast cancer classified according to mammographic findings at the latest screening: a comparison of true interval and missed interval cancers.

The aim of this study was to investigate whether different mammographic categories of interval cancer classified according to findings at the latest screening are associated with different distributions of prognostic factors or with different survival rates. The series consisted of all patients with invasive interval cancer detected from May 1978 to August 1995 (n = 544). The tumours were evaluated with regard to age, radiological category, interval between the latest screen and diagnosis and tumour characteristics at the time of diagnosis. We investigated possible relationships between the survival rate of patients with interval cancer and the interval between the latest screen and diagnosis, tumour characteristics and radiological category of the interval tumours. The study focused on comparison of patients with true interval and missed interval cancer. Women with mammographically occult tumours were younger than those in the other radiological categories. Comparisons of true interval cancers with overlooked or misinterpreted tumours showed equal distributions of age, tumour size, TNM stage and lymph node status. The overlooked or misinterpreted tumours showed significantly higher proportions of grade-I tumours (22 vs 11 %), tumours with low S-phase fraction (SPF; 44 vs 24 %) and oestrogen receptor (ER) positive tumours (72 vs 57 %). However, analyses of survival rates disclosed no clear differences between the two radiological categories. Radiological category and interval between the latest screen and diagnosis were not genuine predictors of the prognosis in patients with invasive interval breast cancer. No certain prognostic difference existed between true interval cancers and overlooked or misinterpreted interval breast cancers, despite higher proportions of grade-I tumours, ER positive tumours and tumours with low SPF in the latter group.

Invasive interval cancers in the Ostergötland Mammographic Screening Programme: radiological analysis.

The aim of this study was to find out by radiological analysis whether it is possible to prevent interval cancers in screening programmes without a concomitant increase in false positives. Re-scrutiny of all mammograms of 544 interval cancers was implemented. Blinded re-scrutiny of screening mammograms of 390 interval cancers mixed with other cases (prevalence of detectable cases 37 %) by two outside radiologists was also carried out. True interval cancers constituted 49 %, missed tumours 25 %, occult tumours 10 % and unclassifiable cases 16 %. There were no significant differences in survival between true interval cancers, overlooked or misinterpreted cancers, and mammographically occult cancers. In the blinded re-scrutiny both examiners reached a positive predictive value for malignancy of 67 %, and a sensitivity of 69 % and 81 %, respectively. In a screening programme such as this, it was estimated that 100-400 additional women would have to be recalled from screening for further workup in order to detect each additional cancer which would otherwise emerge as an interval cancer. The occurrence of interval cancer increased with age and was influenced by the design of screening. The proportion of patients with potential iatrogenic delay in diagnosis (overlooked and misinterpreted cancers) was equivalent to figures from other studies. There were no significant differences in survival between the various categories of interval cancer. The study confirmed the association of criteria for referral for further investigation with the number of false positives. Efforts to reduce the number of interval cancers by lowering the threshold for recall are likely to be counterproductive.

Interval cancers and cancers in non-attenders in the Ostergötland Mammographic Screening Programme. Duration between screening and diagnosis, S-phase fraction and distant recurrence.

The study was based on a population mammographic screening programme for women aged 40-74 years. Metastatic potential was analysed in 843 invasive breast cancers with regard to mode of detection and a number of prognostic factors. There was a higher metastatic capacity in clinically detected cases, but multivariate analyses showed that neither the mode of detection (hazard rate ratio of distant recurrence RR = 1.39, 95% CI 0.78-2.46 interval cancers and RR = 1.6, 95% CI 0.76-3.36 non-attenders) nor the duration between screening and diagnosis for true interval cancers (RR = 0.47, 95% CI 0.16-1.35 in tumours detected later than one year after screening) were independent prognostic factors. A correlation was found between metastatic potential and the SPF (RR = 2.94, 95% CI 1.57-5.50 in tumours with a high SPF), the oestrogen receptor status and the tumour stage. In conclusion, interval cancers intrinsically are not different from other breast cancers with equivalent characteristics; the duration between screening and diagnosis in interval cancers was not clearly correlated to the prognosis, but the S-phase fraction was a powerful predictor of prognosis.

Diagnostic outcome of repeated mammography screening.

Mammographic screening for breast cancer within health service routines was evaluated for the years 1987-1992, with special focus on repeated screening during 1989-1992. The overall attendance rate by women aged 40 to 74 years was 82.8%. During 1989-1992 malignancy was found in 2.6/1000 screened women, giving a 87.4% positive predictive rate at surgery and 95.9% efficiency. Among women aged > or = 45 years, the positive predictive rate was > 94%. Fine-needle aspiration (FNA) biopsy showed invasive cancers in 84% and highly suspected cancer in another 15%; 60% of the lesions were nonpalpable. For first-time (prevalence) screening (1987-1988) the positive predictive rate was 86% and the malignancy yield 6.4/1000. In women aged 40-44 years there were few surgical referrals (1.6%), but the positive predictive rate at surgery was only 48.3%, indicating diagnostic difficulties in young women. The median size of all invasive cancers was 12 mm: 84% were classified as pT1, and 23% had lymph node involvement. Stage II disease was found in 27% of all malignancies. The use of FNA in the diagnostic workup for breast cancer screening is of crucial importance to the maintenance of high positive predictive rates at surgery. Moreover, regular analysis is important even when mammographic screening is incorporated into the routine work of health services.

Comparison of image quality and mean absorbed dose to the breast for two mammographic films.

A film intended for mammography (Kodak SO 155 MRH-1) was tested and compared to one (Kodak SO 177 Ortho M) used earlier. Both films including the cassette could resolve 20 lp/mm. For a processing time of 90 s the new film gave the same overall image quality and irradiation dose to the breast as the old system. If, instead, the processing time was increased to 150 s, a 43% reduction in kerma could be attained with the new film. With a developing temperature of 36 degrees C and 150 s processing time, the noise is clinically acceptable. A nonparametric test showed no significant difference between the 2 films on the 0.01 level. At an X-ray tube potential difference of 25 kV, the mean absorbed dose to a 4.5-cm-thick breast was reduced from 1.7 mGy with the old combination to 1.0 mGy with the new one. The measurements were made with a moving grid.