Cumulative incidence of advanced breast cancer in women aged 40-49 years in the

# Cumulative incidence of advanced breast cancer in women aged 40–49 years in the Japan Strategic Anti-cancer Randomised Trial (J-START) of adjunctive ultrasonography: a prespecified secondary analysis *The Lancet 2026* Articles Cumulative incidence of advanced breast cancer in women aged 40–49 years in the Japan Strategic Anti-cancer Randomised Trial (J-START) of adjunctive ultrasonography: a prespecified secondary analysis Narumi Harada-Shoji*, Akihiko Suzuki*, Takanori Ishida*, Seiichiro Yamamoto, Seiki Kanemura, Takuhiro Yamaguchi, Yoko Shiono-Narikawa, Noriaki Ohuchi, on behalf of the J-START investigators† Summary Lancet 2026; 407: 784–93 Background The J-START randomised controlled trial found that adjunctive ultrasonography was associated with See Comment page 737 significantly higher rates of breast cancer detection than mammography alone. This report aims to evaluate the long- *Joint first authors term eect of adjunctive ultrasonography screening on the cumulative incidence of advanced breast cancer as a prespecified secondary outcome of J-START. †Collaborator list can be found in the appendix (p 79) Department of Breast and Methods We enrolled asymptomatic women in 42 study sites in 23 of 47 prefectures in Japan. Eligible women were Endocrine Surgical Oncology aged 40–49 years without a history of breast cancer, including in-situ cancer, or other cancers in the previous 5 years, (N Harada-Shoji MD PhD, and who had a life expectancy of more than 5 years. Participants were assigned in a 1:1 ratio to undergo Prof T Ishida MD PhD, ultrasonography with mammography (intervention group) or mammography alone (control group) twice during a Y Shiono-Narikawa PhD, Prof N Ohuchi MD PhD) and 2-year screening period by individual or cluster randomisation. Participants attended an initial screening Division of Biostatistics appointment and were then asked to return for a second screening after 2 years. This prespecified secondary analysis (Prof T Yamaguchi PhD), Tohoku assessed the cumulative incidence of stage 2 or higher breast cancers based on the TNM classification up to data University Graduate School of cuto on Oct 4, 2024. Medicine, Sendai, Japan; Department of Breast and Endocrine Surgery, Tohoku Findings Between Aug 2, 2007, and March 31, 2011, 72 661 asymptomatic women aged 40–49 years were randomly Medical and Pharmaceutical assigned (36 723 in the intervention group and 35 938 in the control group). Median follow-up for this secondary University, Sendai, Japan analysis was 11·4 years (range 0·0–16·1; IQR 9·3–12·9) in the intervention group and 11·3 years (0·0–16·1; 8·9–12·9) (Prof A Suzuki MD PhD); Division of Cancer in the control group. Of 894 breast cancers detected in the intervention group, 234 (26%) were advanced cancers, Epidemiology and Prevention, compared with 277 (33%) of 843 detected breast cancers in the control group (hazard ratio 0·83 [95·6% CI 0·70–0·98]; Miyagi Cancer Center Research p=0·026). Kaplan–Meier curves suggested a violation of the proportional hazards assumption, with a significant Institute, Natori, Japan dierence in advanced cancer incidence only between 48 months and 96 months. Divergence between groups (S Kanemura MD PhD); Graduate School of Public emerged around year four, widened until year eight, and remained stable thereafter. Health, Shizuoka Graduate University of Public Health, Interpretation Adjunctive ultrasonography reduced the cumulative incidence of advanced breast cancer in women Shizuoka, Japan (Prof S Yamamoto PhD) aged 40–49 years. These findings highlight the potential value of integrating adjunctive ultrasonography into screening programmes for women with dense breast tissue, particularly in Asian populations, and could inform Correspondence to: Dr Narumi Harada-Shoji, future breast cancer screening guidelines. Department of Breast and Endocrine Surgical Oncology, Funding The Ministry of Health, Labor, and Welfare (Japan) and Japan Agency for Medical Research and Development. Tohoku University Graduate School of Medicine, Sendai 980–8575, Japan Copyright © 2026 Elsevier Ltd. All rights reserved, including those for text and data mining, AI training, and similar narumi.harada.a3@ technologies. tohoku.ac.jp See Online for appendix Introduction aged between 40 years and 49 years have dense breast Breast cancer is the most common malignant tumour tissue,7 highlighting the potential value of adjunctive in women worldwide, and the incidence of breast ultrasound screening in this population. cancer in Japanese women substantially increases after The Japan Strategic Anti-cancer Randomised Trial the age of 40 years.1 Mammography is the only breast (J-START) is a randomised controlled trial designed to cancer screening method proven to reduce mortality.2–5 evaluate the ecacy of adjunctive ultrasonography as a However, mammography does not achieve sucient breast cancer screening method for women aged screening accuracy for young women due to their 40–49 years. The study compared ultrasonography with dense breast tissue—women aged between 40 years mammography (intervention group) and mam mography and 49 years often have dense breast tissue on alone (control group). The primary endpoints were mammography that can obscure lesions and reduce sensitivity, specificity, and cancer detection rate.6 sensitivity.6 In Japan, approximately 60–70% of women Adjunctive ultrasonography with mammography showed 784 Articles Research in context Evidence before this study ultrasonography significantly reduced the incidence of We searched the Cochrane Database of Systematic Reviews, advanced breast cancer for up to 8 years following the second which published a comprehensive analysis in 2023 evaluating screening. The study included 72 661 women followed up over the added value of ultrasonography to mammography in breast more than 15 years (Aug 2, 2007, to Oct 4, 2024), with a cancer screening. The review included eight studies: minimal loss to follow-up. These results suggest that adjunctive two prospective cohort studies, five retrospective cohort ultrasonography with mammography is a highly effective studies, and one randomised controlled trial (J-START), which strategy for the early detection of breast cancer in Asian women was assessed to have low risk of bias. To identify any additional aged between 40 and 49 years with dense breast tissue, who randomised controlled trials published since then, we are the target population of J-START. conducted an updated search of the Cochrane Database of Implications of all the available evidence Systematic Reviews, as well as MEDLINE and PubMed on This study is the first randomised controlled trial to show a Aug 5, 2025 for studies published between Nov 22, 2014, and reduction in the cumulative incidence of advanced breast Aug 5, 2025, using keywords related to “breast cancer”, cancer, which could translate into decreased breast cancer “screening”, “mammography”, “ultrasound”, and “randomised mortality in the future. In cancer screening trials, mortality is controlled trial”. This search identified the BRAID trial, in which often the most important endpoint; however, the favourable abbreviated MRI and contrast mammography detected prognosis of breast cancer limits the number of mortality three times more invasive cancers than automated whole events that can be observed within a short follow-up period. breast ultrasound. BRAID differs from J-START in age, modality, Despite the fact that an observed reduction in advanced and follow-up. cancer might suggest a potential decline in mortality, the Added value of this study validity of using advanced cancer incidence as a surrogate This report presents the extended follow-up results of J-START, endpoint remains under debate. Although current evidence is a large-scale randomised controlled trial investigating the effect still limited, our findings might contribute to the use of of adjunctive ultrasonography in breast cancer screening advanced breast cancer incidence as a potential surrogate among women aged 40–49 years. Unlike previous reports, this indicator. Continued follow-up is warranted to confirm analysis focused specifically on the cumulative incidence of whether adjunctive ultrasonography reduces breast cancer advanced breast cancer. Our findings show that the addition of mortality. higher sensitivity and lower specificity, while detecting a a life expectancy of more than 5 years. Gender was greater number of cancers, including a higher frequency defined as women who were registered as eligible for of cancers at stages 0 and 1, and fewer interval cancers breast cancer screening by local governments or compared to mammography alone.6 workplaces. Data on race and ethnicity were not Several clinical trials have shown that adjunctive collected. In this prespecified follow-up analysis, we ultrasonography improves screening sensitivity and assessed cumulative incidence of advanced breast cancer detection rates.8–11 Furthermore, J-START showed cancers up to Oct 4, 2024. Outcome ascertainment was that adjunctive ultrasonography provides a favourable based on trial screening data, reports from screening screening balance for asymptomatic women aged facilities, participant self-reports, regional cancer 40–49 years, regardless of breast density.7 In this Article, registries, and national vital statistics, with medical we report follow-up data from J-START comparing the record confirmation for self-reported cases. cumulative incidence of advanced breast cancers in the The protocol for J-START was developed in accordance two groups. with the Declaration of Helsinki and complied with the ethical guidelines for clinical studies (Ministry of Methods Education, Culture, Sports, Science and Technology; Study design and participants Ministry of Health, Labor, and Welfare, Japan). Ethical The research design of J-START has been previously approval was obtained from the Tohoku University described.6 Briefly, between August, 2007, and March, School of Medicine Research Ethics Committee 2011, we enrolled asymptomatic women at 42 study sites (2008-340, 2010-279, 2011-421, 2012-1-141, 2013-01–98, (10 hospitals or hospital-aliated screening centres and 2015-1-203, 2015-1-432, 2016-1-324, 2017-1-171, 2017-1-1028, 32 screening organisations, including prefectural cancer 2019-1-225, 2020-1-36, 2020-1-157, 2021-1-224, 2022-1-450, societies, health-service associations, and health 2023-1-066, 2023-1-193) and Japanese Anti-cancer Society promotion centres) in 23 of 47 prefectures in Japan. (gann-002, gann-002-amendment). Written informed Eligible women were aged between 40 and 49 years consent was obtained from all participants, and this without a history of breast cancer, including in-situ clinical trial was registered with UMIN-CTR Japan cancer, or other cancers in the previous 5 years, and had (UMIN000000757). Articles Randomisation and masking 110 communities from eight participating research Participants in J-START were assigned in a 1:1 ratio by institutions were allocated as clusters (44 to the individual or cluster randomisation to undergo intervention group and 66 to the control group). The ultrasonography with mammography (intervention number of communities per participating research group) or mammography alone (control group) across institution ranged from a minimum of 6 to a maximum two screening rounds (an initial screening and a second of 32. The number of individuals per cluster ranged from screening 2 years later). Randomisation was done centrally a minimum of five to a maximum of 781. by the Japan Clinical Research Support Unit. Detailed procedures for this randomisation were decided by the Procedures Director of the Japan Clinical Research Support Unit and As part of J-START, participants underwent two screening have not been revealed to others. At the trial design stage, visits (mammography and ultrasonography in the sites unable to implement individual randomisation due intervention group or mammography alone in the to logistical or ethical constraints were identified and pre- control group). J-START’s primary outcome was assigned to cluster randomisation. The number of clusters sensitivity, specificity, cancer detection rate, and stage and expected cluster sizes were estimated based on distribution at the first round of screening. Interval historical screening data at each site, and sample size was cancers were defined as cancers diagnosed between the calculated assuming all participants were individually first and second screening rounds among women randomised, given the expected negligible intra-cluster categorised as being Breast Imaging Reporting and Data correlation. System categories 1 and 2 at the first round. For sites using individual randomisation, the In this prespecified, long-term, follow-up analysis, we participating research institutions were regarded as extended surveillance for incident breast cancer and vital stratified factors (blocks), and a permuted block method status until Oct 4, 2024. Follow-up information was was used at each institution so that the number of obtained from: 1) the two rounds of screening conducted patients allocated to each group within a single through the trial, 2) reports from screening facilities, participating research institution is approximately equal. 3) self-reports from participants by mail or telephone, In addition, a variable block method was used to vary 4) regional cancer registry, and 5) national vital statistics. block size to avoid predictability. A random number When participants self-reported a breast cancer generator and allocation program were created in an SAS diagnosis, we contacted the relevant medical institutions data step. The block sizes are 4, 6, and 8. Uniform to obtain clinical and pathological details. We identified random numbers are generated, and blocks of 4, 6, and 8 the number of breast cancer cases and analysed the are generated with probabilities of 0·5, 0·25, and 0·25. corresponding data. Before informed consent was obtained, information After completion of the two trial screening rounds, about allocation remained undisclosed (allocation is participants were not routinely rescanned as part of the concealed) to the study participants and to the sta in trial protocol. Following the second round, invitations for charge at each participating research institution. population-based breast cancer screening were issued Cluster randomisation was used at participating according to public health programmes, and participation research institutions where it was dicult to implement in these screenings varied by individual. Long-term screening based on individual randomisation. At the follow-up during the study period, from 2007 to 2023, stage when this study was designed, sample sizes for was conducted to ascertain incident breast cancer and cluster randomisation and each cluster size were not vital status. Follow-up information was collected through specified. Allocation was decided after considering the periodic follow-up questionnaires, supplemented by ability to implement the study at each participating reports from screening facilities, and linkage and research institution. Regarding allocation by the Japan confirmation through regional cancer registries and Clinical Research Support Unit, communities were national vital statistics, up to the data cuto of Oct 4, 2024. regarded as units (clusters), and cluster randomisation For participants whose survival could not be verified was performed for each participating research institution. through mail or telephone interviews, confirmation was The Japan Clinical Research Support Unit confirmed obtained by reviewing the residence records. In the event beforehand the number of eligible patients who were of death, cause was determined using vital statistics. The scheduled for screening in the community and with duration of follow-up was not predefined as a fixed health-care providers at the participating research period, but continued until the occurrence of an outcome institutions. Based on the scheduled number of patients of interest, death, or the data cuto. for each community, allocation was determined so that Initially, we recalculated the incidence rates of breast the number of participants would be nearly the same in cancers detected at the first screening, interval cancers the two groups. Because the correlation of outcomes in diagnosed between the first and second screenings, and clusters was assumed to be lower, greater priority was cancers detected at the second screening up to data given to the number of participants than the number of cuto based on participants who underwent screening clusters. as defined by the J-START protocol. Furthermore, 786 Articles participants who either did not undergo the second screening or who were diagnosed with breast cancer 76 196 women enrolled after the scheduled second screening were classified as being diagnosed at another period. 3198 excluded The TNM staging system was used to classify disease 3101 not randomised stages. In this study, advanced breast cancer was defined 9 enrolled twice in error 88 other reasons as Stage 2 or higher (node positive, T2 or higher, or both), while stages 0 and 1 are defined as early-stage cancer. Disease stages were determined based on 72 998 randomly assigned comprehensive information from clinical and pathological stagings, obtained through medical institutions and regional cancer registry data. Patients 36 859 assigned to undergo 36 139 assigned to undergo with distant metastasis were classified as stage 4. When mammography plus mammography alone† neoadjuvant chemotherapy was administered or the ultrasonography* pathological stage was unavailable, the clinical stage was used (determined by the treating clinicians based on 107 excluded before first screening 174 excluded before first screening pretreatment clinical examination and imaging as 9 had breast cancer 6 had breast cancer recorded in the medical record). 9 had a history of other cancers 11 had a history of other cancers within the last five years within the last five years In cases of bilateral breast cancer where both breasts 70 withdrew 124 withdrew are aected at the same time, the more advanced stage 19 did not receive 33 did not receive was recorded. Conversely, if a new cancer developed on mammography mammography the contralateral side and had reached stage 2 or higher, it was classified as advanced breast cancer. The definition 36 752 screened‡ 35 965 screened§ for disease staging was approved by the statistical analysis committee and data monitoring committee. 29 withdrew 27 excluded Since regional cancer registries generally use summary 1 had breast cancer stage rather than the TNM classification, we applied a set 26 withdrew of predefined conversion rules derived from registry definitions to map the summary stage categories to the 36 723 analysed 35 938 analysed corresponding TNM stages. Figure 1: Trial profile Outcomes Japanese women aged 40–49 years were enrolled and randomly assigned in a 1:1 ratio to undergo mammography The primary endpoint of the trial was sensitivity, with ultrasonography (intervention) or mammography alone (control) every 2 years. *26 434 enrolled by specificity, and cancer detection rate, and the results individual randomisation and 10 245 by cluster randomisation. †26 411 enrolled by individual randomisation and have been reported previously.6 This Article reports on 9278 by cluster randomisation. ‡Four women did not undergo ultrasonography. §Five women underwent ultrasonography. the prespecified secondary outcome of advanced cancer detection and mortality between the two groups. The beginning of the study, we planned the statistical analysis time-to-event was defined from randomisation to the of the cumulative incidence of advanced cancer with first occurrence of advanced breast cancer (stage 2 or strict adjustment for alpha error including one interim higher). Breast cancer death was also counted as an analysis. Owing to one interim analysis, multiplicity was event if advanced breast cancer had not been documented controlled with Lan–DeMets α-spending function before death. Follow-up was considered complete at the (O’Brien–Fleming type). Therefore, we report 95·6% CIs date of the most recent follow-up if events were not for hazard ratios (HR) at the final analysis, corresponding observed. If there was no event, or if the participants had to an adjusted 95% CI after this interim analysis. The not died, time-to-event data were censored as of the date sample size for the analysis of the cumulative incidence of the most recent follow-up. The participants who died of advanced breast cancers was defined to detect a of causes other than breast cancer were censored at the clinically reasonable dierence (HR 0·75) in terms of the date of death. number of events. The target events were 385 with one-sided α=0·025 and a power of 80%. For the long-term Statistical analysis follow-up (secondary endpoint), no additional sample To investigate the ecacy of cancer screening, breast size was planned beyond the original randomised cohort; cancer mortality was compared in the two groups. In this the planned group sizes for this analysis therefore trial, the cumulative incidence of advanced cancer was corresponded to the trial’s original 1:1 allocation. Rather set as a secondary outcome, and the analysis was based than increasing the sample size, follow-up was continued on the modified intention-to-treat principle, which to accrue the required number of events. Additionally, a included predefined exclusions based on post- competing risk analysis was also conducted as a randomisation information or behaviours. At the sensitivity analysis. Articles All participants Intervention Control group All participants Intervention Control group (n=72 661) group (n=35 938) (n=72 661) group (n=35 938) (n=36 723) (n=36 723) Age at 44·0 (5·0) 44·0 (5·0) 44·0 (5·0) (Continued from previous column) randomisation Number of pregnancies delivered (years) Nulliparous 9493 (13·1%) 4849 (13·2%) 4644 (12·9%) Ever undergone breast cancer screening 1 11 010 (15·2%) 5559 (15·1%) 5451 (15·2%) Yes 55 798 (76·8%) 28 289 (77·0) 27 509 (76·5%) 2 32 119 (44·2%) 16 164 (44·0%) 15 955 (44·4%) No 16 851 (23·2%) 8424 (22·9%) 8427 (23·4%) 3 14 969 (20·6%) 7635 (20·8%) 7334 (20·4%) Unknown or 12 (<0·1%) 10 (<0·1%) 2 (<0·1%) data missing 4–8 2155 (3·0%) 1090 (3·0%) 1065 (3·0%) Method of most recent breast cancer screening Unknown or 2915 (4·0%) 1426 (3·9%) 1489 (4·1%) data missing Mammography Age at first birth (years) Yes 39 498 (54·4%) 20 008 (54·5%) 19 490 (54·2%) <20 675 (0·9%) 354 (1·0%) 321 (0·9%) No 16 290 (22·4%) 8279 (22·5%) 8011 (22·3%) 20–24 14 594 (20·1%) 7465 (20·3%) 7129 (19·8%) Unknown or 16 873 (23·2%) 8436 (23·0%) 8437 (23·5%) data missing 25–29 28 461 (39·2%) 14 387 (39·2%) 14 074 (39·2%) Ultrasonography 30–39 15 792 (21·7%) 7883 (21·5%) 7909 (22·0%) Yes 10 555 (14·5%) 5452 (14·8%) 5103 (14·2%) 40–49 430 (0·6%) 222 (0·6%) 208 (0·6%) No 45 233 (62·3%) 22 835 (62·2%) 22 398 (62·3%) Unknown or 12 709 (17·5%) 6412 (17·5%) 6297 (17·5%) data missing Unknown or 16 873 (23·2%) 8436 (23·0%) 8437 (23·5%) data missing Ever breastfed children Clinical breast examination Yes 56 183 (77·3%) 28 415 (77·4%) 27 768 (77·3%) Yes 50 545 (69·6%) 25 652 (69·9%) 24 893 (69·3%) No 14 953 (20·6%) 7575 (20·6%) 7378 (20·5%) No 5243 (7·2%) 2635 (7·2%) 2608 (7·3%) Unknown or 1525 (2·1%) 733 (2·0%) 792 (2·2%) data missing Unknown or 16 873 (23·2%) 8436 (23·0%) 8437 (23·5%) data missing Number of first–degree female relatives with breast cancer 0 69 250 (95·3%) 34 960 (95·2%) 34 290 (95·4%) Age at menarche (years) 1 3342 (4·6%) 1726 (4·7%) 1616 (4·5%) 7–11 15 257 (21·0%) 7656 (20·8%) 7601 (21·1%) >1 69 (0·1%) 37 (0·1%) 32 (0·1%) 12–13 41 653 (57·3%) 21 077 (57·4%) 20 576 (57·3%) 14–18 15 656 (21·5%) 7939 (21·6%) 7717 (21·5%) Ever had breast 1462 (2·0%) 754 (2·1%) 708 (2·0%) surgery Others or data 95 (0·1%) 51 (0·1%) 44 (0·1%) missing Ever had benign 917 (1·3%) 489 (1·3%) 428 (1·2%) neoplasm Menopausal status Ever had breast 538 (0·7%) 264 (0·7%) 274 (0·8%) Premenopausal 54 959 (75·6%) 27 715 (75·5%) 27 244 (75·8%) inflammation Perimenopausal 13 389 (18·4%) 6774 (18·4%) 6615 (18·4%) Table 1: Demographic characteristics Postmenopausal 4269 (5·9%) 2207 (6·0%) 2062 (5·7%) Unknown or 44 (0·1%) 27 (0·1%) 17 (0·1%) data missing O’Brien–Fleming-type α-spending function was used. Number of pregnancies The results were reviewed by an independent committee 0 8740 (12·0%) 4423 (12·0%) 4317 (12·0%) specially organised for the interim analysis. The current 1 8802 (12·1%) 4500 (12·3%) 4302 (12·0%) study was designed and conducted based on one-sided 2 25 206 (34·7%) 12 650 (34·4%) 12 556 (34·9%) testing, but the presented results were expressed using 3–4 24 083 (33·1%) 12 233 (33·3%) 11 850 (33·0%) two-sided p values. All analyses were conducted using 5–10 3422 (4·7%) 1752 (4·8%) 1670 (4·6%) the SAS programme (version 9.4). Unknown or 2408 (3·3%) 1165 (3·2%) 1243 (3·5%) data missing Role of the funding source (Table 1 continues in next column) The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. Time-to-event curves were constructed using the Kaplan–Meier method and compared using Cox Results regression analysis considering correlation due to Between Aug 2, 2007, and March 31, 2011, cluster randomisation using robust variance. An interim 76 196 asymptomatic women aged between 40 and 49 years analysis was conducted based on the pre-specified were initially enrolled, with 72 998 randomly assigned to Lan–DeMets α-spending approach (when approxim ately either the intervention or control groups. 281 participants 56·7% of the total expected events had occurred). The were excluded after randomisation due to withdrawal of 788 Articles consent or ineligibility identified post-randomisation , and intervention group and 280 in the control 55 withdrew or were excluded after screening. Therefore, group—21 deaths in each group were due to breast 72 661participants were included in the analysis (n=36 723 cancer. in the intervention group [26 316 individually randomised, There was a statistically significant dierence between 10 407 cluster randomi sed]) and n=35 938 in the control the groups for the cumulative incidence of advanced group [26 240 individually randomised, 9698 cluster cancer (HR 0·83, 95·6% CI 0·70–0·98; p=0·026). The randomised]; figure 1, table 1). The median follow-up time two-sided p value (p=0·026) corresponds to a one-sided for this follow-up analysis was 11·4 years (range 0·0–16·1; p=0·013, which is below the adjusted one-sided α=0·022 IQR 9·3–12·9) in the intervention group and 11·3 years according to the Lan–DeMets O’Brien–Fleming spending (0·0–16·1; 8·9–12·9) in the control group. function, meeting the statistical threshold for Between the first screening visit and data cuto, there significance. The result by the competing risk analysis were 234 (26·2%) advanced cancers among the was not substantially dierent. Because the Kaplan–Meier 894 participants with breast cancer in the intervention curves suggested a violation of the proportional hazards group and 277 (32·9%) advanced cancers among the assumption, we calculated HRs by follow-up time 843 patients with cancer in the control group (table 2). In interval—the results were 0·91 (0·69–1·21) for the intervention group, 646 (72%) of 894 patients were 0–48 months, 0·67 (0·50–0·91) for 49–96 months, 0·94 diagnosed with early-stage cancer (stage 0 or 1) compared (0·64–1·36) for 97–144 months, and 0·80 (0·44–1·72) for with 551 (65%) of 843 patients in the control group. 145 months or longer. Given the possible violation When analysed by screening period, the intervention of the proportional hazards assumption suggested by group consistently showed higher detection rates of these time-interval results, we additionally calculated early-stage cancer than the control group across both the restricted mean survival time (RMST) as a model- screening rounds (first round: 151 [76%] of 198 identified free summary measure on the time scale. The RMST cancers vs 78 [63%] of 124; second round: 84 [80%] of 105 was 179·08 months (95% CI 178·96–179·20) in the vs 57 [71%] of 80). Of cancers detected in participants who intervention group and 178·90 months (178·77–179·03) did not undergo the second screening or those who were in the control group. The dierence was 0·18 months diagnosed after the second screening, 392 (71%) (p=0·045), corresponding to approximately 5·6 days per of 555 cancers identified in the intervention group and participant, or about 15·3 event-free person-years gained 384 (65%) of 590 cancers identified in the control group per 1000 participants over 15 years. were early stage. However, among interval cancers (ie, This small RMST dierence is consistent with the those diagnosed between the first and second screening small absolute dierence in the cumulative incidence of visit), early-stage disease was observed in advanced breast cancer at 180 months (1·08% 19 (53%) of 36 cases in the intervention group and [95% CI 0·89–1·30] in the intervention group and 1·14% 32 (65%) of 49 cases in the control group. During the [0·99–1·30] in the control group). A sensitivity analysis follow-up period, there were 297 deaths in the comparing the eect of screening in the individually Intervention group (n=894) Control group (n=843) 1st round 1st round 2nd round Other period Total (n=894) 1st round 1st round 2nd round Other period Total (n=843) (n=198) interval (n=105) (n=555) (n=124) interval (n=80) (n=590) cancers cancers (n=36) (n=49) Early cancers 151 (76·3%) 19 (52·7%) 84 (80·0%) 392 (70·6%) 646 (72·3%) 78 (62·9%) 32 (65·4%) 57 (71·3%) 384 (65·1%) 551 (65·4%) Stage 0 57 (28·8%) 7 (19·4%) 30 (28·6%) 116 (20·9%) 210 (23·5%) 31 (25·0%) 9 (18·4%) 18 (22·5%) 119 (20·2%) 177 (21·0%) Stage 1 94 (47·5%) 12 (33·3%) 54 (51·4%) 276 (49·7%) 436 (48·8%) 47 (37·9%) 23 (47·0%) 39 (48·8%) 265 (44·9%) 374 (44·4%) Advanced cancers 44 (22·2%) 17 (47·3%) 21 (20·0%) 152 (27·4%) 234 (26·1%) 45 (36·3%) 16 (32·6%) 23 (28·7%) 193 (32·7%) 277 (32·8%) Stage 2 33 (16·7%) 14 (38·9%) 18 (17·1%) 121 (21·8%) 186 (20·8%) 37 (29·8%) 13 (26·5%) 20 (25·0%) 163 (27·6%) 233 (27·6%) Stage 3 10 (5·0%) 2 (5·6%) 2 (1·9%) 25 (4·5%) 39 (4·3%) 6 (4·9%) 3 (6·1%) 3 (3·7%) 26 (4·4%) 38 (4·5%) Stage 4 1 (0·5%) 1 (2·8%) 1 (1·0%) 6 (1·1%) 9 (1·0%) 2 (1·6%) 0 0 4 (0·7%) 6 (0·7%) Unknown 3 (1·5%) 0 0 11 (2·0%) 14 (1·6%) 1 (0·8%) 1 (2·0%) 0 13 (2·2%) 15 (1·8%) Number of deaths* ·· ·· ·· ·· 297 (100%) ·· ·· ·· ·· 280 (100%) Breast cancer ·· ·· ·· ·· 21 (7·1%) ·· ·· ·· ·· 21 (7·5%) Any cause ·· ·· ·· ·· 223 (75·1%) ·· ·· ·· ·· 207 (73·9%) Unknown ·· ·· ·· ·· 53 (17·8%) ·· ·· ·· ·· 52 (18·6%) *Percentages in the stage-distribution section are calculated using the total number of breast cancer cases in each period as the denominator. Percentages in the deaths section are calculated using total deaths in each group (297 in the intervention group and 280 in the control group) as the denominator. Table 2: Stage distribution of breast cancer and number of deaths stratified by cause Articles A 1·2 1·0 0·8 0·6 0·4 0·2 0 24 48 72 96 120 144 168 192 Number at risk (censored) Study group 36 723 337 94 32 266 31 171 290 36 25 550 14 838 3062 1 (0) (2874) (1484) (1054) (2102) (3455) (10 685) (11 763) (3059) Control group 35 938 32 791 31 155 30 046 276 90 24 345 14 206 2724 1 (0) (3082) (1596) (1048) (2311) (3314) (10 111) (11 466) (2723) Figure 2: Kaplan–Meier curve for cumulative incidence of breast cancer (A) Cumulative incidence of advanced breast cancer in the intervention group (red) and control group (blue). The graph shows the percentage of patients who were diagnosed with advanced breast cancer or who died due to breast cancer of unknown stage (y-axis) over a follow-up period of 192 months (x-axis). (B) Cumulative incidence of overall breast cancer in the intervention group (red) and control group (blue). The graph shows the percentage of patients who were diagnosed with breast cancer (y-axis) or who died over a follow-up period of 192 months (x-axis). Arrows indicate the timing of the first and second screenings. Numbers below the graph indicate the number of participants at risk and number of events at different timepoints for both groups. and cluster-randomised groups showed evidence of group. Although a slight separation between study arms quantitative interaction, with HRs of 0·77 was observed in the cluster-randomised group, no clear (95·6% CI 0·62–0·95) for the individually randomised dierence was evident (appendix pp 5–6). The within- group and 0·95 (0·71–1·27) for the cluster-randomised cluster correlation was negligible (intra-class correlation 790 )%( ecnedicni recnac tsaerb decnavda evitalumuC B 4·0 3·5 3·0 2·5 2·0 1·5 1·0 0·5 0 24 48 72 96 120 144 168 192 Number at risk (censored) Study group 36 723 33 776 32 210 31 086 289 23 25 382 14 682 2994 1 (0) (2724) (1402) (987) (2030) (3424) (10 615) (11 652) (2986) Control group 35 938 32 777 31 101 29 956 275 79 24 180 14 061 2674 1 (0) (2985) (1540) (997) (2232) (3284) (10 037) (11 342) (2672) )%( ecnedicni recnac tsaerb llarevo evitalumuC First Second screening screening Number 10-year cumulative of events incidence rate (95% CI) Study group 246 0·64 (0·56–0·73) Control group 286 0·79 (0·69–0·89) Hazard ratio 0·83 (95·6% CI 0·70 –0·98); p=0·026 Number 5-year cumulative of events incidence rate (95% CI) Study group 902 1·33 (1·21–1·46) Control group 848 1·20 (1·09–1·33) Hazard ratio 1·02 (95% CI 0·93–1·13); p=0·65 Time since randomisation (months) Articles coecient <0·0001). These findings indicate that the benefit of screening. Although we did not evaluate cost- estimated eect diered by randomisation design, and eectiveness, the RMST dierence should be interpreted this interaction should be considered when interpreting alongside the reduction in advanced breast cancer the overall results. incidence and other clinical outcomes rather than as a In the control group, 20 additional advanced breast standalone metric. cancer cases were newly detected after the second There was no dierence in the number of advanced screening between 49 and 72 months, and 12 more cases cancers between the two groups until 48 months after the were identified between 73 and 96 months. However, the second-round screening. However, there was a dierence incidence of advanced cancer beyond 97 months was between the two groups from 49 to 72 months as 20 more comparable between the two groups. The 10-year advanced cancers were found in the control group, and cumulative incidence rate of advanced breast cancer, from 73 to 96 months, as 12 more advanced cancers were defined as stage 2 or higher breast cancer or death from found in the control group. The decrease of advanced breast cancer, was 0·64% (95% CI 0·56–0·73; 246 events cancer in the intervention group from 49 to 96 months in 36 723 participants) in the intervention group might indicate that the adjunctive ultrasonography and 0·79% (0·69–0·89; 286 events in 35 938 participants) screenings with 2-year interval detected the cases that in the control group (appendix p 4). These event numbers would become advanced cancer in the future, as both include 234 (95%) cases of stage 2 or higher breast cancer populations were initially expected to be similar in terms in the intervention group and 277 (97%) in the control of advanced cancer prevalence by randomisation. Among group, plus an additional 12 (5%) and nine people (3%) the breast cancers detected at early stage during the who died of breast cancer with unknown stage. In follow-up, they might have potential to progress to more contrast, there was no dierence between the groups for advanced stages over time. Early detection of these the cumulative incidence of overall breast cancer cancers through ultrasonography screening might have (HR 1·02 [95% CI 0·93–1·13]). The 5-year cumulative contributed to the reduction in advanced cancer incidence of overall breast cancer was 1·33% (902 events incidence. No dierence in the incidence of advanced in 36 723 individuals [95% CI 1·21–1·46]) in the cancer was observed between the two groups from the intervention group and 1·20% (848 events in 35 938 first screening to 48 months. However, the dierence [1·09–1·33]) in the control group (figure 2B). The total widened between 48 and 96 months, and it remained at number of events exceeds the number of participants that level thereafter. These results suggest that adjunctive with breast cancer because some breast cancer deaths ultrasonography may be eective between 4 and 8 years were counted as events even without a corresponding after the initial screening. registration of advanced breast cancer. The individually and cluster-randomised groups showed dierent point estimates of eect, indicating a Discussion quantitative interaction. Because the intra-class coecient Mammography has limited accuracy in breast cancer was essentially zero, within-cluster correlation was screening especially for women with dense breast negligible and is unlikely to account for this dierence. tissue.12,13 Adjunctive ultrasonography, when used in As shown in the appendix (pp 2–3), no substantial conjunction with mammography screening, increases imbalance was observed in measured baseline the sensitivity and detection rate of breast cancers and characteristics between the study arms under either reduces interval cancers in asymptomatic women with randomisation method. Nevertheless, we cannot exclude dense and non-dense breasts.6 In this follow-up study we the possibility that unmeasured factors diered in the observed that adjunctive ultrasonography screening cluster-randomised group. In addition, although we do significantly reduces the incidence of advanced breast not have sucient data to confirm this, it is possible that cancer in asymptomatic women aged 40 to 49 years after the second screening, more women in the compared with mammography alone (HR 0·83 mammography-only group at cluster-randomised sites [95·6% CI 0·70–0·98]; p=0·026). The RMST analysis underwent opportunistic ultrasound examinations, (between-group dierence 0·18 months [approximately which could have diluted the apparent eect of the 5·6 days] per participant [p=0·045]) provided an intervention. Chance variation is also a possible additional, model-free summary on the time scale and explanation. Taken together, these findings suggest that complemented the interpretation of HRs and absolute although a slight dierence was observed in the cluster- risk dierences. This modest per-person estimate is randomised group, no clear dierence was evident. The expected because advanced breast cancer was a rare event overall findings from the combined analysis should in this screening trial and the per-person RMST is therefore be interpreted with consideration of the averaged over all participants, most of whom remained dierences in trial design and the potential for event-free. Expressed on a population scale, this contamination in the cluster-randomised setting. corresponds to 15·3 event-free person-years gained per Mortality reduction is a crucial endpoint indicator for 1000 participants over 15 years, providing a more intuitive cancer-screening evaluation in randomised controlled representation of the cumulative programme-level trials. However, breast cancer generally has a better Articles prognosis compared to other cancers, and tracking death round onward, in accordance with standard public events is time-consuming.14 There is ongoing debate as to screening protocols in Japan. However, due to the whether a reduction in advanced breast cancer incidence absence of detailed information on the screening can be considered a valid surrogate marker for reduced frequency and the potential use of additional breast cancer mortality. Some studies, including a meta- ultrasonography examinations in both groups, the analysis,15,16 have reported a strong correlation between possibility of contamination remains a concern. Third, reductions in advanced cancers and subsequent declines the higher incidence rate of early-stage breast cancer in in mortality. In these studies, advanced cancer was the intervention group cannot rule out the possibility of defined as stage 2 or higher. Feng and colleagues, who overdiagnosis. Although adjunctive ultrasonography defined stage 3 and 4 as advanced cancer, showed a strong could contribute to early detection, some of the correlation between stage and mortality in ovarian and additionally detected cancers might have been clinically lung cancers, and a moderate correlation in breast cancer harmless. To evaluate the extent of overdiagnosis and its as compared to colorectal and prostate cancers.17 clinical significance, further long-term follow-up Furthermore, others have cautioned against relying too regarding breast cancer-specific mortality is needed. heavily on surrogate endpoints without validation based In conclusion, adjunctive ultrasonography reduced the on randomised controlled trials,18,19 citing concerns about incidence of advanced breast cancer between overdiagnosis and lead-time bias. In this context, the four years and eight years after the first screening, but long-term follow-up of J-START oers valuable real-world showed no superiority after 8 years. This finding suggests evidence to examine whether the observed stage shift that the combined screening modality is highly eective translates into survival benefits. Further evaluation of for the early detection of breast cancer in women aged mortality outcomes is necessary. 40–49 years, in an Asian population where the prevalence The main strength of this study is that many participants of dense breast tissue is high. To our knowledge, this enrolled had long-term follow-up. The 72 661 women study is the first randomised controlled trial showing were followed up for more than 15 years, with only lower incidence of advanced breast cancer with adjunctive 2671 (3·7%) participants lost to follow-up, which allowed ultrasonography. Further follow-up is needed to us to investigate the ecacy of adding the adjunctive determine whether adjunctive ultrasonography ultrasonography over time. As a result, the dierence contributes to a reduction in breast cancer mortality. between the two groups was confirmed with high Contributors statistical power. J-START focused on women aged NO and SY designed this study. NO, AS, and TI undertook the literature 40–49 years as Asian women generally have high density search. NO and TI contributed to funding acquisition. NO, AS, YS-N, and TI collected the data with collaboration sites. SY, SK, and TY breast tissue and age-specific incidence of breast cancer planned the statistical analysis. SK, SY, NH-S, and YS-N managed the peaks during this period.20,21 Another strength of this trial data. SK, SY, and NH-S accessed and verified the data. TY did the is the standardisation of ultrasonography screening statistical analysis. NH-S, NO, AS, TI, SK, and SY aided in interpreting across sites. In collaboration with the relevant professional the results and worked on the manuscript. All authors had full access to all the included data in the study and had final responsibility for the societies, we developed and implemented guidelines and decision to submit for publication. training programmes covering equipment requirements, Declaration of interests image acquisition, interpretation and assessment We declare no competing interests. categories, and quality management across trial sites.22 Data sharing We have established training programmes for doctors Due to the nature of this research, participants of this study did not agree and technologists to ensure accurate and high-quality for their data to be shared publicly, so supporting data are not available. screenings.12,13 We anticipate this trial contributing to Acknowledgments early detection of breast cancer in women with dense We thank the trial participants and the research associates at all research breast tissue, improving their prognosis. sites for their contributions, and Junki Mizusawa for his helpful The limitations of this study are as follows. First, the comments on the statistical aspects of this paper. We are also grateful to Tomotaka Sobue and Hiroshi Saito for their support in conducting this biological characteristics of breast cancer were not study. Finally, we express our deepest gratitude to the late Yasuo Ohashi, evaluated in this study. Currently, survival risk who was involved from the beginning in developing the statistical assessment should be tailored to each breast cancer analysis plan and made invaluable contributions as the person subtype, given the availability of highly eective, subtype- responsible for data management throughout this study. specific drug therapies. The subtype-based classification References 1 Kawai M, Ohtani S, Iwasaki M, et al. The Japanese Breast Cancer was established in the 2011 St Gallen Consensus,23 Society clinical practice guidelines for epidemiology and whereas J-START was developed before the extensive prevention of breast cancer, 2022 edition. Breast Cancer 2024; adoption of this classification system in clinical practice. 31: 166–78. Consequently, pathological records do not have consistent 2 Nyström L, Andersson I, Bjurstam N, Frisell J, Nordenskjöld B, Rutqvist LE. Long-term eects of mammography screening: updated molecular subtype data. Second, J-START conducted overview of the Swedish randomised trials. Lancet 2002; 359: 909–19. screenings every 2 years up to the second round. 3 Tabar L, Duy SW, Yen M-F, et al. All-cause mortality among breast Consequently, it is presumed that most participants cancer patients in a screening trial: support for breast cancer mortality as an end point. J Med Screen 2002; 9: 159–62. received mammography-only screenings from the third 792 Articles 4 Frisell J, Lidbrink E, Hellström L, Rutqvist L-E. Followup after 14 Miller KD, Nogueira L, Devasia T, et al. Cancer treatment and 11 years—update of mortality results in the Stockholm survivorship statistics, 2022. CA Cancer J Clin 2022; 72: 409–36. mammographic screening trial. Breast Cancer Res Treat 1997; 15 Autier P, Héry C, Haukka J, Boniol M, Byrnes G. Advanced breast 45: 263–70. cancer and breast cancer mortality in randomised controlled trials 5 Bjurstam N, Björneld L, Warwick J, et al. The Gothenburg breast on mammography screening. J Clin Oncol 2009; 27: 5919–23. screening trial. Cancer 2003; 97: 2387–96. 16 Kerlikowske K, Bissell MCS, Sprague BL, et al. Advanced breast 6 Ohuchi N, Suzuki A, Sobue T, et al. Sensitivity and specificity of cancer definitions by staging system examined in the breast cancer mammography and adjunctive ultrasonography to screen for surveillance consortium. J Natl Cancer Inst 2021; 113: 909–16. breast cancer in the Japan Strategic Anti-Cancer Randomised Trial 17 Feng X, Zahed H, Onwuka J, et al. Cancer stage compared with (J-START): a randomised controlled trial. Lancet 2016; 387: 341–48. mortality as end points in randomised clinical trials of cancer 7 Harada-Shoji N, Suzuki A, Ishida T, et al. Evaluation of adjunctive screening. JAMA 2024; 331: 1910–17. ultrasonography for breast cancer detection among women aged 18 Webb AB, Berg CD, Castle PE, et al. Considerations for using 40–49 years with varying breast density undergoing screening potential surrogate endpoints in cancer screening trials. mammography. JAMA Netw Open 2021; 4: e2121505. Lancet Oncol 2024; 25: e183–92. 8 Kolb TM, Lichy J, Newhouse JH. Comparison of the performance of 19 Kumar SH. Alternative endpoints to mortality in cancer screening screening mammography, physical examination, and breast US and trials. Mol Oncol 2024; 18: 1817–20. evaluation of factors that influence them: an analysis 20 Leong SPL, Shen Z-Z, Liu T-J, et al. Is breast cancer the same of 27,825 patient evaluations. Radiology 2002; 225: 165–75. disease in Asian and Western countries? World J Surg 2010; 9 Scheel JR, Lee JM, Sprague BL, Lee CI, Lehman CD. Screening 34: 2308–24. ultrasound as an adjunct to mammography in women with 21 Suzuki A, Kuriyama S, Kawai M, et al. Age-specific interval breast mammographically dense breasts. Am J Obstet Gynecol 2015; 212: 9–17. cancers in Japan: estimation of the proper sensitivity of screening 10 Berg WA, Zhang Z, Lehrer D, et al. Detection of breast cancer with using a population-based cancer registry. Cancer Sci 2008; addition of annual screening ultrasound or a single screening MRI 99: 2264–67. to mammography in women with elevated breast cancer risk. 22 Tohno E, Ueno E, Watanabe H. Ultrasound screening of breast JAMA 2012; 307: 1394–404. cancer. Breast Cancer 2009; 16: 18–22. 11 Glechner A, Wagner G, Mitus JW, et al. Mammography in 23 Goldhirsch A, Wood WC, Coates AS, Gelber RD, Thürlimann B, combination with breast ultrasonography versus mammography for Senn HJ. Strategies for subtypes—dealing with the diversity of breast cancer screening in women at average risk. breast cancer: highlights of the St Gallen International Expert Cochrane Database Syst Rev 2023; 3: CD009632. Consensus on the primary therapy of early breast cancer 2011. 12 Tohno E, Takahashi H, Tamada T, Fujimoto Y, Yasuda H, Ohuchi N. Ann Oncol 2011; 22: 1736–47. Educational program and testing using images for the standardisation of breast cancer screening by ultrasonography. Breast Cancer 2012; 19: 138–46. 13 Ishida T, Suzuki A, Kawai M, et al. A randomised controlled trial to verify the ecacy of the use of ultrasonography in breast cancer screening aged 40-49 (J-START): 76 196 women registered. Jpn J Clin Oncol 2014; 44: 134–40. --- [PDF原文](https://sci-net.xyz/storage/7932541/fa4ac13132c37ef3e88f285d3507f9b6c108c6a142ebe74ef5da0b5311af6c93/Cumulative-incidence-of-advanced-breast-cancer-in-women-aged-40-49-years-in-the-Japan-Strategic.pdf) DOI: 10.1016/S0140-6736(25)02319-0