- Email: [email protected]
Early detection of second breast cancers improves prognosis in breast cancer survivors
original article
Annals of Oncology 20: 1505–1510, 2009 doi:10.1093/annonc/mdp037 Published online 17 March 2009
Early detection of second breast cancers improves prognosis in breast cancer survivors N. Houssami1,2* , S. Ciatto1 , F. Martinelli1, R. Bonardi1 & S. W. Duffy3 1
Istituto per lo Studio e la Prevenzione Oncologica, Istituto Scientifico della Regione Toscana, Florence, Italy; 2Screening and Test Evaluation Program, School of Public Health, University of Sydney, Sydney, Australia and 3Cancer Research UK, Centre for Epidemiology, Mathematics and Statistics, Wolfson Institute of Preventive Medicine, London, UK
Received 26 November 2008; accepted 27 January 2009
introduction The population of women with a personal history of breast cancer is progressively expanding; it is estimated that breast cancer is the most prevalent cancer in the world [1]. Gains in life expectancy, improved cancer therapy, and population breast screening contribute to the increasing number of women who have experienced, and survived, breast cancer. Breast cancer survivors are at increased future risk of developing locoregional recurrences and contralateral breast cancers (CBCs) [2–4]. Surveillance aimed at early detection of distant metastases does not improve survival [5, 6]; however, the effect on survival of early detection of breast recurrences is debated [7–10]. *Correspondence to: Assoc. Prof. N. Houssami, Screening and Test Evaluation Program, School of Public Health, Edward Ford Building (A27), University of Sydney, Sydney NSW 2006, Australia. Tel: +61-02-419-273510; Fax: +61-02-93517420; E-mail: [email protected]
These authors have equally contributed to the development and reporting of this work.
Studies that have reported on the prognosis of mammographydetected ipsilateral relapse or contralateral cancer [2, 11, 12] have not shown good or consistent evidence that this translates into a survival benefit. A review by De Bock et al. [7] highlighted the discordance among studies on the basic ‘assumption that the detection and treatment of locoregional recurrences before symptomatic presentation has any beneficial effect on overall survival’. Grunfeld et al. [9] reported that many of the studies examining detection of second breast cancers did not consider long-term outcomes. Surveillance of women with a history of breast cancer may potentially have clinical and psychological benefits; however, it has not been possible to quantify the impact of early detection (if any) without the biases inherent in nonrandomised studies. The main biases are those associated with cancer screening, namely lead-time and length-time bias. Since randomised trials of the impact of early detection in breast cancer survivors may not be feasible, alternate approaches are warranted to estimate
ª The Author 2009. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For permissions, please email: [email protected]
original article
cancer is unknown. We examined the prognostic effect of detection of ipsilateral breast relapse (IBR) or contralateral breast cancer (CBC) in the asymptomatic relative to symptomatic phase. Patients and methods: Subjects were women with histology-verified second (invasive or in situ) breast cancer (N = 1044) in a breast centre in Florence (1980–2005). Symptom status, test, tumour stage, and outcomes data were obtained from clinical records and linkage with mortality registry. Disease-specific survival was measured from first cancer diagnosis to avoid lead-time bias. Sensitivity analysis was used to allow for length-time bias. Results: Second cancers (IBR = 455; CBC = 589; median age 60 years) were diagnosed in 699 asymptomatic and 345 symptomatic women (67% versus 3%, P < 0.0001). Mammography was more sensitive than clinical examination (86% versus 57%, P < 0.0001); however, 13.8% of cases were only identified clinically. Asymptomatic cancers were smaller than symptomatic for both IBR (P < 0.001) and CBC (P < 0.001). Early-stage tumours were more frequent in asymptomatic (58.1%) than symptomatic (22.6%) women (P < 0.0001). Fewer women with asymptomatic than symptomatic CBC had node metastases (P = 0.0001). Hazard ratio (HR) for asymptomatic (relative to symptomatic) detection was 0.51 (0.32–0.80) for IBR, 0.53 (0.36–0.78) for CBC, and 0.53 (0.40–0.72) in all subjects (P < 0.0001). Length bias-adjusted HRs ranged from 0.53 to 0.73. Conclusion: Detection of second breast cancers in the asymptomatic phase leads to detection of early-stage cancer and improves relative survival by between 27% and 47%. Key words: contralateral cancer, disease-specific survival, early detection, ipsilateral breast relapse, lead-time bias, second breast cancer
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Background: The impact of early detection of second breast cancers in women who have survived a primary breast
original article the benefit associated with early detection of second breast cancer events and to ensure that estimates are free of bias. We present a retrospective study of the impact of early detection—detection in the asymptomatic relative to the symptomatic phase—of second breast cancer events in women with a past history of breast cancer. Our objectives are to estimate the effect of early detection of ipsilateral breast relapse (IBR) or CBC and to report on the long-term survival benefit attributable to early detection of second cancer events allowing for lead-time and length-time bias.
patients and methods
analysis To establish whether detection in the asymptomatic relative to symptomatic phase was associated with earlier detection, we compared second cancers by symptom status for stage at diagnosis. When comparing second cancers according to symptom status, we used the chi-square test for frequencies and McNemar’s chi-square test for paired data. Detection patterns for the time frame of the study compared proportions detected by mammography alone or by BCE alone over four time periods (1980–1989, 1990–1994, 1995–1999, and 2000–2005).
1506 | Houssami et al.
Breast cancer deaths were identified. For subjects alive at the end of follow-up or for those who had died from a different cause, observations were censored on the date of their last observation or date of death. Vital status was assessed directly for subjects regularly followed up at CSPO or according to the regional mortality registry. Linkage with the mortality registry for specific cause of death was updated in December 2005. Our primary outcome measure for quantifying the effect of earlydetected (asymptomatic relative to symptomatic) cancer was diseasespecific survival measured from the ‘first’ cancer diagnosis to eliminate lead-time bias in detection of the ‘second’ cancer. Kaplan–Meier survival curves were generated to compare differences in survival probabilities over time according to whether the second cancer was detected in asymptomatic or symptomatic women. To estimate the effect of early detection, we used Cox regression analysis initially including the following covariates in the model: (i) age at first and second cancer diagnosis (<50, 50–69, and ‡70 years), (ii) stage (pT and pN category) of the first breast cancer, (iii) symptom status at diagnosis of the second cancer, and (iv) time period in which the second cancer was diagnosed (1980–1997 and 1998–2007). Regression models were examined for IBR and CBC separately to provide estimates for each group (available from authors). A global model was then developed for all second breast cancer events retaining covariates that are significant and/or clinically relevant. These models were used to calculate the hazard ratio (HR) and 95% confidence intervals (CIs) associated with the detection status of second breast cancers while allowing for variables influencing survival in our data. Models were tested for proportionality assumption. To allow for potential length bias, we applied a sensitivity analysis [14] to estimates of survival differences between asymptomatic and symptomatic women. The sensitivity analysis hypothesises two latent tumour populations. One of these, the length bias group, has a greater probability of being screen detected in the asymptomatic phase and a proportionally smaller prior probability of proving fatal [14]. This group will tend to be over-represented in the screen-detected cancers, thus exaggerating the survival advantage of detection in the asymptomatic phase. Our sensitivity analyses re-estimated the HRs of breast cancer death for asymptomatic versus symptomatic tumours, over a range of plausible values of the increased probability of asymptomatic detection (and reduced probability of fatality) and of the mixing fraction of the two populations. We posited a length bias group with relative size ranging from 10% of the population to 50% and with prior relative risk of death from 0.9 to 0.5 (and therefore relative risk of screen detection from 1.11 to 2.00). The methods of this sensitivity analysis have been described [14], and further details are available from the authors. All statistical analyses were carried out using the STATA software, release 8.0.
results Histology-verified second breast cancers were identified in 1044 women (from 24 278 breast cancer histology records) with a median age of 60 years [interquartile range (IQR) 51–70 years]: IBR occurred in 455 subjects and CBC in 589 subjects. Median follow-up from diagnosis of the first cancer was 13.7 years (IQR 9.0–18.1 years). Second cancers were detected in 699 asymptomatic (67.0%) relative to 345 symptomatic (33.0%) women (P < 0.0001). A summary of the characteristics of IBR and CBC in our cohort is included as Appendix 1.
Volume 20 | No. 9 | September 2009
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Eligible subjects consisted of all women diagnosed with a second metachronous breast cancer (a cancer occurring ‡6 months after first cancer) from January 1980 to December 2005, in women with a prior primary breast cancer (invasive or in situ cancer). We searched the clinical and pathology archives of the study centre Centro per lo Studio e la Prevenzione Oncologica (CSPO), identifying women with two breast cancer (surgical or needle) histology episodes separated by at least 6 months, and reviewed medical and mammography records to verify eligibility. CSPO is Florence’s main breast screening and diagnostic centre, has one of the longest established population screening programmes in Europe, and has provided a programme of periodic surveillance for women with a history of breast cancer since 1970 [12]. The following surveillance has been recommended at CSPO throughout the time frame of the study: clinical examination (6-monthly) and annual mammography for the initial 5 years in women treated with breast conservation, thereafter annual clinical examination with annual or biennial mammography; annual clinical examination and annual or biennial mammography in women treated with mastectomy. CSPO archives have ongoing data linkage to population cancer and mortality registries in the Tuscan district, and CSPO provides population-based mammography services for the region. We refer to second breast cancers when describing all subjects and define two groups eligible for inclusion: (i) ipsilateral in-breast relapse (IBR), comprising women who had a primary breast cancer treated with breast conservation and developed a second cancer in the affected breast (including those with breast relapse and concomitant axillary disease) and (ii) CBC, comprising women who had a primary breast cancer and subsequently developed a second cancer in the opposite breast. We excluded women presenting primarily with metastatic disease and those presenting with chest wall recurrences following mastectomy for the first cancer. The distinction between an ipsilateral recurrence and an ipsilateral second primary is difficult, thus we classified all ipsilateral cancers as IBR. Information on the data collected and the characteristics of this cohort of women with second breast cancers has been detailed in a recent publication [13]. Data on diagnosis of the second cancer included mammography and breast clinical examination (BCE) findings and whether the woman was asymptomatic or had presented with symptoms (including lump or focal mass, nipple discharge, skin change, and axillary mass).
Annals of Oncology
original article
Annals of Oncology
The corresponding proportion of cancers detected with BCE only did not change much over time, being 16.2%, 16.6%, 19.0%, and 12.6% (P for trend = 0.45). Data reported on long-term survival excluded 36 cases (3.4% of subjects) lost to follow-up. Figure 1 shows Kaplan– Meier (unadjusted) survival probabilities over time according to whether the second cancer was detected in asymptomatic or symptomatic women (P < 0.0001). Cox regression models (available from authors) showed that HRs for asymptomatic relative to symptomatic detection were 0.51 (95% CI 0.32–0.80) for IBR (P = 0.004) and 0.53 (95% CI 0.36–0.78) for CBC (P < 0.0001). Sensitivity analyses for length bias gave ranges of HRs from 0.51 to 0.66 for IBR and from 0.59 to 0.77 for CBC. The global regression model is shown in Table 2: the HR for asymptomatic relative to symptomatic detection is 0.53 (95% CI 0.40–0.72, P < 0.0001). Sensitivity analysis for length bias gave a range of HRs from 0.53 to 0.73.
discussion We have shown that detection of IBR or CBC in the asymptomatic phase, in women with a personal history of breast cancer, is strongly associated with early detection—that is detection of tumours at an earlier stage than those diagnosed once cancer symptoms have developed. We have also provided evidence on the impact of early detection of second breast events that factors both lead-time and length bias in estimates of effect. The risk of breast cancer death is almost halved (HR = 0.53) if the second cancer event is detected early when the woman is asymptomatic relative to symptomatic disease. Allowing for possible length bias, the estimate lies between 0.53 and 0.73. Estimates of a survival benefit of similar magnitude are evident for early detection of IBR and CBC. However, in clinical practice, a woman who has experienced a breast cancer will need to be given advice on the potential benefit of early detection that applies to either breast; therefore, our global model provides a more useful estimate, indicating that early detection of second cancers improves relative survival by between 27% and 47%.
Table 1. Cancer stage (pT and pN) of second breast cancers in asymptomatic versus symptomatic women (N = 1044) Tumour size distribution for IBR (n = 455) Asymptomatic Symptomatic Total IBR (%)
Tis 63 7 70 (15.4)
Total (%) T1a–b 119 37 156 (34.3)
T1c 80 47 127 (27.9)
T2 26 14 40 (8.8)
T3–4 12 19 31 (6.8)
Tx 14 17 31 (6.8)
Tumour size and node status distribution for CBC (n = 589) Asymptomatic Symptomatic Total CBC (%)
Tis 51 4 55 (9.3)
T1a–b 160 21 181 (30.7)
T1c 127 89 216 (36.7)
T2 31 51 82 (13.9)
314 (69.0) 141 (40.0) 455 (100.0) Total (%)
T3–4 7 16 23 (3.9)
Tx 9 23 32 (5.4)
N0 230 100 330 (56.0)
N+ 57 59 116 (19.7)
Nx 98 45 143 (24.3)
385 (65.3) 204 (34.7) 589 (100.0)
v2 for tumour size distribution for IBR (asymptomatic versus symptomatic, excludes Tx): v2(df = 4) = 34.55; P < 0.0001. v2 for tumour size distribution for CBC (asymptomatic versus symptomatic, excludes Tx): v2(df = 4) = 107.33; P < 0.0001. v2 for node status distribution for CBC (asymptomatic versus symptomatic, excludes Nx): v2(df = 1) = 15.09; P = 0.001. IBR, ipsilateral breast relapse; CBC, contralateral breast cancer; Tx, tumour size unknown; Nx, node status unknown.
Volume 20 | No. 9 | September 2009
doi:10.1093/annonc/mdp037 | 1507
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Table 1 summarises stage-related variables of all 1044 second breast cancers according to whether detection occurred in asymptomatic or symptomatic women. The distribution of prognostic variables differed significantly between asymptomatic and symptomatic cancers for both IBR and CBC occurrence (Table 1). Asymptomatic cancers were significantly smaller than symptomatic for both IBR (P < 0.001) and CBC (P < 0.001). Counting all second cancers, early-stage tumours (Tis + T1a–b) were significantly more frequent in asymptomatic (58.13%) than symptomatic (22.6%) subjects (P < 0.0001). Of those with tumour size determined, 12.7% of the asymptomatic IBR were pT2 or larger, compared with 26.6% of symptomatic IBR (P = 0.0007). For CBC, the corresponding figures were 10% and 37% (P < 0.0001). In CBC, of subjects with nodes examined, 20% with asymptomatic cancer had node metastases compared with 37% in symptomatic cancer (P = 0.0001). Data for both mammography and BCE findings were available in 882 subjects: mammography was positive in 760 and BCE was positive in 500. Mammography had a sensitivity of 86.1% which was significantly higher than the sensitivity of BCE of 56.7% (P < 0.0001); however, BCE was the only positive finding in 13.8% of subjects. In IBR, the sensitivity of mammography was 80.5% and was higher than the sensitivity of BCE of 59.5% (P < 0.0001); BCE was the only positive finding in 19.5% of IBR. In CBC, the sensitivity of mammography was 90.5% and was higher than BCE sensitivity of 54.5% (P < 0.0001); BCE was the only positive finding in 9.5% of CBC. In asymptomatic women, the proportion of cancers detected with mammography only increased significantly over time, comprising 33.3%, 44.2%, 50.5%, and 60.0% from the earliest to the most recent time period (P for trend < 0.0001). The proportion of cancers detected on BCE only decreased over time, being 15.7%, 13.4%, 11.1%, and 6.0% from the earliest to the most recent time period (P for trend = 0.002). In symptomatic women, the proportion of cancers detected with mammography only slightly changed over time but this was nonsignificant, comprising 2.3%, 3.5%, 4.7%, and 7.4% from the earliest to the most recent time period (P for trend = 0.15).
original article
Annals of Oncology
Table 2. Breast cancer-specific survival in women with second breast cancers (measured from diagnosis of the first breast cancer): global multivariate Cox proportional hazards model Variable Age at first cancer (years) <50 50–69 >69 pT first cancer T1 Tis T2–4 Tx pN first cancer pN0 pN+ pNx Time period of second cancer 1998–2005 1980–1997 Second cancer CBC IBR Symptomatic Asymptomatic
Hazard ratio
95% CI
P value
1.00 0.92 2.23
0.66–1.28 1.44–3.44
0.624 <0.0001
1.00 0.413 1.42 0.88
0.17–0.98 1.01–1.99 0.50–1.56
0.046 0.044 0.666
1.00 1.64 1.62
1.17–2.30 0.86–3.06
0.004 0.135
1.00 2.92
2.07–4.11
<0.0001
1.05–1.99
0.023
0.40–0.71
<0.0001
1.00 1.45 1.00 0.53
CI, confidence interval; CBC, contralateral breast cancer; IBR, ipsilateral breast relapse.
We have estimated the effect of early detection where a second breast cancer is detected in the asymptomatic (preclinical) phase, through a combination of mammography and BCE surveillance. A randomised controlled trial (RCT) is the best design to quantify the effect of early detection of cancer, but this may not be feasible in breast cancer survivors since they have an elevated risk of IBR or CBC. While the
1508 | Houssami et al.
Volume 20 | No. 9 | September 2009
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Figure 1. Kaplan–Meier disease-specific survival curves (1008 second breast cancer events) in asymptomatic and symptomatic women.
present study is limited by its nonrandomised design, we have used methods and a modelling approach that avoid or minimise biases associated with cancer surveillance when evaluating its impact in a nonrandomised design. We have dealt with lead-time bias (the length of time diagnosis is advanced by early detection), which would otherwise be expected as the major source of bias in surveillance for second cancer events, by defining time origin at the first cancer diagnosis. This methodology is aimed at changing the total follow-up time differentially between asymptomatic (early detected) and symptomatic cases, in order to account for lead time. For the former (the group in which, otherwise, lead time potentially has an effect), our method adds the same time minus the lead time since the lead time will be counted in the time from diagnosis of the second cancer to end point. Thus, the differential treatment of the two groups being compared by the use of time from first cancer renders them comparable in estimating the effect of early detection of the second cancer. A similar approach was applied by Ciatto et al. [12] and Schootman et al. [15], in studies of contralateral cancer, but neither study factored for length-time bias. An alternate method of dealing with lead-time bias in an observational study is to consider survival from the second cancer, then analytically adjust for lead-time bias using the methods of Duffy et al. [14]; however, this involves assumptions about the length of lead time. While one might not expect length-time bias to be a major source of bias in this setting, particularly for IBR (it is reasonable to assume that only a minority of these cancers would be the indolent cancers associated with length bias), we nevertheless allowed for this analytically. We applied a sensitivity analysis [14] to our estimates of the impact of early detection to quantify the associated uncertainty from any potential length-time effect, providing a range of estimates (27%–47%) based on our global model. Length bias-adjusted HRs encompassed a wider range of estimates for CBC, suggesting that the adjusted benefit of asymptomatic detection in CBC may be smaller than that for IBR. Follow-up of breast cancer survivors is aimed at management of a broad range of clinical and quality-of-life factors [16]. We have focused on detection of subsequent breast events since little evidence is available on its potential impact. We modelled ‘relative’ survival in asymptomatic and symptomatic detection of second breast cancer events. We have not reported ‘absolute’ survival: to do so would require a comparison of breast cancer survivors who developed second events with those who did not, which is not within the aims of our study. Future work needs to explore research methods for evaluating the impact of follow-up in this setting and to determine optimal surveillance strategies in terms of frequency and duration of follow-up. The magnitude of the estimates of effect we attribute to early detection of second breast cancer events is substantial but plausible. Mammography-screening RCTs have clearly demonstrated mortality reductions of about 25%–30% in women aged 50–69 years [17]. Estimates of service screening (in women at average risk) in organised population screening programmes have also shown improved outcomes in women participating in screening [18, 19]. It is therefore
original article
Annals of Oncology
Volume 20 | No. 9 | September 2009
appendix 1 Characteristics of first and second cancers in the Tuscan cohort of women with second breast cancers [13] Ipsilateral breast relapse (N = 455) Second breast cancer Median age (IQR) at second cancer, years Detection (%) Asymptomatic Symptomatic pT category (%) pTis pT1a–c pT2+ pTx Node status (%) Negative Positive Not examined Histology (%) DCIS Invasive ductal Invasive lobular Other special types (invasive) Other breast cancersa Surgery (%) Mastectomy WLE Data missing First breast cancer Median age (IQR) at first cancer, years pT category (%) PTis pT1a–c pT2+ PTx Node status (%) Negative Positive Not examined Histology (%) DCIS Invasive ductal Invasive lobular Other special types (invasive) Other breast cancersa
59 (50–69)
Contralateral breast cancer (N = 589) 61 (53–71)
314 (69) 141 (31)
385 (65) 204 (35)
70 283 71 31
55 397 105 32
(15) (62) (16) (7)
NA NA NA 72 238 55 22
(9) (67) (18) (5)
330 (56) 116 (20) 143 (24) (16) (52) (12) (5)
55 346 115 43
(9) (59) (20) (7)
68 (15)
30 (5)
248 (55) 183 (40) 24 (5)
233 (40) 342 (58) 14 (2)
51 (43–63)
85 257 85 28
(19) (56) (19) (6)
53 (45–62)
37 271 227 54
(6) (46) (39) (9)
283 (62) 74 (16) 98 (22)
394 (67) 156 (26) 39 (7)
78 273 58 28
34 402 98 36
(17) (60) (13) (6)
18 (4)
(6) (68) (17) (6)
19 (3)
a
Includes cases where histological type of breast malignancy was not specified and those with missing data on type of cancer histology. IQR, interquartile range; NA, not applicable; DCIS, ductal carcinoma in situ; WLE, wide local excision.
doi:10.1093/annonc/mdp037 | 1509
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
reasonable to expect that women at elevated risk due to a personal history of breast cancer will have a comparable (or greater) benefit from early detection of a second breast cancer event, in line with the magnitude of benefit we report in this study. Schootman et al. [15] reported an 81% survival benefit for early detection of contralateral cancer, a higher estimate than what we have reported—this study was primarily a comparison of early-stage and advanced-stage disease [15] and did not examine how this related to detection in the asymptomatic phase or screening. Lash et al. [20] used a case–control design and demonstrated an association between mammography surveillance and reduced conditional odds of breast cancer death in older breast cancer survivors. Our work provides additional evidence in showing the impact of asymptomatic relative to symptomatic detection, based on follow-up with mammography and clinical examination. In our global model, the strongest predictors of poor prognosis were age (‡69 years), earlier relative to more recent time period in which the second cancer was diagnosed, and symptom status. We presume that the association between earlier time period of diagnosis and poorer survival is, in part, indicative of the effect of therapy; this has been reported by others in population-based studies of both early and metastatic breast cancer [21–23]. We acknowledge that our analyses are potentially limited by lack of data on therapy. Significant, though less powerful, predictors of poor outcomes in the global model were large first cancers (pT2–4), positive nodes, and an ipsilateral breast event. Although we report that mammography has a higher relative sensitivity than BCE in surveillance for second cancers, it should be noted that a proportion of second cancers were identified only on BCE. In particular, about one-fifth of IBR cases were identified only on BCE. Detection patterns over the time frame of the study showed a progressive increase in the proportion of second cancers that are detected with mammography only, and this was associated with asymptomatic detection. These findings are in keeping with the evidence we have reported on detection of second cancers in the asymptomatic phase and associated early-stage disease. It is likely that the time-related increase in the proportion of cancers identified only with mammography (in asymptomatic women) is more a reflection of increasing uptake of mammography surveillance in this population of women and to a lesser extent a result of improvement in mammography sensitivity. Due to the paucity of evidence on the impact of breast surveillance in women with a personal history of breast cancer, current recommendations vary substantially between countries and organisations [2, 24–26]. We have provided evidence that detection of second cancers in asymptomatic relative to symptomatic women is associated with detection of early-stage tumours and confers a survival benefit. This ranges between a 27% and a 47% reduction in the hazard of breast cancer death after adjusting for length bias. Recommendations on follow-up after treatment of early breast cancer should consider our findings, which suggest that early detection of second breast cancer events improves prognosis in this ever-increasing group of women.
original article funding National Health and Medical Research Council (Program grant no. 402764 to the Screening and Test Evaluation Program) to NH.
references
1510 | Houssami et al.
13. Ciatto S, Houssami N, Martinelli F et al. Second breast cancers in a Tuscan case series: characteristics, prognosis and predictors of survival. Br J Cancer 2008; 99: 539–544. 14. Duffy SW, Nagtegaal ID, Wallis M et al. Correcting for lead time and length bias in estimating the effect of screen-detection on cancer survival. Am J Epidemiol 2008; 168: 98–104. 15. Schootman M, Fuortes L, Aft R. Prognosis of metachronous contralateral breast cancer according to stage at diagnosis: the importance of early detection. Breast Cancer Res Treat 2006; 99: 91–95. 16. Muss HB. Challenges in survivorship: appropriate standards of follow-up care. In Proceedings of the 2008 Breast Cancer Symposium September 5–7 2008, Washington DC, p. 50. Alexandria, VA: American Society of Clinical Oncology 2008. 17. International Agency for Research on Cancer (IARC). IARC Handbooks of Cancer Prevention: Breast Cancer Screening. Lyon, France: International Agency for Research on Cancer 2002. 18. Gabe R, Duffy SW. Evaluation of service screening mammography in practice: the impact on breast cancer mortality. Ann Oncol 2005; 16 (Suppl): ii153–ii162. 19. Roder DM, Houssami N, Farshid G et al. Population screening and intensity of screening are associated with reduced breast cancer mortality: evidence of efficacy of mammography screening in Australia. Breast Cancer Res Treat 2008; 108(3): 409–416. 20. Lash TL, Fox MP, Buist DSM et al. Mammography surveillance and mortality in older breast cancer survivors. J Clin Oncol 2007; 25: 3001–3006. 21. Dabakuyo TS, Bonnetain F, Roignot P et al. Population-based study of breast cancer survival in Cote d’Or (France): prognostic factors and relative survival. Ann Oncol 2008; 19: 276–283. 22. Ernst MF, van de Poll-Franse LV, Roukema JA et al. Trends in the prognosis of patients with primary metastatic breast cancer diagnosed between 1975 and 2002. Breast 2007; 16: 344–351. 23. Chen SL, Hoehne FM, Giuliano AE. The prognostic significance of micrometastases in breast cancer: a SEER population-based analysis. Ann Surg Oncol 2007; 14: 3378–3384. 24. The Association of Breast Surgery @ BASO, Royal College of Surgeons of England. Guidelines for the management of symptomatic breast disease. Eur J Surg Oncol 2005; 31 (Suppl 1): 1–21. 25. Khatcheressian JL, Wolff AC, Smith TJ et al. American Society of Clinical Oncology 2006 update of the breast cancer follow-up and management guidelines in the adjuvant setting. J Clin Oncol 2006; 24: 5091–5097. 26. Hayes DF. Clinical practice. Follow-up of patients with early breast cancer. N Engl J Med 2007; 356(24): 2505–2513.
Volume 20 | No. 9 | September 2009
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin 2005; 55: 74–108. 2. Montgomery DA, Krupa K, Jack WJL et al. Changing pattern of the detection of locoregional relapse in breast cancer: the Edinburgh experience. Br J Cancer 2007; 96: 1802–1807. 3. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005; 366: 2087–2106. 4. Kurtz JM, Amalric R, Brandone H et al. Contralateral breast cancer and other second malignancies in patients treated by breast-conserving therapy with radiation. Int J Radiat Oncol Biol Phys 1988; 15: 277–284. 5. Rosselli Del Turco M, Palli D, Cariddi A et al. Intensive diagnostic follow-up after treatment of primary breast cancer: a randomised trial—National Research Council Project on Breast Cancer follow-up. JAMA 1994; 271: 1593–1597. 6. The GIVIO Investigators. Impact of follow-up testing on survival and healthrelated quality of life in breast cancer patients: a multicenter randomized trial. JAMA 1994; 271: 1587–1592. 7. De Bock GH, Bonnema J, van Der Hage J et al. Effectiveness of routine visits and routine tests in detecting isolated locoregional recurrences after treatment for early stage invasive breast cancer: a meta-analysis and systematic review. J Clin Oncol 2004; 22: 4010–4018. 8. Te Boekhorst DS, Peer NG, Van der Sluis RF et al. Periodic follow-up after breast cancer and the effect on survival. Eur J Surg 2001; 167: 490–496. 9. Grunfeld E, Noorani H, McGahan L et al. Surveillance mammography after treatment of primary breast cancer: a systematic review. Breast 2002; 11: 228–235. 10. Jacobs HJ, van Dijck JA, de Kleijn EM et al. Routine follow-up examinations in breast cancer patients have minimal impact on life expectancy: a simulation study. Ann Oncol 2001; 12: 1107–1113. 11. Lash TL, Clough-Gorr K, Silliman RA. Reduced rates of cancer-related worries and mortality associated with guideline surveillance after breast cancer therapy. Breast Cancer Res Treat 2005; 89: 61–67. 12. Ciatto S, Miccinesi G, Zappa M. Prognostic impact of the early detection of metachronous contralateral breast cancer. Eur J Cancer 2004; 40: 1496–1501.
Annals of Oncology
Annals of Oncology 20: 1505–1510, 2009 doi:10.1093/annonc/mdp037 Published online 17 March 2009
Early detection of second breast cancers improves prognosis in breast cancer survivors N. Houssami1,2* , S. Ciatto1 , F. Martinelli1, R. Bonardi1 & S. W. Duffy3 1
Istituto per lo Studio e la Prevenzione Oncologica, Istituto Scientifico della Regione Toscana, Florence, Italy; 2Screening and Test Evaluation Program, School of Public Health, University of Sydney, Sydney, Australia and 3Cancer Research UK, Centre for Epidemiology, Mathematics and Statistics, Wolfson Institute of Preventive Medicine, London, UK
Received 26 November 2008; accepted 27 January 2009
introduction The population of women with a personal history of breast cancer is progressively expanding; it is estimated that breast cancer is the most prevalent cancer in the world [1]. Gains in life expectancy, improved cancer therapy, and population breast screening contribute to the increasing number of women who have experienced, and survived, breast cancer. Breast cancer survivors are at increased future risk of developing locoregional recurrences and contralateral breast cancers (CBCs) [2–4]. Surveillance aimed at early detection of distant metastases does not improve survival [5, 6]; however, the effect on survival of early detection of breast recurrences is debated [7–10]. *Correspondence to: Assoc. Prof. N. Houssami, Screening and Test Evaluation Program, School of Public Health, Edward Ford Building (A27), University of Sydney, Sydney NSW 2006, Australia. Tel: +61-02-419-273510; Fax: +61-02-93517420; E-mail: [email protected]
These authors have equally contributed to the development and reporting of this work.
Studies that have reported on the prognosis of mammographydetected ipsilateral relapse or contralateral cancer [2, 11, 12] have not shown good or consistent evidence that this translates into a survival benefit. A review by De Bock et al. [7] highlighted the discordance among studies on the basic ‘assumption that the detection and treatment of locoregional recurrences before symptomatic presentation has any beneficial effect on overall survival’. Grunfeld et al. [9] reported that many of the studies examining detection of second breast cancers did not consider long-term outcomes. Surveillance of women with a history of breast cancer may potentially have clinical and psychological benefits; however, it has not been possible to quantify the impact of early detection (if any) without the biases inherent in nonrandomised studies. The main biases are those associated with cancer screening, namely lead-time and length-time bias. Since randomised trials of the impact of early detection in breast cancer survivors may not be feasible, alternate approaches are warranted to estimate
ª The Author 2009. Published by Oxford University Press on behalf of the European Society for Medical Oncology. All rights reserved. For permissions, please email: [email protected]
original article
cancer is unknown. We examined the prognostic effect of detection of ipsilateral breast relapse (IBR) or contralateral breast cancer (CBC) in the asymptomatic relative to symptomatic phase. Patients and methods: Subjects were women with histology-verified second (invasive or in situ) breast cancer (N = 1044) in a breast centre in Florence (1980–2005). Symptom status, test, tumour stage, and outcomes data were obtained from clinical records and linkage with mortality registry. Disease-specific survival was measured from first cancer diagnosis to avoid lead-time bias. Sensitivity analysis was used to allow for length-time bias. Results: Second cancers (IBR = 455; CBC = 589; median age 60 years) were diagnosed in 699 asymptomatic and 345 symptomatic women (67% versus 3%, P < 0.0001). Mammography was more sensitive than clinical examination (86% versus 57%, P < 0.0001); however, 13.8% of cases were only identified clinically. Asymptomatic cancers were smaller than symptomatic for both IBR (P < 0.001) and CBC (P < 0.001). Early-stage tumours were more frequent in asymptomatic (58.1%) than symptomatic (22.6%) women (P < 0.0001). Fewer women with asymptomatic than symptomatic CBC had node metastases (P = 0.0001). Hazard ratio (HR) for asymptomatic (relative to symptomatic) detection was 0.51 (0.32–0.80) for IBR, 0.53 (0.36–0.78) for CBC, and 0.53 (0.40–0.72) in all subjects (P < 0.0001). Length bias-adjusted HRs ranged from 0.53 to 0.73. Conclusion: Detection of second breast cancers in the asymptomatic phase leads to detection of early-stage cancer and improves relative survival by between 27% and 47%. Key words: contralateral cancer, disease-specific survival, early detection, ipsilateral breast relapse, lead-time bias, second breast cancer
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Background: The impact of early detection of second breast cancers in women who have survived a primary breast
original article the benefit associated with early detection of second breast cancer events and to ensure that estimates are free of bias. We present a retrospective study of the impact of early detection—detection in the asymptomatic relative to the symptomatic phase—of second breast cancer events in women with a past history of breast cancer. Our objectives are to estimate the effect of early detection of ipsilateral breast relapse (IBR) or CBC and to report on the long-term survival benefit attributable to early detection of second cancer events allowing for lead-time and length-time bias.
patients and methods
analysis To establish whether detection in the asymptomatic relative to symptomatic phase was associated with earlier detection, we compared second cancers by symptom status for stage at diagnosis. When comparing second cancers according to symptom status, we used the chi-square test for frequencies and McNemar’s chi-square test for paired data. Detection patterns for the time frame of the study compared proportions detected by mammography alone or by BCE alone over four time periods (1980–1989, 1990–1994, 1995–1999, and 2000–2005).
1506 | Houssami et al.
Breast cancer deaths were identified. For subjects alive at the end of follow-up or for those who had died from a different cause, observations were censored on the date of their last observation or date of death. Vital status was assessed directly for subjects regularly followed up at CSPO or according to the regional mortality registry. Linkage with the mortality registry for specific cause of death was updated in December 2005. Our primary outcome measure for quantifying the effect of earlydetected (asymptomatic relative to symptomatic) cancer was diseasespecific survival measured from the ‘first’ cancer diagnosis to eliminate lead-time bias in detection of the ‘second’ cancer. Kaplan–Meier survival curves were generated to compare differences in survival probabilities over time according to whether the second cancer was detected in asymptomatic or symptomatic women. To estimate the effect of early detection, we used Cox regression analysis initially including the following covariates in the model: (i) age at first and second cancer diagnosis (<50, 50–69, and ‡70 years), (ii) stage (pT and pN category) of the first breast cancer, (iii) symptom status at diagnosis of the second cancer, and (iv) time period in which the second cancer was diagnosed (1980–1997 and 1998–2007). Regression models were examined for IBR and CBC separately to provide estimates for each group (available from authors). A global model was then developed for all second breast cancer events retaining covariates that are significant and/or clinically relevant. These models were used to calculate the hazard ratio (HR) and 95% confidence intervals (CIs) associated with the detection status of second breast cancers while allowing for variables influencing survival in our data. Models were tested for proportionality assumption. To allow for potential length bias, we applied a sensitivity analysis [14] to estimates of survival differences between asymptomatic and symptomatic women. The sensitivity analysis hypothesises two latent tumour populations. One of these, the length bias group, has a greater probability of being screen detected in the asymptomatic phase and a proportionally smaller prior probability of proving fatal [14]. This group will tend to be over-represented in the screen-detected cancers, thus exaggerating the survival advantage of detection in the asymptomatic phase. Our sensitivity analyses re-estimated the HRs of breast cancer death for asymptomatic versus symptomatic tumours, over a range of plausible values of the increased probability of asymptomatic detection (and reduced probability of fatality) and of the mixing fraction of the two populations. We posited a length bias group with relative size ranging from 10% of the population to 50% and with prior relative risk of death from 0.9 to 0.5 (and therefore relative risk of screen detection from 1.11 to 2.00). The methods of this sensitivity analysis have been described [14], and further details are available from the authors. All statistical analyses were carried out using the STATA software, release 8.0.
results Histology-verified second breast cancers were identified in 1044 women (from 24 278 breast cancer histology records) with a median age of 60 years [interquartile range (IQR) 51–70 years]: IBR occurred in 455 subjects and CBC in 589 subjects. Median follow-up from diagnosis of the first cancer was 13.7 years (IQR 9.0–18.1 years). Second cancers were detected in 699 asymptomatic (67.0%) relative to 345 symptomatic (33.0%) women (P < 0.0001). A summary of the characteristics of IBR and CBC in our cohort is included as Appendix 1.
Volume 20 | No. 9 | September 2009
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Eligible subjects consisted of all women diagnosed with a second metachronous breast cancer (a cancer occurring ‡6 months after first cancer) from January 1980 to December 2005, in women with a prior primary breast cancer (invasive or in situ cancer). We searched the clinical and pathology archives of the study centre Centro per lo Studio e la Prevenzione Oncologica (CSPO), identifying women with two breast cancer (surgical or needle) histology episodes separated by at least 6 months, and reviewed medical and mammography records to verify eligibility. CSPO is Florence’s main breast screening and diagnostic centre, has one of the longest established population screening programmes in Europe, and has provided a programme of periodic surveillance for women with a history of breast cancer since 1970 [12]. The following surveillance has been recommended at CSPO throughout the time frame of the study: clinical examination (6-monthly) and annual mammography for the initial 5 years in women treated with breast conservation, thereafter annual clinical examination with annual or biennial mammography; annual clinical examination and annual or biennial mammography in women treated with mastectomy. CSPO archives have ongoing data linkage to population cancer and mortality registries in the Tuscan district, and CSPO provides population-based mammography services for the region. We refer to second breast cancers when describing all subjects and define two groups eligible for inclusion: (i) ipsilateral in-breast relapse (IBR), comprising women who had a primary breast cancer treated with breast conservation and developed a second cancer in the affected breast (including those with breast relapse and concomitant axillary disease) and (ii) CBC, comprising women who had a primary breast cancer and subsequently developed a second cancer in the opposite breast. We excluded women presenting primarily with metastatic disease and those presenting with chest wall recurrences following mastectomy for the first cancer. The distinction between an ipsilateral recurrence and an ipsilateral second primary is difficult, thus we classified all ipsilateral cancers as IBR. Information on the data collected and the characteristics of this cohort of women with second breast cancers has been detailed in a recent publication [13]. Data on diagnosis of the second cancer included mammography and breast clinical examination (BCE) findings and whether the woman was asymptomatic or had presented with symptoms (including lump or focal mass, nipple discharge, skin change, and axillary mass).
Annals of Oncology
original article
Annals of Oncology
The corresponding proportion of cancers detected with BCE only did not change much over time, being 16.2%, 16.6%, 19.0%, and 12.6% (P for trend = 0.45). Data reported on long-term survival excluded 36 cases (3.4% of subjects) lost to follow-up. Figure 1 shows Kaplan– Meier (unadjusted) survival probabilities over time according to whether the second cancer was detected in asymptomatic or symptomatic women (P < 0.0001). Cox regression models (available from authors) showed that HRs for asymptomatic relative to symptomatic detection were 0.51 (95% CI 0.32–0.80) for IBR (P = 0.004) and 0.53 (95% CI 0.36–0.78) for CBC (P < 0.0001). Sensitivity analyses for length bias gave ranges of HRs from 0.51 to 0.66 for IBR and from 0.59 to 0.77 for CBC. The global regression model is shown in Table 2: the HR for asymptomatic relative to symptomatic detection is 0.53 (95% CI 0.40–0.72, P < 0.0001). Sensitivity analysis for length bias gave a range of HRs from 0.53 to 0.73.
discussion We have shown that detection of IBR or CBC in the asymptomatic phase, in women with a personal history of breast cancer, is strongly associated with early detection—that is detection of tumours at an earlier stage than those diagnosed once cancer symptoms have developed. We have also provided evidence on the impact of early detection of second breast events that factors both lead-time and length bias in estimates of effect. The risk of breast cancer death is almost halved (HR = 0.53) if the second cancer event is detected early when the woman is asymptomatic relative to symptomatic disease. Allowing for possible length bias, the estimate lies between 0.53 and 0.73. Estimates of a survival benefit of similar magnitude are evident for early detection of IBR and CBC. However, in clinical practice, a woman who has experienced a breast cancer will need to be given advice on the potential benefit of early detection that applies to either breast; therefore, our global model provides a more useful estimate, indicating that early detection of second cancers improves relative survival by between 27% and 47%.
Table 1. Cancer stage (pT and pN) of second breast cancers in asymptomatic versus symptomatic women (N = 1044) Tumour size distribution for IBR (n = 455) Asymptomatic Symptomatic Total IBR (%)
Tis 63 7 70 (15.4)
Total (%) T1a–b 119 37 156 (34.3)
T1c 80 47 127 (27.9)
T2 26 14 40 (8.8)
T3–4 12 19 31 (6.8)
Tx 14 17 31 (6.8)
Tumour size and node status distribution for CBC (n = 589) Asymptomatic Symptomatic Total CBC (%)
Tis 51 4 55 (9.3)
T1a–b 160 21 181 (30.7)
T1c 127 89 216 (36.7)
T2 31 51 82 (13.9)
314 (69.0) 141 (40.0) 455 (100.0) Total (%)
T3–4 7 16 23 (3.9)
Tx 9 23 32 (5.4)
N0 230 100 330 (56.0)
N+ 57 59 116 (19.7)
Nx 98 45 143 (24.3)
385 (65.3) 204 (34.7) 589 (100.0)
v2 for tumour size distribution for IBR (asymptomatic versus symptomatic, excludes Tx): v2(df = 4) = 34.55; P < 0.0001. v2 for tumour size distribution for CBC (asymptomatic versus symptomatic, excludes Tx): v2(df = 4) = 107.33; P < 0.0001. v2 for node status distribution for CBC (asymptomatic versus symptomatic, excludes Nx): v2(df = 1) = 15.09; P = 0.001. IBR, ipsilateral breast relapse; CBC, contralateral breast cancer; Tx, tumour size unknown; Nx, node status unknown.
Volume 20 | No. 9 | September 2009
doi:10.1093/annonc/mdp037 | 1507
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Table 1 summarises stage-related variables of all 1044 second breast cancers according to whether detection occurred in asymptomatic or symptomatic women. The distribution of prognostic variables differed significantly between asymptomatic and symptomatic cancers for both IBR and CBC occurrence (Table 1). Asymptomatic cancers were significantly smaller than symptomatic for both IBR (P < 0.001) and CBC (P < 0.001). Counting all second cancers, early-stage tumours (Tis + T1a–b) were significantly more frequent in asymptomatic (58.13%) than symptomatic (22.6%) subjects (P < 0.0001). Of those with tumour size determined, 12.7% of the asymptomatic IBR were pT2 or larger, compared with 26.6% of symptomatic IBR (P = 0.0007). For CBC, the corresponding figures were 10% and 37% (P < 0.0001). In CBC, of subjects with nodes examined, 20% with asymptomatic cancer had node metastases compared with 37% in symptomatic cancer (P = 0.0001). Data for both mammography and BCE findings were available in 882 subjects: mammography was positive in 760 and BCE was positive in 500. Mammography had a sensitivity of 86.1% which was significantly higher than the sensitivity of BCE of 56.7% (P < 0.0001); however, BCE was the only positive finding in 13.8% of subjects. In IBR, the sensitivity of mammography was 80.5% and was higher than the sensitivity of BCE of 59.5% (P < 0.0001); BCE was the only positive finding in 19.5% of IBR. In CBC, the sensitivity of mammography was 90.5% and was higher than BCE sensitivity of 54.5% (P < 0.0001); BCE was the only positive finding in 9.5% of CBC. In asymptomatic women, the proportion of cancers detected with mammography only increased significantly over time, comprising 33.3%, 44.2%, 50.5%, and 60.0% from the earliest to the most recent time period (P for trend < 0.0001). The proportion of cancers detected on BCE only decreased over time, being 15.7%, 13.4%, 11.1%, and 6.0% from the earliest to the most recent time period (P for trend = 0.002). In symptomatic women, the proportion of cancers detected with mammography only slightly changed over time but this was nonsignificant, comprising 2.3%, 3.5%, 4.7%, and 7.4% from the earliest to the most recent time period (P for trend = 0.15).
original article
Annals of Oncology
Table 2. Breast cancer-specific survival in women with second breast cancers (measured from diagnosis of the first breast cancer): global multivariate Cox proportional hazards model Variable Age at first cancer (years) <50 50–69 >69 pT first cancer T1 Tis T2–4 Tx pN first cancer pN0 pN+ pNx Time period of second cancer 1998–2005 1980–1997 Second cancer CBC IBR Symptomatic Asymptomatic
Hazard ratio
95% CI
P value
1.00 0.92 2.23
0.66–1.28 1.44–3.44
0.624 <0.0001
1.00 0.413 1.42 0.88
0.17–0.98 1.01–1.99 0.50–1.56
0.046 0.044 0.666
1.00 1.64 1.62
1.17–2.30 0.86–3.06
0.004 0.135
1.00 2.92
2.07–4.11
<0.0001
1.05–1.99
0.023
0.40–0.71
<0.0001
1.00 1.45 1.00 0.53
CI, confidence interval; CBC, contralateral breast cancer; IBR, ipsilateral breast relapse.
We have estimated the effect of early detection where a second breast cancer is detected in the asymptomatic (preclinical) phase, through a combination of mammography and BCE surveillance. A randomised controlled trial (RCT) is the best design to quantify the effect of early detection of cancer, but this may not be feasible in breast cancer survivors since they have an elevated risk of IBR or CBC. While the
1508 | Houssami et al.
Volume 20 | No. 9 | September 2009
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
Figure 1. Kaplan–Meier disease-specific survival curves (1008 second breast cancer events) in asymptomatic and symptomatic women.
present study is limited by its nonrandomised design, we have used methods and a modelling approach that avoid or minimise biases associated with cancer surveillance when evaluating its impact in a nonrandomised design. We have dealt with lead-time bias (the length of time diagnosis is advanced by early detection), which would otherwise be expected as the major source of bias in surveillance for second cancer events, by defining time origin at the first cancer diagnosis. This methodology is aimed at changing the total follow-up time differentially between asymptomatic (early detected) and symptomatic cases, in order to account for lead time. For the former (the group in which, otherwise, lead time potentially has an effect), our method adds the same time minus the lead time since the lead time will be counted in the time from diagnosis of the second cancer to end point. Thus, the differential treatment of the two groups being compared by the use of time from first cancer renders them comparable in estimating the effect of early detection of the second cancer. A similar approach was applied by Ciatto et al. [12] and Schootman et al. [15], in studies of contralateral cancer, but neither study factored for length-time bias. An alternate method of dealing with lead-time bias in an observational study is to consider survival from the second cancer, then analytically adjust for lead-time bias using the methods of Duffy et al. [14]; however, this involves assumptions about the length of lead time. While one might not expect length-time bias to be a major source of bias in this setting, particularly for IBR (it is reasonable to assume that only a minority of these cancers would be the indolent cancers associated with length bias), we nevertheless allowed for this analytically. We applied a sensitivity analysis [14] to our estimates of the impact of early detection to quantify the associated uncertainty from any potential length-time effect, providing a range of estimates (27%–47%) based on our global model. Length bias-adjusted HRs encompassed a wider range of estimates for CBC, suggesting that the adjusted benefit of asymptomatic detection in CBC may be smaller than that for IBR. Follow-up of breast cancer survivors is aimed at management of a broad range of clinical and quality-of-life factors [16]. We have focused on detection of subsequent breast events since little evidence is available on its potential impact. We modelled ‘relative’ survival in asymptomatic and symptomatic detection of second breast cancer events. We have not reported ‘absolute’ survival: to do so would require a comparison of breast cancer survivors who developed second events with those who did not, which is not within the aims of our study. Future work needs to explore research methods for evaluating the impact of follow-up in this setting and to determine optimal surveillance strategies in terms of frequency and duration of follow-up. The magnitude of the estimates of effect we attribute to early detection of second breast cancer events is substantial but plausible. Mammography-screening RCTs have clearly demonstrated mortality reductions of about 25%–30% in women aged 50–69 years [17]. Estimates of service screening (in women at average risk) in organised population screening programmes have also shown improved outcomes in women participating in screening [18, 19]. It is therefore
original article
Annals of Oncology
Volume 20 | No. 9 | September 2009
appendix 1 Characteristics of first and second cancers in the Tuscan cohort of women with second breast cancers [13] Ipsilateral breast relapse (N = 455) Second breast cancer Median age (IQR) at second cancer, years Detection (%) Asymptomatic Symptomatic pT category (%) pTis pT1a–c pT2+ pTx Node status (%) Negative Positive Not examined Histology (%) DCIS Invasive ductal Invasive lobular Other special types (invasive) Other breast cancersa Surgery (%) Mastectomy WLE Data missing First breast cancer Median age (IQR) at first cancer, years pT category (%) PTis pT1a–c pT2+ PTx Node status (%) Negative Positive Not examined Histology (%) DCIS Invasive ductal Invasive lobular Other special types (invasive) Other breast cancersa
59 (50–69)
Contralateral breast cancer (N = 589) 61 (53–71)
314 (69) 141 (31)
385 (65) 204 (35)
70 283 71 31
55 397 105 32
(15) (62) (16) (7)
NA NA NA 72 238 55 22
(9) (67) (18) (5)
330 (56) 116 (20) 143 (24) (16) (52) (12) (5)
55 346 115 43
(9) (59) (20) (7)
68 (15)
30 (5)
248 (55) 183 (40) 24 (5)
233 (40) 342 (58) 14 (2)
51 (43–63)
85 257 85 28
(19) (56) (19) (6)
53 (45–62)
37 271 227 54
(6) (46) (39) (9)
283 (62) 74 (16) 98 (22)
394 (67) 156 (26) 39 (7)
78 273 58 28
34 402 98 36
(17) (60) (13) (6)
18 (4)
(6) (68) (17) (6)
19 (3)
a
Includes cases where histological type of breast malignancy was not specified and those with missing data on type of cancer histology. IQR, interquartile range; NA, not applicable; DCIS, ductal carcinoma in situ; WLE, wide local excision.
doi:10.1093/annonc/mdp037 | 1509
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
reasonable to expect that women at elevated risk due to a personal history of breast cancer will have a comparable (or greater) benefit from early detection of a second breast cancer event, in line with the magnitude of benefit we report in this study. Schootman et al. [15] reported an 81% survival benefit for early detection of contralateral cancer, a higher estimate than what we have reported—this study was primarily a comparison of early-stage and advanced-stage disease [15] and did not examine how this related to detection in the asymptomatic phase or screening. Lash et al. [20] used a case–control design and demonstrated an association between mammography surveillance and reduced conditional odds of breast cancer death in older breast cancer survivors. Our work provides additional evidence in showing the impact of asymptomatic relative to symptomatic detection, based on follow-up with mammography and clinical examination. In our global model, the strongest predictors of poor prognosis were age (‡69 years), earlier relative to more recent time period in which the second cancer was diagnosed, and symptom status. We presume that the association between earlier time period of diagnosis and poorer survival is, in part, indicative of the effect of therapy; this has been reported by others in population-based studies of both early and metastatic breast cancer [21–23]. We acknowledge that our analyses are potentially limited by lack of data on therapy. Significant, though less powerful, predictors of poor outcomes in the global model were large first cancers (pT2–4), positive nodes, and an ipsilateral breast event. Although we report that mammography has a higher relative sensitivity than BCE in surveillance for second cancers, it should be noted that a proportion of second cancers were identified only on BCE. In particular, about one-fifth of IBR cases were identified only on BCE. Detection patterns over the time frame of the study showed a progressive increase in the proportion of second cancers that are detected with mammography only, and this was associated with asymptomatic detection. These findings are in keeping with the evidence we have reported on detection of second cancers in the asymptomatic phase and associated early-stage disease. It is likely that the time-related increase in the proportion of cancers identified only with mammography (in asymptomatic women) is more a reflection of increasing uptake of mammography surveillance in this population of women and to a lesser extent a result of improvement in mammography sensitivity. Due to the paucity of evidence on the impact of breast surveillance in women with a personal history of breast cancer, current recommendations vary substantially between countries and organisations [2, 24–26]. We have provided evidence that detection of second cancers in asymptomatic relative to symptomatic women is associated with detection of early-stage tumours and confers a survival benefit. This ranges between a 27% and a 47% reduction in the hazard of breast cancer death after adjusting for length bias. Recommendations on follow-up after treatment of early breast cancer should consider our findings, which suggest that early detection of second breast cancer events improves prognosis in this ever-increasing group of women.
original article funding National Health and Medical Research Council (Program grant no. 402764 to the Screening and Test Evaluation Program) to NH.
references
1510 | Houssami et al.
13. Ciatto S, Houssami N, Martinelli F et al. Second breast cancers in a Tuscan case series: characteristics, prognosis and predictors of survival. Br J Cancer 2008; 99: 539–544. 14. Duffy SW, Nagtegaal ID, Wallis M et al. Correcting for lead time and length bias in estimating the effect of screen-detection on cancer survival. Am J Epidemiol 2008; 168: 98–104. 15. Schootman M, Fuortes L, Aft R. Prognosis of metachronous contralateral breast cancer according to stage at diagnosis: the importance of early detection. Breast Cancer Res Treat 2006; 99: 91–95. 16. Muss HB. Challenges in survivorship: appropriate standards of follow-up care. In Proceedings of the 2008 Breast Cancer Symposium September 5–7 2008, Washington DC, p. 50. Alexandria, VA: American Society of Clinical Oncology 2008. 17. International Agency for Research on Cancer (IARC). IARC Handbooks of Cancer Prevention: Breast Cancer Screening. Lyon, France: International Agency for Research on Cancer 2002. 18. Gabe R, Duffy SW. Evaluation of service screening mammography in practice: the impact on breast cancer mortality. Ann Oncol 2005; 16 (Suppl): ii153–ii162. 19. Roder DM, Houssami N, Farshid G et al. Population screening and intensity of screening are associated with reduced breast cancer mortality: evidence of efficacy of mammography screening in Australia. Breast Cancer Res Treat 2008; 108(3): 409–416. 20. Lash TL, Fox MP, Buist DSM et al. Mammography surveillance and mortality in older breast cancer survivors. J Clin Oncol 2007; 25: 3001–3006. 21. Dabakuyo TS, Bonnetain F, Roignot P et al. Population-based study of breast cancer survival in Cote d’Or (France): prognostic factors and relative survival. Ann Oncol 2008; 19: 276–283. 22. Ernst MF, van de Poll-Franse LV, Roukema JA et al. Trends in the prognosis of patients with primary metastatic breast cancer diagnosed between 1975 and 2002. Breast 2007; 16: 344–351. 23. Chen SL, Hoehne FM, Giuliano AE. The prognostic significance of micrometastases in breast cancer: a SEER population-based analysis. Ann Surg Oncol 2007; 14: 3378–3384. 24. The Association of Breast Surgery @ BASO, Royal College of Surgeons of England. Guidelines for the management of symptomatic breast disease. Eur J Surg Oncol 2005; 31 (Suppl 1): 1–21. 25. Khatcheressian JL, Wolff AC, Smith TJ et al. American Society of Clinical Oncology 2006 update of the breast cancer follow-up and management guidelines in the adjuvant setting. J Clin Oncol 2006; 24: 5091–5097. 26. Hayes DF. Clinical practice. Follow-up of patients with early breast cancer. N Engl J Med 2007; 356(24): 2505–2513.
Volume 20 | No. 9 | September 2009
Downloaded from http://annonc.oxfordjournals.org/ at Ondokuz MayisUniversity on May 10, 2014
1. Parkin DM, Bray F, Ferlay J, Pisani P. Global cancer statistics, 2002. CA Cancer J Clin 2005; 55: 74–108. 2. Montgomery DA, Krupa K, Jack WJL et al. Changing pattern of the detection of locoregional relapse in breast cancer: the Edinburgh experience. Br J Cancer 2007; 96: 1802–1807. 3. Early Breast Cancer Trialists’ Collaborative Group (EBCTCG). Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: an overview of the randomised trials. Lancet 2005; 366: 2087–2106. 4. Kurtz JM, Amalric R, Brandone H et al. Contralateral breast cancer and other second malignancies in patients treated by breast-conserving therapy with radiation. Int J Radiat Oncol Biol Phys 1988; 15: 277–284. 5. Rosselli Del Turco M, Palli D, Cariddi A et al. Intensive diagnostic follow-up after treatment of primary breast cancer: a randomised trial—National Research Council Project on Breast Cancer follow-up. JAMA 1994; 271: 1593–1597. 6. The GIVIO Investigators. Impact of follow-up testing on survival and healthrelated quality of life in breast cancer patients: a multicenter randomized trial. JAMA 1994; 271: 1587–1592. 7. De Bock GH, Bonnema J, van Der Hage J et al. Effectiveness of routine visits and routine tests in detecting isolated locoregional recurrences after treatment for early stage invasive breast cancer: a meta-analysis and systematic review. J Clin Oncol 2004; 22: 4010–4018. 8. Te Boekhorst DS, Peer NG, Van der Sluis RF et al. Periodic follow-up after breast cancer and the effect on survival. Eur J Surg 2001; 167: 490–496. 9. Grunfeld E, Noorani H, McGahan L et al. Surveillance mammography after treatment of primary breast cancer: a systematic review. Breast 2002; 11: 228–235. 10. Jacobs HJ, van Dijck JA, de Kleijn EM et al. Routine follow-up examinations in breast cancer patients have minimal impact on life expectancy: a simulation study. Ann Oncol 2001; 12: 1107–1113. 11. Lash TL, Clough-Gorr K, Silliman RA. Reduced rates of cancer-related worries and mortality associated with guideline surveillance after breast cancer therapy. Breast Cancer Res Treat 2005; 89: 61–67. 12. Ciatto S, Miccinesi G, Zappa M. Prognostic impact of the early detection of metachronous contralateral breast cancer. Eur J Cancer 2004; 40: 1496–1501.
Annals of Oncology