|Year : 2017 | Volume
| Issue : 3 | Page : 562-569
Tumor size distribution of invasive breast cancers and the sensitivity of screening methods in the Canadian National Breast Screening Study
Dana Shaevitch1, Sharareh Taghipour2, Anthony B Miller1, Neil Montgomery3, Bart Harvey1
1 Dalla Lana School of Public Health, University of Toronto, Toronto, Ontario M5T 3M7, Canada
2 Department of Mechanical and Industrial Engineering, Ryerson University, Toronto, Ontario M5B 2K3, Canada
3 Department of Mechanical and Industrial Engineering, University of Toronto, Toronto, Ontario M5S 3G8, Canada
|Date of Web Publication||31-Aug-2017|
Department of Mechanical and Industrial Engineering, Ryerson University, Eric Palin Hall, Room 338A, 87 Gerrard Street East, Toronto, Ontario M5B 2K3
Source of Support: None, Conflict of Interest: None
Introduction: This study set out to explore if breast cancers of different sizes are detected with varying sensitivity. In addition, we attempt to determine the effect of tumor size on screening detectability.
Subjects and Methods: Data arising from the Canadian National Breast Screening Study (CNBSS) was used to perform all analyses. The CNBSS consists of two randomized controlled trials, which includes data on detection methods, age, and allocation groups. We stratified tumor size by 5 mm; age into 40–49 and 50–59 years age groups; and cancer detection or presentation methods into mammography only, physical breast examination only, both mammography and physical breast examination, interval cancers, and incident cancers.
Results: This study revealed that a difference in tumor size exists for age (smaller tumor sizes are found in older women) and breast cancer detection or presentation modes. More specifically, breast cancers detected by mammography screening are statistically smaller than those detected by physical breast examination or those presenting as incident or interval cancers. This study also found that tumor size affects screening detectability for women in their 50's but not in their forties. That is, a statistically significant difference between mammography screening alone and physical examination alone was observed for women between the ages of 50–59 for tumor sizes up to 20 mm, including prevalent cases, and up to 15 mm when prevalent cases were excluded.
Conclusion: The results of this study suggest that smaller breast cancers are more likely to be detected among women in their 50s.
Keywords: Clinical breast examination, detectability, detection method, invasive breast cancer, mammography, mode of detection, sensitivity, tumor size
|How to cite this article:|
Shaevitch D, Taghipour S, Miller AB, Montgomery N, Harvey B. Tumor size distribution of invasive breast cancers and the sensitivity of screening methods in the Canadian National Breast Screening Study. J Can Res Ther 2017;13:562-9
|How to cite this URL:|
Shaevitch D, Taghipour S, Miller AB, Montgomery N, Harvey B. Tumor size distribution of invasive breast cancers and the sensitivity of screening methods in the Canadian National Breast Screening Study. J Can Res Ther [serial online] 2017 [cited 2020 May 27];13:562-9. Available from: http://www.cancerjournal.net/text.asp?2017/13/3/562/174539
| > Introduction|| |
The objective of breast cancer screening is to increase the possibility of curing breast cancers through treatment by identifying cancers at an earlier stage. Currently, several breast imaging methods such as mammography, ultrasound, and magnetic resonance imaging (MRI) are offered to detect breast cancer at earlier stage, in order to subsequently provide more effective treatment afterwards. X-ray mammography is the key diagnostic tool for breast cancer detection. However, sensitivity of mammography is about 70%, and poor specificity may result in further unnecessary investigation such as biopsy. Ultrasound imaging is another detection technique which is more beneficial for women with dense breasts, especially if it is combined with mammography, but it increases false positive results. MRI is the recommended detection tool for breast cancer; however, it is also associated to low specificity, positive predictive value, and high cost.
Current breast cancer screening modelling often utilize a single overall estimate of sensitivity for the screening modalities that are modelled, such as for mammography screening, and are often a function of some variable such as age., We suggest that more accurate models can be achieved if sensitivity is estimated as a function of tumor size, ideally as a continuous function. Several studies attempted to model continuous screening test sensitivity and tumor progression using tumor size and mammography screening data, yet modelling sensitivity as a continuous function has not widely performed and is scarce in the literature.,, The objective of this study is to investigate tumor sizes in the Canadian National Breast Screening Study (CNBSS) for different risk-group age subpopulations, with the larger future goal of modeling continuous or stage-based cancer development and screening sensitivity over time. We set out to explore the mean and median tumor sizes that are detected or presented through the five differing methods within the CNBSS, including through mammography or physical examination detection, both mammography and physical examination detection, or presentation as incident or interval cancers. We first aimed to determine if average breast cancer sizes differ according to how they are detected or presented by age or by CNBSS allocation group. We then set out to explore if the ability to detect breast cancers of differing sizes differs for mammography alone as compared to physical breast examination alone. That is, we sought to determine if tumor size has an effect on the sensitivity of the two screening methods used in the CNBSS.
| > Subjects and Methods|| |
Canadian National Breast Screening Study
The analyses were carried out using data collected during the CNBSS, which has been described elsewhere.,,,,, The CNBSS consists of two randomized controlled trials designed to investigate the effect of annual mammography screening over and above the annual physical breast examination of breast cancer in women aged 40–59 years. The primary objective of these randomized trials was to assess the effect of mammography on breast cancer mortality. A total of 89,948 women participated in the study; 50,430 were aged 40–49, and 39,405 were aged 50–59. These women were recruited through 15 Canadian centers between 1980 and 1985. Eligibility criteria included women with no prior history of breast cancer and no mammography screening in the previous 12 months. Women who were pregnant at the time of recruitment were excluded. Women who met the inclusion criteria were enrolled after they signed an informed consent form granting agreement for data analysis in the future and linkage to vital statistics records. At enrollment, women were asked to complete an epidemiologic questionnaire on demographics, lifestyle, and family and personal history of breast disease. All eligible women underwent a physical examination of the breast performed by a specially trained nurse or, in the province of Quebec, by a physician. Women in the 40–49 years age group were randomly assigned to either receive a mammogram and physical breast examination (intervention group) annually for 4 or 5 years or to have only a single physical breast examination at enrollment (control group). Women in the 50–59 years age group were randomly assigned to receive both annual mammograms and physical breast examinations (intervention group), or physical breast examinations alone for 4 or 5 years.
From a total of 89,948 women who agreed to participate, 113 women were excluded for various reasons described elsewhere. Therefore, a total of 89,835 women were randomized in the CNBSS. Cancer diagnoses were reported and identified through active follow-up of each participant until 1989 and also by periodic record linkages with the national cancer registry. Deaths were identified through active follow-up of each participant until 1989 and by periodic record linkages with the Canadian mortality database at statistics Canada.
Study population and period
This analysis used 1980–1989 as the study period because this was the active follow-up period of the study and the time within which women received breast screening as a part of the CNBSS. Analyses of all women were performed with both the inclusion and exclusion of women diagnosed with breast cancer within the first 6 months of enrollment. The exclusion of these women was done to avoid long-term so-called “prevalent cancers.” We assumed that the exclusion of prevalent cancers would result in a more accurate estimate of the probability of developing breast cancer, as women did not have a detectable breast cancer at the beginning of follow-up. As a result, out of 89,835 women, 1332 were considered to have invasive breast cancer and of these, 388 were prevalent cancers. As well, women with breast cancers with size not specified (n = 150) were excluded from the analyses, resulting in a total of 1182 invasive breast cancers including prevalent breast cancers, and a total of 834 breast cancers with prevalent breast cancers excluded. An additional participant with no data on mode of detection was also excluded, resulting in a total of 1181 invasive breast cancers, with the inclusion of prevalent tumors. [Figure 1] is a flow diagram of the CNBSS data and shows the analytical cohort of the study, including how many women were randomized to each group.
|Figure 1: Flow diagram of the Canadian National Breast Screening Study and this study|
Click here to view
General linear modeling (GLM) was used to compare the cancer sizes by allocation group, age group, and the five modes of breast cancer presentation or detection: screen-detected (mammography only, physical breast examination only, or both mammography and physical breast examination), interval, and incident cancers. This was carried out using PROC GLM in SAS version 9.3 (SAS Institute, Cary, NC, USA). Analyses were considered significant if they reached a level of 0.05. Statistically significant overall results were further analyzed with Holm-adjusted pair-wise t-tests to determine which group(s) differed statistically from the others.
To address the second study objective – if a statistically significant difference exists between detection methods for various tumor sizes – we performed McNemar's testsand odds ratios (ORs). McNemar's tests were performed separately for the 40–49 and 50–59 age groups, with both the inclusion and exclusion of prevalent cancers, stratified by tumor size, comparing cancers detected only by mammography with those detected only by physical breast examination. McNemar's test for changes is used to analyze dependent dichotomous measures such as when individuals receive two different diagnostic tests. McNemar's tests were carried out using PROC FREQ and EXACT MCNEM in SAS version 9.3 (SAS Institute, Cary, NC, USA). In addition to supplement McNemar's tests, we performed ORs and 95% confidence intervals (CIs) separately for the 40–49 and 50–59 age groups, with both the inclusion and exclusion of prevalent cancers, stratified by tumor size, comparing cancers detected only by mammography with those detected only by physical breast examination. The 95% CIs were calculated as described by Schlesselman.
| > Results|| |
A total of 89,671 women were considered in this analysis with the inclusion of prevalent cases, and 89,283 women were considered with the exclusion of prevalent cases. Of these, 1.3% (1181 women) were diagnosed with invasive breast cancer if prevalent cases were included, and 0.9% (834 women) were diagnosed with invasive breast cancer if prevalent cases were excluded. Summary statistics of tumor sizes stratified by age group, method of detection, and allocation group are presented in [Table 1]. The reported tumor sizes in the CNBSS ranged from 1 to 90 mm.
Difference in tumor size by detection method, age group, and allocation
The overall ANOVA model was significant (F (15, 1165) = 3.84, P < 0.001); [Table 2]. More specifically, a difference between groups was found for age (F = 6.87 and P = 0.009) and modes of detection or presentation (F = 11.53 and P < 0.001). As shown in [Table 3], a significant difference was observed between mammography screening and all other detection and presentation methods: t(440) = −5.89, P < 0.001 in relation to those detected only by physical breast examination; The difference in size of mammography vs. physical examination detection is probably due to overdetection by mammography; t(203.77) = −4.23, P < 0.001 when compared to those detected by both physical breast examination and mammography; t(352.03) = −7.41, P < 0.001 in comparison to those presenting as incident cancers; and t(331.54) = −6.07, P < 0.001 in relation to those presenting as interval cancers. No other statistically significant differences between detection method groups were observed. None of the other detection or presentation groups were found to differ from one another.
|Table 2: Difference in tumor size by allocation group, detection method, and age group|
Click here to view
These results are consistent with the mean and standard deviation (SD) for the sizes of breast cancers detected only by mammography for both age groups and allocation methods [Table 4] where mean = 14.4 mm (SD = 9.9 mm); compared to cancers detected only by physical breast examination where mean = 20.5 mm (SD = 10.6 mm); cancers detected by both mammography and physical breast examination where mean = 20.8 mm (SD = 14.0 mm); those presenting as interval cancers where mean = 22.0 mm (SD = 13.2 mm); and those presenting with incident cancers where mean = 22.0 mm (SD = 13.3 mm). When we stratified mammography detected tumor sizes by age, we observed that the mean tumor size for women aged 40–49 years was 16.0 mm (SD = 13.8 mm) for women aged 40 to 49 years, and 13.6 mm (SD = 7.1 mm) for women aged 50 to 59 years – a difference that was not statistically significant [Table 3].
|Table 4: Tumor sizes in millimeters for detection method and age, independently|
Click here to view
As shown in [Table 2] and [Table 3], we found a statistically significant difference between the 40–49 and 50–59 age groups t(1054.4) =3.41, P < 0.001. Further analyses were performed to determine differences between the 40–49 and 50–59 age groups; no such difference was detected between the 40–44 and 45–49 age groups, or the 50–54 and 55–59 age groups [Table 3].
Effect of tumor size on screening detectability
Tumour size was stratified by 5 mm groups up to 50 mm. [Table 5] and [Table 6] show the results of the McNemar's tests and ORs, stratified by tumor size, and indicate the number of individuals falling under each detection method, as well as the P values and CIs for each strata. For women aged 40–49 years, we observed no statistically significant difference in cancer detectability between mammography screening alone and physical breast examination alone for any tumor sizes, irrespective of the inclusion or exclusion of prevalent cancers. However, we observed statistically significant differences in cancer detectability between mammography screening alone and physical breast examination alone for women 50–59 years of age, including prevalent cancers, for tumor sizes ranging from 0 to 5 mm (OR = 3.8, 95% CI: 1.23–13.97, P = 0.007), 5–10 mm (OR = 22, 95% CI: 2.97–163.22, P < 0.001), 10–15 mm (OR = 3.33, 95% CI: 1.43–8.28, P = 0.001), and 15–20 mm (OR = 2.07, 95% CI: 1.01–4.38, P = 0.03). No statistically significant differences in cancer detectability were observed between mammography screening alone and physical breast examination alone for tumour sizes that were greater than 20 mm for women in the 50-59 year age group. When prevalent cancers were excluded in the 50–59 age group, we observed statistically significant differences in cancer detectability between mammography screening alone and physical breast examination alone for tumor sizes ranging from 5 to 10 mm (OR = 17, 95% CI: 2.26–127.75, P < 0.001), and 10–15 mm (OR = 5.75, 95% CI: 1.72–24.96, P < 0.001). No statistically significant difference in cancer detectability were observed between mammography screening alone and physical breast examination alone for tumor sizes >15 mm when prevalent cancers were excluded.
|Table 5: McNemar's tests and odds ratios: The effect of tumor size on screening sensitivity (ages 40-49)|
Click here to view
|Table 6: McNemar's tests and odds ratios: The effect of tumor size on screening sensitivity (ages 50-59)|
Click here to view
| > Discussion|| |
The results of these analyses indicated a statistically significant difference for tumor size between detection or presentation methods and between age groups. The results of the t-tests indicated that tumors detected by mammography alone are statistically significantly smaller than those found by all other detection or presentation methods, with the other detection and presentation methods not statistically differing from one another. [Figure 2] shows a cumulative distribution plot of tumor size by various detection methods. There does not appear to be a point on the plot where all detection methods overlap to indicate that a certain percent of tumors are detected by all detection methods at a particular tumor size. However, it is clear from the plot that mammography screening is the most sensitive detection method for all tumor sizes. These results are consistent with other literature.,
|Figure 2: Cumulative distribution plot for tumor size (mm) by method of detection. (1) Mammography only; (2) physical examination only; (3) physical examination and mammography; (4) incident; (5) interval|
Click here to view
Michaelson et al. estimated that the median tumor size of mammography-detectable tumours is 7.5 mm, while the median tumor size of cancers presenting clinically is 15 mm. The median tumor size for mammography-detected tumors in our study was 12 mm for women aged 40–49 years, and 13 mm for women aged 50–59 years. The median tumor sizes for all other groups was 20 mm, with the exception of cancers detected by only physical examination in the 40–49 years age group (18 mm) and cancers detected by both physical breast examination and mammography in the 50–59 years age group (16 mm). However, Michaelson et al. used a method of back-calculation which they believed made it possible to determine the size of each intervening and subsequently screen-detected tumor at the previous breast cancer screening. Thus, their estimates are not comparable with our findings, which were based on the actual tumor sizes as reported by the pathologist, and we believe their results are underestimated. The indication that mammography detection identifies smaller tumors than other detection methods is verified by Güth et al. who found median tumor sizes of 12 mm for tumors identified via radiological breast examination, and 21 mm for tumors identified through a clinical breast examination. Elmore et al. conducted a review of randomized controlled trials assessing the effectiveness of breast cancer screening. They also reviewed meta-analyses, systematic reviews, community studies, and guidelines and found a general trend whereby tumors identified through mammography screening were significantly smaller than tumors identified by other methods.
Our analysis more formally indicates a statistically significant difference in tumor sizes by age [Table 2], [Table 3], [Table 4]. However, a similar trend in tumor size distribution was observed across all detection methods for both age groups, resulting in no statistically significant difference in tumor sizes overall across the two age groups for each respective detection method [Table 2]. Taking into consideration all invasive breast cancers, women aged 40–49 years had a mean tumor size of 21.9 mm (SD = 14.0), while women aged 50–59 years had a mean tumor size of 19.4 mm (SD = 11.1). In other words, tumors identified in women aged 40–49 years were larger than those identified in women aged 50–59 years, t(1054.4) = 3.41, P < 0.001. Detecting breast cancer in younger women is more challenging because they tend to have denser breast tissue. Despite the fact that this study did not explicitly measure and assess breast density, the results of this study are consistent with other studies finding breast density positively associated with tumor size.,,
The present analysis indicates that when a breast cancer presents clinically (i.e., as an interval or incident cancer) it will be, on average, just >20 mm in size [Table 4]. Some studies have found that interval cancers are more likely to be diagnosed in younger women or those with denser breasts, while screen-detected cancers are more likely to be diagnosed in women with lower breast densities., In addition, Aiello et al. indicated that screen-detected tumors are more likely to be larger in dense breasts, but this association is not found for interval cancers. In our study, there appeared to be more screen-detected cancers identified in older women, who would be expected to have lower breast density. However, in our study, there were also more women with interval breast cancers that presented in the 50–59 age group (112 in total women; or 94.8/1000 breast cancers identified as interval cancers) compared to the 40–49 age group (73 in total women; 61.8/1000 breast cancers identified as interval cancers). Consistent with Aiello et al., we found that screen-detected tumors were larger in younger women, but we did not find this association for interval cancers, where the median is 20 mm for interval cancers for both age groups, but with a mean of 23.5 mm (SD = 15.6) for women aged 40–49 and 21.2 mm (SD = 13.1) for women aged 50–59. This difference was not statistically significant in our analysis [Table 7].
|Table 7: Difference in tumor size for interval cancers by age group and allocation group|
Click here to view
For women aged 40–49 years, we found no statistically significant differences between the two detection methods for all tumor sizes. In contrast, for women aged 50–59 years, a statistically significant difference in screening detection methods was observed for tumors up to 20 mm with the inclusion of prevalent cases, and from 5 to 15 mm when prevalent cancers were excluded. It is possible that there is insufficient statistical power to detect the difference for the 0–5mm and 15–20 mm tumor size groups when prevalent cases are excluded. This explanation also applies to the ORs performed to detect differences between mammography screening alone and physical examination alone. For example, the fact that a statistically significant difference was observed for the 15–20 mm group, when prevalent cancers were included for women aged 50-59 years but not when the prevalent cases were excluded, may be explained by the fact that the sample size was larger when prevalent cases were included (29 vs. 14 cases), resulting in a narrower CI [Table 6].
In this analysis, we compared screen-detected cancers that are palpable and nonpalpable, and the results are consistent with other literature. The lack of a statistically significant difference between mammography screening alone and physical examination alone for women aged 40–49 years may be due to the lack of statistical power as the result of the small number of individuals in each tumour size stratum. However, we found a very little difference when tumor sizes were classified in increments of 10 mm instead of 5 mm, with women aged 40–49 still showing no statistically significant differences between mammography screening alone and physical examination alone across all tumor sizes. Therefore, the lack of statistically significant difference between mammography screening alone and physical examination alone for women aged 40–49 years is most likely because mammography is less sensitive in terms of tumor identification for this age group. As discussed earlier, younger women have denser breasts which may be one explanation for the observed lack of significance. For women aged 50–59 years, we observed a significant difference between detection methods for smaller tumors and found no difference for larger tumors. If we only take into account the results obtained with the inclusion of prevalent cases, we can conclude that the difference in the sensitivity of mammography screening alone and physical breast examination alone disappears as tumors reach 20 mm, the size of stage II breast cancers. Therefore, we suggest that women, particularly those in their 50s, may have the greatest potential to benefit from mammography screening. From these results, there appears to be a higher possibility of identifying smaller breast tumors due to the less dense biological nature of the breast in women aged 50–59 years.
This study consists of some clear and arguable limitations. The choppy appearance of the cumulative distribution plot [Figure 2] is an indication of digit preference bias for tumor size. More specifically, there is a preference for tumor sizes ending in “0” and “5.” It appears that the pathologists responsible for determining the sizes of tumors, irrespective of the detection method, have a decided preference to round up or down the observed tumor sizes to the nearest five. Digit preference is a major clinical problem previously described in the literature and commonly associated with the recording of hypertension, weight, height, and emergency department times.,,,, To our knowledge, this is the first clinical case to describe digit preference in the context of breast cancer tumor size recording. This is a major limitation in the use of sample medians on noncontinuous data and is responsible for the similar median tumor sizes observed in our analysis in [Table 1], especially for cancers that presented clinically (interval and incident cancers). In addition, this may contribute to the fact that mammography appears capable of identifying tumors approximately 12 mm in size 50% of the time (median) for the 40–49 years age group, and 13 mm in size 50% of the time for women aged 50–59 years. This suggests that tumors detected by mammography in younger women are smaller than those detected in older women. This nonintuitive result is not as widely supported by the literature and is not consistent with the mean tumor sizes observed in this analysis.
We thank Dr. Elizabeth Thompson for editing the manuscript.
Financial support and sponsorship
The Collaborative Health Research Program funded by the Natural Sciences and Engineering Research Council of Canada and the Canadian Institute of Health Research of Canada.
Conflicts of interest
There are no conflicts of interest.
| > References|| |
Van Engen R, Skaane P, Thibault F, Wilson AR. Breast imaging and screening –Technology update. EJC Suppl 2008;6:4-6.
Pisano ED, Gatsonis C, Hendrick E, Yaffe M, Baum JK, Acharyya S, et al
. Diagnostic performance of digital versus film mammography for breast-cancer screening. N
Engl J Med 2005;353:1773-83.
Alakhras M, Bourne R, Rickard M, Ng KH, Pietrzyk M, Brennan PC. Digital tomosynthesis: A new future for breast imaging? 2013;68:e225-36.
Berg WA, Zhang Z, Lehrer D, Jong RA, Pisano ED, Barr RG, et al
. Detection of breast cancer with addition of annual screening ultrasound or a single screening MRI to mammography in women with elevated breast cancer risk. JAMA 2012;307:1394-404.
Saslow D, Boetes C, Burke W, Harms S, Leach MO, Lehman CD, et al
. American Cancer Society guidelines for breast screening with MRI as an adjunct to mammography. CA Cancer J Clin 2007;57:75-89.
Sardanelli F, Giuseppetti GM, Panizza P, Bazzocchi M, Fausto A, Simonetti G, et al
. Sensitivity of MRI versus mammography for detecting foci of multifocal, multicentric breast cancer in fatty and dense breasts using the whole-breast pathologic examination as a gold standard. AJR Am J Roentgenol 2004;183:1149-57.
Moore SG, Shenoy PJ, Fanucchi L, Tumeh JW, Flowers CR. Cost-effectiveness of MRI compared to mammography for breast cancer screening in a high risk population. BMC Health Serv Res 2009;9:9.
Taghipour S, Banjevic D, Miller AB, Montgomery N, Jardine AK, Harvey BJ. Parameter estimates for invasive breast cancer progression in the Canadian national breast screening study. Br J Cancer 2013;108:542-8.
Chen Y, Brock G, Wu D. Estimating key parameters in periodic breast cancer screening-application to the Canadian national breast screening study data. Cancer Epidemiol 2010;34:429-33.
Weedon-Fekjær H, Tretli S, Aalen OO. Estimating screening test sensitivity and tumour progression using tumour size and time since previous screening. Stat Methods Med Res 2010;19:507-27.
Weedon-Fekjær H, Lindqvist BH, Vatten LJ, Aalen OO, Tretli S. Breast cancer tumor growth estimated through mammography screening data. Breast Cancer Res 2008;10:R41.
Michaelson J, Satija S, Moore R, Weber G, Halpern E, Garland A, et al
. Estimates of breast cancer growth rate and sojourn time from screening database information. JWI 2003;5:11-9.
Taghipour S, Banjevic D, Fernandes J, Miller AB, Montgomery N, Harvey BJ, et al
. Incidence of invasive breast cancer in the presence of competing mortality: The Canadian national breast screening study. Breast Cancer Res Treat 2012;134:839-51.
Taghipour S, Banjevic D, Fernandes J, Miller AB, Montgomery N, Jardine AK, et al
. Predictors of competing mortality to invasive breast cancer in the Canadian national breast screening study. BMC Cancer 2012;12:299.
Taghipour S, Caudrelier LN, Miller AB, Harvey B. Modelling and validating invasive breast cancer progression in women in the study and control groups of the CNBSS I and II. J R Soc Med Decis Making 2016; [Under second round of review].
Miller AB, Baines CJ, To T, Wall C. Canadian National Breast Screening Study: 1. Breast cancer detection and death rates among women aged 40 to 49 years. CMAJ 1992;147:1459-76.
Miller AB, Baines CJ, To T, Wall C. Canadian national breast screening study: 2. Breast cancer detection and death rates among women aged 50 to 59 years. CMAJ 1992;147:1477-88.
Glantz SA. Primer of Biostatistics. 7th
ed. New York: McGraw-Hill; 2012.
Schlesselman JJ. Case Control Studies: Design Conduct Analysis. New York: Oxford University Press; 1982.
Michaelson J, Satija S, Moore R, Weber G, Halpern E, Garland A, et al
. Estimates of the sizes at which breast cancers become detectable on mammographic and clinical grounds. JWI 2003;5:3-10.
Güth U, Huang DJ, Huber M, Schötzau A, Wruk D, Holzgreve W, et al
. Tumor size and detection in breast cancer: Self-examination and clinical breast examination are at their limit. Cancer Detect Prev 2008;32:224-8.
Elmore JG, Armstrong K, Lehman CD, Fletcher SW. Screening for breast cancer. JAMA 2005;293:1245-56.
Heine JJ, Malhotra P. Mammographic tissue, breast cancer risk, serial image analysis, and digital mammography. Part 2. Serial breast tissue change and related temporal influences. Acad Radiol 2002;9:317-35.
Roubidoux MA, Bailey JE, Wray LA, Helvie MA. Invasive cancers detected after breast cancer screening yielded a negative result: Relationship of mammographic density to tumor prognostic factors. Radiology 2004;230:42-8.
Sala E, Solomon L, Warren R, McCann J, Duffy S, Luben R, et al
. Size, node status and grade of breast tumours: Association with mammographic parenchymal patterns. Eur Radiol 2000;10:157-61.
Aiello EJ, Buist DS, White E, Porter PL. Association between mammographic breast density and breast cancer tumor characteristics. Cancer Epidemiol Biomarkers Prev 2005;14:662-8.
Mandelson MT, Oestreicher N, Porter PL, White D, Finder CA, Taplin SH, et al
. Breast density as a predictor of mammographic detection: Comparison of interval- and screen-detected cancers. J Natl Cancer Inst 2000;92:1081-7.
Ciatto S, Visioli C, Paci E, Zappa M. Breast density as a determinant of interval cancer at mammographic screening. Br J Cancer 2004 26;90:393-6.
Gocgun Y, Banjevic D, Taghipour S, Montgomery N, Harvey BJ, Jardine AK, et al
. Cost-effectiveness of breast cancer screening policies using simulation. Breast 2015;24:440-8.
Wen SW, Kramer MS, Hoey J, Hanley JA, Usher RH. Terminal digit preference, random error, and bias in routine clinical measurement of blood pressure. J Clin Epidemiol 1993;46:1187-93.
Burt VL, Cutler JA, Higgins M, Horan MJ, Labarthe D, Whelton P, et al
. Trends in the prevalence, awareness, treatment, and control of hypertension in the adult US population. Data from the health examination surveys, 1960 to 1991. Hypertension 1995;26:60-9.
Thavarajah S, White WB, Mansoor GA. Terminal digit bias in a specialty hypertension faculty practice. J Hum Hypertens 2003;17:819-22.
Niedhammer I, Bugel I, Bonenfant S, Goldberg M, Leclerc A. Validity of self-reported weight and height in the French GAZEL cohort. Int J Obes Relat Metab Disord 2000;24:1111-8.
Rowland ML. Self-reported weight and height. Am J Clin Nutr 1990;52:1125-33.
[Figure 1], [Figure 2]
[Table 1], [Table 2], [Table 3], [Table 4], [Table 5], [Table 6], [Table 7]