Cancer Causes & Control (2018) 29:233–241

https://doi.org/10.1007/s10552-017-0995-7

ORIGINAL PAPER

The performance of mobile screening units in a breast cancer

screening program in Brazil
Z. R. Greenwald1,2 · J. H. Fregnani3 · A. Longatto‑Filho3,4,5,6 · A. Watanabe7 · J. S. C. Mattos7 · F. L. Vazquez3 ·
E. L. Franco1,2

Received: 15 May 2017 / Accepted: 9 December 2017 / Published online: 18 December 2017
© Springer International Publishing AG, part of Springer Nature 2017

Abstract
Purpose In Brazil, access to breast cancer screening outside of urban centers is limited. This study aims to describe the cov-
erage and performance of a breast cancer screening program implemented with Mobile Screening Units (MSU) in northern
São Paulo state.
Methods This is a retrospective cohort study of a population-based mammography program targeting women ages 40–69
in 108 municipalities from 12/2010 to 07/2015. Screening coverage rates were estimated using the Brazil 2010 census data.
We calculated performance measures for the number of exams, recalls, and detected cases of cancer. Screen-detected cases
were compared to clinically detected cases using hospital cancer registry data and a propensity-score matching method. The
down-staging of screen-detected cases relative to clinically detected cases was assessed using logistic regression to calculate
risk ratios (RRs) with 95% confidence intervals.
Results 122,634 women were screened through the MSU program, representing a cumulative coverage rate of 54.8% in the
target population. For initial and subsequent rounds, recall rates were 12.25 and 6.10% and cancer detection rates were 3.63
(95% CI 3.23–4.10) and 1.94 (95% CI 1.59–2.41), respectively. 92.51% of referrals were successful. Screen-detected cases
had more favorable prognoses than clinically detected cases, including smaller tumor size and a decreased risk of late-stage
detection (RR 0.14 95% CI 0.074–0.25).
Conclusions MSUs are a feasible method for the delivery of mammography services in this setting. Patients who had breast
cancer detected on an MSU had favorable prognostic factors when compared with clinically detected cases arising from the
same target population.

Keywords Breast cancer · Mammography · Screening and early detection · Mobile screening units · Cancer prevention ·
Brazil

Introduction

There is an urgent need to improve early detection of breast

Electronic supplementary material The online version of this cancer in low- and middle-income countries (LMICs), which
article (https://doi.org/10.1007/s10552-017-0995-7) contains typically do not have structured screening programs yet bear
supplementary material, which is available to authorized users.

4
* Z. R. Greenwald Laboratory of Medical Investigation (LIM) 14, Faculty
zoe.greenwald@mail.mcgill.ca of Medicine, São Paulo University, FMUSP, São Paulo,
Brazil
1
Division of Cancer Epidemiology, McGill University, 5
Life and Health Sciences Research Institute, ICVS, School
5100 Maisonneuve Blvd West, Suite 720, Montréal,
of Health Sciences, Uminho University, Braga, Portugal
Québec H4A3T2, Canada
6
2 ICVS/3B’s - PT Government Associate Laboratory,
Department of Oncology, McGill University, Montréal,
Braga/Guimarães, Portugal
Québec, Canada
7
3 Prevention Department, Barretos Cancer Hospital, Barretos,
Department of Teaching and Research, Barretos Cancer
São Paulo, Brazil
Hospital, Barretos, São Paulo, Brazil

13
Vol.:(0123456789)
234
the highest burden of cancer morbidity and mortality [1, 2].
In Brazil, breast cancer is the most common cancer among
women with an estimated 67,300 cases in 2012 [3]. Demo-
graphic shifts in the population age distribution are projected
to lead to a 70.5% increase to 114,700 new cases per year
by 2035. Additionally, shifts in lifestyle factors and repro-
ductive factors among Brazilian women, including younger
age at menarche, delayed childbirth, decreased parity, and
increasing body weight, are likely to augment breast cancer
incidence rates in coming years, contributing to future chal-
lenges the healthcare system faces with respect to cancer
prevention [4].
Mammography screening in combination with appropri-
ate treatment significantly decreases cancer mortality [5].
However, mammography screening programs are commonly
alleged to be infeasible in LMICs due to issues of cost,
logistical and cultural barriers [6]. In Brazil, the National
Cancer Institute (INCA) recommends breast cancer screen-
ing beginning at age 40 via annual clinical breast exams for
women ages 40–49 and biennial mammography screening
for women ages 50–69 [7]. Women can access breast cancer
prevention services opportunistically, free of charge through
the Unified Health System (Sistema Único de Saúde, SUS)
regardless of age [8]. As a result of the availability of early
detection services in combination with treatment, the breast
cancer survival rates in Brazil have improved markedly over
the past 10 years [9]. However, inequalities in access to can-
cer screening persist, often in association with geographic
and socio-economic disparities [10]. As there is no national
organized breast cancer screening program, individuals most
commonly access breast cancer prevention services oppor-
tunistically through the public or private systems [11].
Rural populations have fewer opportunities to participate
in cancer screening and consequently tend to receive can-
cer diagnoses upon clinical manifestations of disease (late
stage), when therapy is less effective and survival probability
is lower [12]. Brazil presents unique challenges in the imple-
mentation of organized mammography screening, including
a large territory, limited health infrastructure in the more
rural interior regions of the country, limited medical equip-
ment and health personnel to provide mammography ser-
vices, and a marked increase in demand for screening, due to
the large population of women entering targeted age ranges.
Since 2003, a population-based mammography screening
program, implemented with Mobile Screening Units (MSUs)
and designed by Barretos Cancer Hospital (Brazil), has oper-
ated in rural and urban regions of northern São Paulo state.
Previous studies have described the operations of the pro-
gram in the 19 cities of the Barretos Regional Health Dis-
trict (Departamento Regional da Saúde, DRS V), logistical
aspects with respect to attaining coverage during the initial
implementation, and quality performance measures [13–16].
Performance indicators were found to be within acceptable
13
Cancer Causes & Control (2018) 29:233–241
ranges based on guidelines of the European Breast Cancer
Network [16]. The clinical quality of the program is moni-
tored by biennial external audits from the Dutch Refer-
ence Centre for Screening to assure performance quality of
radiologists and radiographers, and the technical quality of
mammography equipment [17]. Since 2010, the program has
expanded by adding two new MSUs to provide screening in
additional regions of São Paulo State. The target population
for screening has quadrupled, from 54,238 eligible women
ages 40–69 residing in the Barretos (DRS V) to 223,467
women residing in 108 municipalities across three districts
(also including parts of DRS II—Araçatuba and DRS XV—
São Jose do Rio Preto).
This study aims to assess the performance of the
expanded MSU screening program with respect to coverage
rates within the target population; performance indicators
including abnormal call rates, biopsy rates, and case detec-
tion rates; factors associated with the successful completion
of referrals; and stage distribution of screen-detected cases
relative to cases arising from the same target population
among women who did not participate in the program.
Methods
This is a retrospective cohort study of the MSU breast cancer
program performance. Patients who accessed mammography
screening were followed over time to measure the detec-
tion of breast cancer. We compared clinical characteristics
at diagnosis of cases detected through the screening program
to a control group of cases detected outside of the screen-
ing program and included in the hospital cancer registry, to
estimate whether risk of late-stage detection differs between
patients exposed or unexposed to screening. Data on mam-
mography screening taking place on MSUs and clinical
follow-up taking place at fixed-site breast clinics in hospi-
tals are routinely collected in the electronic medical health
records of the Barretos Cancer Hospital (BCH). The BCH
also has a hospital cancer registry database, which includes
data on incident cases of cancer. We extracted data retro-
spectively from these two databases in August 2015 for the
period of December 2010–July 2015. Databases were linked
using patient-identification numbers to identify breast cancer
cases detected among participants of the screening program.
The MSU breast cancer program operates via population-
based screening whereby women in the target population
(ages 40–69, targeted municipalities) are invited for an ini-
tial screening exam which is coordinated with the assistance
of community healthcare agents within a given municipal-
ity. Four MSUs were used in the screening program: two
equipped with digital mammography and two with analogue
film-screen mammography. Appointments are pre-booked
and women are screened by a radiologist-technician, at a rate
Cancer Causes & Control (2018) 29:233–241
of approximately 10 patients per hour, and 60 patients per
day. If clients also wish to access cervical cancer screening,
they can receive a Pap smear in an exam room adjacent to
the mammography room on the MSU. Clients access ser-
vices without any fees using their SUS—Unified Health
System card (for the purposes of fee reimbursement to the
Barretos Cancer Hospital by the government), and a medical
record with complete contact information is entered into an
electronic system to facilitate the follow-up of exam results.
Mammography results are reported using the Breast
Imaging Reporting and Data System (BIRADS) [18].
Exams of BIRADS scores 1 and 2 are normal/benign (nega-
tive results), BIRADS 3 warrants a short-interval follow-up
(within 6 months), BIRADS scores 4 and 5 are considered to
be suspicious of malignancy, while BIRADS 0 is an incon-
clusive result. Abnormal or inconclusive exams (BIRADS
0, 3, 4, or 5) requiring follow-up were referred to the nearest
fixed-site breast clinic for additional imaging (ultrasound
or diagnostic mammography), or biopsy and histopathol-
ogy exam if warranted. For normal exams, clients were re-
invited to screen at 1-year intervals for women ages 40–49
and 2-year intervals for women ages 50–69, in accordance
with policies from the Barretos Cancer Hospital.
The size of the targeted study population was estimated
using the Brazilian 2010 National Census to measure the
number of women ages 40–69 residing in each of 108
municipalities where the MSU operates [19]. The number
of screening exams conducted by the Barretos Cancer Hos-
pital program per age group, city, and year was extracted
from SISMAMA (an information system for breast cancer
detection programs managed by SUS) [8, 20]. We calcu-
lated coverage rates as the ratio of the number of screen-
ing mammography exams performed relative to the number
of eligible women per age group (40–49, 50–59, or 60–69)
per municipality. Cumulative coverage rates were calcu-
lated to estimate the proportion of the target population that
was screened at least once during the period of December
2010–July 2015. Test productivity rates were also calculated
as the number of screening exams performed relative to the
eligible population during recommended intervals; annual
test productivity rates were calculated for women ages 40–49
and biennial rates for women ages 50–69 from January 2011
to December 2014. The number of patients accessing screen-
ing during the study period determined the final sample size.
Performance measures were analyzed for women receiv-
ing an initial screening exam between December 2010 and
December 2012. We restricted to these dates in order to
ensure a sufficient length of follow-up time was available to
monitor interval cancers, occurring within 24 months of a
screen-negative (BIRADS 1 or 2) result. Patients reporting
any breast-related symptoms during mammography screen-
ing were excluded from the analysis to minimize bias in the
estimation of detection rates. Performance indicators were
235
calculated following standardized methods reported in the
European Guidelines [21] and included recall rates, cancer
detection rates, positive predictive values (PPV) for screen-
ing mammography and for performed biopsies, interval can-
cer rates, and tumor sizes. Cases were confirmed by histopa-
thology reports and cross-checked with the hospital cancer
registry to access data on clinical staging and tumor size
measured by Tumour Node Metastasis (TNM) system [22].
Data from the hospital registry were used to investigate the
occurrence of interval cancers, measured as breast cancers
occurring within 24 months following a normal (BIRADS
1 or 2) mammography result [21].
Referrals were issued for all screen-positive women
(BIRADS 0, 3, 4, 5) and were considered successful if the
patient reported to the breast clinic for a follow-up visit to
clarify or rule out a breast cancer diagnosis. Follow-up visits
may have included ultrasound exams, diagnostic mammog-
raphy and/or biopsy as needed. An analysis of the referral
system investigated referral effectiveness, measured as the
proportion of abnormal exams successfully followed up, and
referral efficiency, measured in number of days from the
first abnormal screening exam to the first follow-up visit
in hospital. The efficiency of referrals was plotted with
Kaplan–Meier curves to model the cumulative probabil-
ity of a successful exam follow-up by BIRADS score [23].
To investigate factors associated with more rapid follow-
up times, a Cox Proportional Hazards model was used to
estimate Hazard Ratios (HRs) and 95% confidence intervals
(CIs) for successful follow-up adjusting for potential effects
due to BIRADS score, client age, and distance in kilome-
tres from the client’s municipality of residence to the breast
clinic where the follow-up exam was performed (either the
Barretos Cancer Hospital or a satellite breast clinic in Jales
or Fernandópolis, São Paulo State).
Data from the hospital cancer registry were used to com-
pare the staging of screen-detected cases of breast cancer
to clinically detected cases. In order to reduce bias in the
comparison of screen-detected and clinically detected cases,
we used propensity-score matching to identify one clinically
detected control for each screen-detected case on the basis
of covariates that may affect probability to participate in
screening including age, regional health district, and educa-
tion level [24]. Registry data were first restricted to include
only cases arising from the same base population eligi-
ble for screening (women ages 40–69 residing in munici-
palities where screening is offered). Next, screen-detected
cases were matched to clinically detected controls using the
nearest-neighbor 1:1 matching method with a 0.05 caliper.
Differences between socio-demographic and clinical vari-
ables in the matched sample were measured using the Chi-
square test. It is too early to assess survival outcomes among
this cohort of patients, so our comparison between screen-
detected and clinically detected cases focused on evidence
13
236 Cancer Causes & Control (2018) 29:233–241

of down-staging, which is a proxy for improved survival
[21]. Down-staging was assessed using logistic regression
to measure the relative risk of having a case detected at a
late stage (Stage III or IV, according to UICC/TNM criteria)
[22]. Data were analyzed using Stata 14 Software [25].
Results
Among a target of 223,467 women eligible for screening, a
total of 122,634 women completed at least one exam on an
MSU between December 2010 and January 2015 (Fig. 1).
Among 717 patients reporting symptoms at initial screening,
24 breast cancers were detected. The symptomatic patients
and cases arising from this group were excluded from subse-
quent analyses. A total of 382 breast cancers were detected
following a positive screen during the initial screening
round. Additionally, 55 interval breast cancer cases were
detected within 24 months following a negative screening
exam. Among non-participants, 834 cases were detected
during the same period.
Table 1 outlines coverage rates by age group. The cumu-
lative coverage rate across all age groups was 54.8% and
the participation rate was highest for younger women
40–49 years (57.5%), followed by women 50–59 (54.7%)
and women 60–69 (54.8%). The test performance rates

Fig. 1  Flow chart of study

population during first round of
screening 2011–2015

Table 1  Coverage rates by Age group Number screened (%) Size of target Coverage rate by screen- Cumulative
examination for women ages population ing interval (%)a coverage rate
40–49, 50–59, 60–69 from (%)b
December 2010 to July 2015
40–49 53,743 (43.9) 93,513 35.3 57.5
50–59 41,819 (34.1) 76,417 54.4 54.7
60–69 26,934 (22.0) 53,537 49.7 50.3
Total 122,634 (100) 223,467 46.5 54.8
a
Coverage rate by screening interval is calculated as the number of mammograms performed relative to the
size of target group (annually for women 40–49; biennially for women 50–69) averaged across full calendar
years from January 2011 to December 2014
b
Cumulative coverage rate is calculated as the total number of women screened relative to the size of target
group during the entire study period from December 2010 to July 2015

13
Cancer Causes & Control (2018) 29:233–241 237

Table 2  Mammography Initial screen Dec 2010– Repeat screen

performance indicators for Dec 2012 Dec 2011–Dec
initial and repeat screens. 2014
This analysis was restricted
to women receiving an initial Number of women screened 74,503 44,648
screen on an MSU during from
Age (Mean, standard deviation) 52.50 (8.17) 53.98 (7.66)
December 2010 to December
2012 Recall rate (n, %) 9,127 (12.25%) 2,724 (6.10%)
Biopsy rate (n, %) 1,695 (2.27%) 387 (0.86%)
Screen-detected cancers (n) 271 88
Case detection rate per 1000 (95% CI) 3.63 (3.23–4.10) 1.96 (1.59–2.41)
Interval cancers (n) 45 –
Positive predictive value (mammography overall)a 2.83% (271/9,127) 3.23% (88/2,724)
PPV mammography (BIRADS 0) 1.45% (120/8,174) 0.95% (23/2,429)
PPV mammography (BIRADS 3) 2.20% (11/500) 0% (0/115)
PPV mammography (BIRADS 4) 23.20% (92/396) 28.75% (44/153)
PPV mammography (BIRADS 5) 84.21% (48/57) 77.77% (21/27)
Positive predictive value (biopsy)b 29.67% (271/1695) 22.74% (88/387)
a
Mammography PPV is calculated as number of screen-detected cases/number of patients who received an
abnormal BIRADS result
b
Biopsy PPV is calculated as the number of detected cases/number of performed biopsies

Fig. 3  Mammography positive predictive value (PPV) by BIRADS

Fig. 2  Breast cancer case detection rates by screening round and age
group
are displayed in Table 2, for the period restricted from
December 2010 to December 2012. Among 74 503 women
screened from December 2010 to December 2012, 44 648
returned for a second screen from 2011 to 2014 (59.9%).
The recall rate was 12.25% in initial screening and 6.10% in
subsequent screening rounds. In the initial screening round,
271 breast cancers were detected by mammography exams,
for a case detection rate of 3.63 per 1000 women screened
(95% CI 3.23–4.10). During repeat screening rounds, the
case detection rate was 1.96 per 1000 women screened (95%
CI 1.59–2.41). Figure 2 illustrates how case detection rates
increase consistently by age and were higher during the
initial screening round relative to the subsequent screening
rounds. The clinical characteristics of breast cancer cases
diagnosed during the initial and repeat rounds were similar,
as shown in Supplemental Table 1.
The positive predictive value (PPV) of mammography
overall was 2.83% during initial screening rounds and 3.23%
score and age group. The test PPV increases by age. BIRADS score 0
has the lowest PPV, relative to BIRADS 4 and 5
in following rounds. However, this figure is downwardly
biased by the inclusion of the BIRADS 0 category in calcu-
lations for PPV. At initial screening, the positive predictive
values of BIRADS 4 or 5 reported mammography was 23.20
and 84.21%, respectively. Additionally, across all BIRADS
categories, PPV increases with age, as shown in Fig. 3.
Analysis of the effectiveness of the referral system
showed that overall 92.35% of women who were recalled
due to an abnormal screening exam on an MSU success-
fully reported to a fixed-site hospital for a follow-up visit
(Table 3). The proportion of successful referrals was high-
est among the women whose BIRADS scores indicated the
greatest suspicion of breast cancer; 100, 97.97, 93.60, and
91.95% of patients returned for a follow-up exam among
BIRADS 5, 4, 3, and 0 categories, respectively. Similarly,
the median number of days between the date of an MSU
exam and date of a first follow-up exam at the hospital breast
clinic follows the level of urgency that is associated with
each BIRADS category. Results from the time-to-event
analysis of the referral system are modeled in Fig. 4. Due

13
238 Cancer Causes & Control (2018) 29:233–241

Table 3  Proportion of successful follow-up among abnormal exams Table 4  Multivariate Cox Proportional-Hazard Model for factors
associated with delays in follow-up
Test category Rate of follow-up exams Median days to
% (n referred)a referral (interquartile Covariate Hazard ratio 95% CI (lower) 95% CI (upper)
range)
Exam result
BIRADS 0 91.95% (7,516/8,174) 152 (82–181) BIRADS 0 Ref. – –
BIRADS 3 93.60% (468/500) 107 (63–156) BIRADS 3 0.72 0.27 0.65
BIRADS 4 97.97% (388/396) 42 (34–67) BIRADS 4 0.27 0.25 0.30
BIRADS 5 100% (57/57) 38 (28–51) BIRADS 5 0.16 0.12 0.21
Total 92.35% (8,429/9,127) – Age (years) 1.00 1.00 1.00
a Kilometers traveled to hospital
Includes initial screening round (December 2010–December 2012)
< 50 KM REF – –
50–100 KM 0.88 0.85 0.95
100–175 KM 1.64 1.53 1.72
1.00

> 175 KM 1.88 1.75 2.00

Proportion with referral visit completed
0.75

Screen-detected cases had clinical characteristics at diag-

0.50

nosis of more favorable prognosis including smaller tumor

size and significantly decreased risk of detection at stage III
or IV (RR = 0.14 95% CI 0.074, 0.25). However, 25.3% of
0.25

screen-detected cases were detected at stage 0 as compared

with 15.2% of clinically detected cases, indicating possi-
0.00

0 30 60 90 180 270 365 ble over-diagnosis. The groups were evenly balanced with
Time (days)
respect to age and education.
BIRADS 0 BIRADS 3
BIRADS 4 BIRADS 5

Fig. 4  Kaplan–Meier plot of cumulative probability of loss to follow-

Discussion
up visit by BIRADS Score. Time interval measured as days between
screen-positive mammography on an MSU to first appointment at This study of the Barretos Cancer Hospital MSU mammog-
a fixed clinic. Event in the Kaplan–Meier analysis is a report for a raphy program identifies a high coverage rate among the
follow-up visit within 1 year of an abnormal or inconclusive mam-
target population. The cumulative coverage rate of 54.8%
mography exam
found in the MSU program falls below the World Health
Organization’s recommendation of 70% coverage [26]. We
to the nature of the referral system, women with BIRADS hypothesize that non-participating women may decline
results of 4 and 5 are booked for appointments more rapidly mammography screening because they are privately insured
than women with BIRADS 0 (inconclusive) and BIRADS and able to access screening elsewhere. Estimates from the
3 (short-interval follow-up) results. When modeled using National Agency for Supplemental Health plans indicate that
a Cox analysis (Table 4), adjusted for BIRADS score, the approximately 30% of women 40–69 years old in interior
effect of distance can be seen, with delays in successful (rural) areas of South-Eastern Brazil hold private insurance
follow-up experienced for women residing further from the plans [27]. Therefore, it is likely that the coverage rate meas-
hospital. Living in a municipality at 100–175 km away from ured in our study is an underestimate of the true coverage
a hospital led to longer delays in referral times relative to rate in this region. The hypothesis that some women in this
municipalities within 50 km of a hospital (HR 1.64, 95% region access mammography screening through the private
CI 1.53–1.72). Probability of delay was also greater among system supports high rates of stage 0–I breast cancer cases
women living > 175 km from a hospital (HR 1.88, 95% CI registered among women not participating in the MSU pro-
1.75–2.00). Paradoxically, clients residing in municipali- gram (Fig. 1), which were likely diagnosed via mammogra-
ties 50–100 km from a hospital had quicker referral times phy referrals rather than symptomatically.
(HR 0.88, 95% CI 0.85–0.95) than the municipalities within The performance indicators of the program indicate a
50 km. Age had no effect on speed of referral (HR 1.00, 95% high recall rate at first screen, of approximately 12% which
CI 1.00–1.00). is above the recommended rate of 6% stipulated by the Euro-
In the matched analysis, 217 screen-detected cases were pean guidelines for quality assurance in breast cancer screen-
compared with 217 clinically detected cases (Table 5). ing and diagnosis [21] but not far above the average recall

13
Cancer Causes & Control (2018) 29:233–241 239

Table 5  Comparison of Clinically MSU-detected Total p value

demographic and clinical detected cases cases
factors between screen-detected
and clinically detected breast No Col % No Col % No Col %
cancer cases
Age group 1.00
40–49 71 32.7 71 32.7 142 32.7
50–59 84 38.7 84 38.7 168 38.7
60–69 62 28.6 62 28.6 124 28.6
Educational attainment 1.00
Illiterate/primary incomplete 42 19.4 42 19.4 84 19.4
Primary school 56 25.8 56 25.8 112 25.8
Secondary school 71 32.7 71 32.7 142 32.7
College or university 42 19.4 42 19.4 84 19.4
Unknown 6 2.8 6 2.8 12 2.8
Morphology 0.039
8500/3—invasive ductal carcinoma 164 75.6 121 65.1 285 70.7
8500/2—ductal carcinoma in situ 30 13.8 48 25.8 78 19.4
8520/3—lobular Carcinoma 12 5.5 10 5.4 22 5.5
Others 11 5.1 38 20.4 49 11.2
Clinical stage < 0.01
Stage 0 33 15.2 55 25.3 88 20.3
Stage I 39 18 73 33.6 112 25.8
Stage II 57 26.3 49 22.6 106 24.4
Stage III 48 22.1 13 6 61 14.1
Stage IV 26 12 1 0.5 27 6.2
Unknown (X) 14 6.5 26 12 39 9
Tumor size < 0.01
In situ 33 15.2 55 25.6 88 20.4
< 20 mm 44 20.3 77 35.8 121 28
20–50 mm 56 25.8 46 21.4 102 23.6
> 50 mm 62 28.6 11 5.1 73 16.9
Unknown 22 10.1 27 12.4 50 11.5
Treatment ­receiveda
Surgery (mastectomy/lumpectomy) 158 72.8 166 89.2 324 80.4 < 0.01
Radiotherapy 108 49.8 92 49.5 200 49.6 0.95
Chemotherapy 121 55.8 86 46.2 207 51.4 0.06
Endocrine therapy 116 53.5 112 60.2 228 56.6 0.17
Total 217 100 217 100 432 100

Groups were matched by age and education using propensity-score nearest-neighbor matching without
replacement and caliper of 0.05
a
Types of treatments received are not mutually exclusive

rate among US radiologists in community settings (mean
9.8%, range 6.4–13.3%) [28]. We hypothesize the high recall
rate in the MSU program may occur due to the new initiation
of the screening program in a region where women had lim-
ited access beforehand, which means that a large proportion
of women were accessing screening for the first time, con-
tributing to an augmented recall rate. Also, women below
the age of 50 in this screening program have denser breast
tissue which generally leads to decreased clarity of mam-
mographic images and increased rates of BIRADS 0 recalls
for inconclusive images. When restricting the analysis of the
recall rate to only women above age 50, the rate during the
initial rate is 10.75%. The recall rate may also be affected by
the experience levels of radiologists. The practice of double
reading of mammograms is often recommended in order to
minimize bias in recall rates [21]; however, due to cost limi-
tations, the Barretos Cancer Hospital only performs single
readings.
The fact that the screening program was newly initiated
in the region also means that for some of the breast cancer

13
240
cases detected during the initial round of screening, exams
may have served as diagnostic mammograms rather than
screening mammograms. Efforts to limit selection bias in the
detection rates due to this effect were made by the exclusion
of any women reporting symptoms prior to screening.
Strengths of the program include high levels of effective
referrals, which are above rates commonly reported in other
MSU studies [29]. Rates of loss to follow-up as ranging from
10 to 21% have been reported in other MSU studies [29–33].
Our analysis also provided indications of whether distance
between a patient’s city of residence and the fixed-site hos-
pital leads to delays in follow-up. We found an association
of increased delay in follow-up corresponding with distance.
However, it was less than we had initially hypothesized. This
may be due to the nature of the referral system, whereby the
public health department of each city is required to agree to
provide free transport to the Barretos Cancer Hospital for all
patients who have abnormal screening exams.
The median number of days between an abnormal screen-
ing exam and a clinical follow-up visit exceeded the stipula-
tion of the European Guidelines which recommend follow-
up of abnormal screening mammograms within 20 working
days (15 working days from screening mammography to
result + 5 working days from the result of the screening
mammography to the offered assessment) [21]. We were
not able to measure timing of referrals following the recom-
mended method of days between exam and date of offered
assessment, because our data only revealed the date of com-
pleted assessment. Our results therefore reflect the reality
of the timing of the follow-up that occurred, and it is not
possible to separate effects on whether delays were causes
by healthcare providers in terms of offering late dates of
assessment or due to delays that the client is responsible for.
Additionally, we measured our time intervals in total number
of days, without regard to working days versus weekends/
holidays.
We found screen-detected cases of breast cancer to have
favorable prognostic characteristics relative to clinically
detected cases, which indicates a stage shift in detection
associated with participation in screening. It is too soon to
observe whether participation in the screening program is
associated with a survival advantage. However, our findings
of decreased stage at diagnosis in the short term may serve
as a proxy for survival. We interpret the larger total number
of breast cancer cases diagnosed among non-participants
relative to women accessing screening to the fact that cases
can accumulate among populations with infrequent screen-
ing or no screening and become detected at later stages.
This study has limitations with respect to individual-
level data on socio-demographic indicators such as age at
menarche, parity, and age at menopause, factors which may
have allowed risk-factor stratified analyses. Data were also
unavailable for screening history, socio-economic status, and
13
Cancer Causes & Control (2018) 29:233–241
insurance status of study participants, which prevented an
analysis of whether the program served to increase access to
women who had not previously been screened.
It was beyond the scope of this study to include informa-
tion on the cost of MSU implementation, which is a key
consideration impacting the feasibility of MSU operations in
other settings. For participants in this program, no financial
barriers are directly associated with the cost of care as mam-
mography screening and any potential costs of breast cancer
treatment are covered by a mixture of public funding (base-
line reimbursement by SUS and supplemental fees covered
by hospital fundraising). Given the highly contextual nature
of the program implementation, it is difficult to generalize
the results of this study to other settings.
In conclusion, in the northern regions of São Paulo State,
the MSU mammography program provided services for the
majority of the target population. The performance measures
of the MSU were satisfactory and the referral system was
effective in linking patients with abnormal screening exams
to care in the fixed center. The length of time required to
completed referrals remains a challenge. Most importantly,
prognostic characteristics are favorable for screen-detected
breast cancer cases relative to clinically detected cases,
indicating that the mammography screening program offers
the potential to contribute to down-staging of breast cancer
cases in this setting. These findings are based upon opera-
tions from 2010 to 2015 and future projects aimed at meas-
uring survival among screen-detected cases may provide
further evidence of program effectiveness.
Compliance with ethical standards
Conflict of interest The authors have no conflicts of interest to declare.
Ethical approval All procedures in studies involving human partici-
pants were in accordance with the ethical standards of the Barretos
Cancer Hospital, the Brazilian National Research Council (CNPq), and
the McGill University Institutional Review Board and with the 1964
Helsinki declaration and its later amendments or comparable ethical
standards. For this type of study formal consent is not required.
References
1. Bray F, McCarron P, Parkin DM (2004) The changing global
patterns of female breast cancer incidence and mortality. Breast
Cancer Res 6(6):229–239. https://doi.org/10.1186/bcr932
2. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM
(2010) Estimates of worldwide burden of cancer in 2008: GLO-
BOCAN 2008. Int J Cancer 127(12):2893–2917. https://doi.
org/10.1002/ijc.25516
3. Ferlay J, Soerjomataram I, Ervik M, Dikshit R, Eser S, Mathers
C, Rebelo M, Parkin DM, Forman D, Bray F (2013) GLOBOCAN
2012 v1.0, Cancer incidence and mortality worldwide: IARC Can-
cerBase No. 11 [Internet]. International Agency for Research on
Cancer. http://globocan.iarc.fr. Accessed 4 September 2015
Cancer Causes & Control (2018) 29:233–241
4. McCormack VA, Boffetta P (2011) Today’s lifestyles, tomorrow’s
cancers: trends in lifestyle risk factors for cancer in low- and mid-
dle-income countries. Ann Oncol 22(11):2349–2357. https://doi.
org/10.1093/annonc/mdq763
5. Berry DA, Cronin KA, Plevritis SK, Fryback DG, Clarke L,
Zelen M, Mandelblatt JS, Yakovlev AY, Habbema JDF, Feuer
EJ (2005) Effect of screening and adjuvant therapy on mortality
from breast cancer. N Engl J Med 353(17):1784–1792. https://doi.
org/10.1056/NEJMoa050518
6. Anderson BO, Cazap E, El Saghir NS, Yip C-H, Khaled HM,
Otero IV, Adebamowo CA, Badwe RA, Harford JB (2011) Opti-
misation of breast cancer management in low-resource and mid-
dle-resource countries: executive summary of the Breast Health
Global Initiative consensus, 2010. Lancet Oncol 12(4):387–398.
https://doi.org/10.1016/s1470-2045(11)70031-6
7. INCA (2004) Controle do câncer de mama: documento de con-
senso [Internet] Cited 07-11-2017 Availabile from: http://www.
inca.gov.br/publicacoes/Consensointegra.pdf Insituto Nacional de
Câncer, Rio de Janeiro
8. INCA (2011) Sistemas de informação do controle do câncer de
mama (SISMAMA) e do câncer do colo do útero (SISCOLO):
manual gerencial. Instituto Nacional de Câncer. Coordenação
Geral de Ações Estratégicas, Rio de Janeiro
9. Allemani C, Weir HK, Carreira H, Harewood R, Spika D, Wang
X-S, Bannon F, Ahn JV, Johnson CJ, Bonaventure A (2015) Global
surveillance of cancer survival 1995–2009: analysis of individual
data for 25 676 887 patients from 279 population-based registries
in 67 countries (CONCORD-2). Lancet 385(9972):977–1010
10. De Castro Mattos JS, Caleffi M, Da Costa Vieira RE (2013) Ras-
traemento mamográfico no Brasil: resultados preliminares. Rev
Brasil Mastol 23(1):22–27
11. Marchi AA, Gurgel MSC (2010) Adesão ao rastreamento
mamográfico oportunístico em serviços de saúde públicos e pri-
vados. Rev Brasil de Ginecol e Obstetr 32:191–197
12. Lee BL, Liedke PER, Barrios CH, Simon SD, Finkelstein DM,
Goss PE (2012) Breast cancer in Brazil: present status and future
goals. Lancet Oncol 13(3):e95–e102. https://doi.org/10.1016/
S1470-2045(11)70323-0
13. Mauad EC, Nicolau SM, Moreira LF, Haikel RL, Longatto-Filho
A, Baracat EC (2009) Adherence to cervical and breast cancer
programs is crucial to improving screening performance. Rural
Remote Health 9(3):1241–1241
14. Silva TB, Mauad EC, Carvalho AL, Jacobs LA, Shulman LN
(2013) Difficulties in implementing an organized screening pro-
gram for breast cancer in Brazil with emphasis on diagnostic
methods. Rural Remote Health 13(2):2321
15. Haikel RL Jr, Mauad EC, Silva TB, Mattos JS, Chala LF, Lon-
gatto-Filho A, de Barros N (2012) Mammography-based screen-
ing program: preliminary results from a first 2-year round in a
Brazilian region using mobile and fixed units. BMC Women’s
Health 12:32. https://doi.org/10.1186/1472-6874-12-32
16. Tsunoda A, Fregnani J, Oliveira J, Alencar M, Brentani R, Mauad
E (2009) Quality performance indicators in breast cancer screen-
ing and treatment in a Brazilian non-metropolitan setting. Int J
Gynecol Obstet 107:S608
17. Souza Sabino S, Silva TB, Watanabe AH, Syrjanen K, Carvalho
AL, Mauad EC (2014) Implementation of a clinical quality control
241
program in a mammography screening service of Brazil. Antican-
cer Res 34(9):5057–5065
18. Sickles E, D’Orsi C, Bassett L, al. e (2013) ACR BI-RADS®
mammography. In: ACR BI-RADS® Atlas, breast imaging report-
ing and data system. American College of Radiology, Reston
19. IBGE (2013) Atlas do censo demográfico 2010. IBGE, Rio de
Janeiro
20. Passman LJ, Farias AM, Tomazelli JG, de Abreu DM, Dias MB,
de Assis M, de Almeida PF, da Silva RC, Santini LA (2011) SIS-
MAMA–implementation of an information system for breast can-
cer early detection programs in Brazil. Breast 20(Suppl 2):S35–
S39. https://doi.org/10.1016/j.breast.2011.02.001
21. Perry N, Broeders M, De Wolf C, Törnberg S, Holland R, von
Karsa L, Puthaar E (2006) European guidelines for quality assur-
ance in breast cancer screening and diagnosis Fourth Edition.
Office for Official Publications of the European Communities,
Luxembourg
22. Greene F, Page D, Fleming I, Fritz A, Balch C, Haller D (2002)
AJCC cancer staging handbook, 6th edn. Springer, New York
23. Peto R, Pike M, Armitage P, Breslow NE, Cox D, Howard S, Man-
tel N, McPherson K, Peto J, Smith P (1977) Design and analysis
of randomized clinical trials requiring prolonged observation of
each patient. II. analysis and examples. Br J Cancer 35(1):1
24. Stuart EA (2010) Matching methods for causal inference: A
review and a look forward. Stat Sci 25(1):1
25. StataCorp (2015) Stata statistical software: release 14. StataCorp
LP, College Station
26. World Health Organization (2007) Cancer control knowledge into
action. WHO guide for effective programmes. Early Detection,
vol 2. World Health Organization, Geneva
27. ANS (2016) Agência Nacional de Saúde Suplementar; Cobertura
de planos supplementar. http://www.ans.gov.br/anstabnet/cgi-bin/
tabnet?dados/tabnet_tx.def
28. Rosenberg RD, Yankaskas BC, Abraham LA, Sickles EA, Lehman
CD, Geller BM, Carney PA, Kerlikowske K, Buist DS, Weaver
DL, Barlow WE, Ballard-Barbash R (2006) Performance bench-
marks for screening mammography. Radiology 241(1):55–66.
https://doi.org/10.1148/radiol.2411051504
29. Greenwald ZR, El-Zein M, Bouten S, Ensha H, Vazquez FL,
Franco EL (2017) Mobile screening units for the early detection
of cancer: a systematic review. Cancer Epidemiol Biomark Prev.
https://doi.org/10.1158/1055-9965.epi-17-0454
30. Rodriguez-Cuevas S, Guisa-Hohenstein F, Labastida-Almendaro
S (2009) First breast cancer mammography screening program in
Mexico: initial results 2005–2006. Breast J 15(6):623–631
31. Garas I, Pateras H, Triandafilou D, Georgountzos V, Mihas A,
Abatzoglou M, Trichopoulos D (1994) Breast cancer screening
in southern Greece. Eur J Cancer Prev 3(Suppl 1):35–39
32. Séguret F, Daures JP, Guizard AV, Mathieu-Daude H, Bonifacj
JC, Cherifcheik J, Lamarque JL (1995) Herault breast screening
programme: results after 30 months of a mobile French schedule.
Eur J Cancer Prev 4(4):299–305
33. Van Oyen H, Verellen W (1994) Breast cancer screening in the
Flemish region, Belgium. Eur J Cancer Prev 3(Suppl 1):7–12
13