Early Detection and Diagnosis
Outcome of screening by clinical examination of the breast in a trial in the Philippines
The value of screening by Clinical Examination of the Breast (CBE) as a means of reducing mortality from breast cancer (BC) is not established. The issue is relevant, as CBE may be a suitable option for countries in economic transition, where incidence rates are on the increase but limited resources do not permit screening by mammography. Our aims were to assess whether mass screening by CBE carried out by trained para-medical personnel is feasible in an urban population of a low-income country, and its efficacy in reducing BC mortality. Our study was designed as a randomised controlled trial of the effect on BC mortality of 5 annual CBE carried out by trained nurses. The target population was women aged 35–64 years, resident in 12 municipalities of the National Capital Region of Manila, Philippines. The units of randomization were the 202 health centres (HC) within the selected municipalities. During 1995 nurses and midwives were recruited and trained in performing CBE. The first round of screening took place in 1996–1997. The intervention however showed a refractory attitude of the population with respect to clinical follow-up and was discontinued after the completion of the first screening round. Cases of breast cancer occurring in the study population during 1996–1999 were identified by the 2 local population-based registries. In the single screening round 151,168 women were interviewed and offered CBE, 92% accepted (138,392), 3,479 were detected positive for a lump and referred for diagnosis. Of these only 1220 women (35%) completed diagnostic follow-up, whereas 42.4% actively refused further investigation even with home visits, and 22.5% were not traced. Of 53 cases that occurred among screen-positive women in the 2 years after CBE only 34 were diagnosed through the intervention. Eighty cases occurred among screen-negative women. The test sensitivity for CBE repeated annually was 53.2%. The actual sensitivity of the programme was 25.6% and positive predictive value 1%. Screen-detected cases were non-significantly less advanced than the others. Previous studies have shown that most breast cancer cases in the Philippines present at advanced stages and have an unfavourable outcome. Although CBE undertaken by health workers seems to offer a cost-effective approach to reducing mortality, the sensitivity of the screening programme in the real context was low. Moreover, in this relatively well-educated population, cultural and logistic barriers to seeking diagnosis and treatment persist and need to be addressed before any screening programme is introduced. © 2005 Wiley-Liss, Inc.
In the year 2000 breast cancer accounted for over 1 million new cases per year worldwide; it is the most common cancer in women, and incidence rates are rising in low-risk countries.1 These trends are likely to continue, because the current pattern of later childbearing, decreasing fertility, increasing height and weight and ‘westernization’ of diets will all be associated with increased risk.
Significant improvements in the prognosis of early breast cancer have been achieved in the 1980s and 1990s2, 3 and have substantially contributed to the initial reduction of mortality observed in some high-risk countries.4, 5, 6, 7 For treatment to be highly effective however, it is essential that the disease is detected at an early clinical stage.
Possibly because of the low burden relative to other diseases, cancer awareness in low-risk developing countries is generally poor. Cases tend to present at an advanced stage and have an unfavourable outcome. This may induce a general sort of pessimism in the medical community about the capacity of the health system to impact on cancer prognosis even for sites that can be successfully treated. Such pessimism is not justified because even when resources are limited, at least 60% of breast cancer cases presenting with disease localised to the breast survive 5 years from diagnosis.8 A shift toward a more favourable distribution of stage would therefore have a significant impact on mortality.
Our study tests the feasibility and the effect of systematic screening of the population by clinical breast examination (CBE) on stage at presentation and ultimately on breast cancer mortality, in the urban area of Manila, the Philippines. This was done where the 2 local cancer registries had reported relatively high incidence rates (an age standardised rate of 48.7 per 100,000 in 1993–19979), and over 60% of the cases were at Stage III or VI at diagnosis.8 Our study was designed as a randomised trial and was planned to involve 5 rounds of screening for women in the intervention group, at intervals of 1–2 years.
We describe the study population, the intervention and its results in terms of breast cancer detection, cumulative incidence in 3 years of follow-up in the group examined and sensitivity and specificity of the examination in that setting. Due to a very low compliance with clinical follow-up, however, the intervention ceased after completion of the first round of examinations. We discuss reasons for the outcome and implications for the development of cancer control plans in developing countries.
MATERIAL AND METHODS
Our study began in 1995. It was designed as a randomised controlled trial of the efficacy of five annual clinical examinations of the breasts carried out by trained nurses/midwives, in reducing mortality from breast cancer. Women 35–64 years of age, resident in the 12 central, more urbanized municipalities of the National Capital Region of Manila were the target population. Young women were included because of the high proportion of cases below age 50.9 In 1990, the estimated size of the female population 35–64 years of age was about 340,000. The units of randomization were 202 health centres (HC) within the selected municipalities. These were randomly assigned to intervention or control arm by block randomization. Blocks were defined based on population size and a deprivation index indicating the presence of squatters areas within the administrative borders.
Identification of the eligible population
Lists of women resident in the 12 municipalities and who were included in the electoral rolls were obtained from the Department Of Health (DOH). Women were identified by family and first name, date of birth and complete address (street and administrative area that coincided with the area served by a health centre). Electoral rolls had last been updated during 1994–95.
In the first year (1995) a coordinating centre was set up. Nurses and midwives were recruited and trained in the technique of CBE using the MAMMACARE™10 programme tested previously in the Philippines, that uses silicone models of the breast for training purposes11 and has been shown to enhance performance of examiners in previous studies.12, 13, 14 Training was repeated for selected groups of examiners who missed or over-reported by 20% the lumps in the silicone models.
The first round of examinations took place in 1996–97 (24 months) and included 151,168 women. Eligible women resident in the intervention HC were contacted in 2 ways: at the HC among those women who were attending for a variety of reasons, and, for those who did not, by systematic home visits. Basic demographic characteristics (age, marital status, socio-economic level) of eligible women were recorded and the nature and purpose of the intervention were explained. Women were asked to give a signed assent to participation. They were interviewed and CBE was carried out by the trained examiners. The interview addressed socio-demographic variables and classical risk factors for breast cancer. Women were also instructed in the technique of breast self-examination (BSE) and provided with a leaflet in the local language explaining the purpose and methodology of BSE.
Women in whom abnormalities were detected and classified “positive” for a suspected lump were referred for diagnosis to special clinics established in 3 major hospitals and staffed by project personnel. The costs of transport to the clinic and of all medical procedures required to reach diagnosis were covered by the project. In addition, in the last year of the intervention period, a mobile team, comprising a doctor and a nurse and equipped to perform needle biopsies, carried out home visits for all positive women who had not reported to the referral centre, to obtain a final diagnosis. The standard diagnostic process consisted in a physical examination by a specialist doctor followed by fine needle aspiration or excision biopsy if indicated. Mammography was not available to the large majority of the women judged positive.
Women in the control area received no active intervention but were exposed to the general health education campaigns carried out by municipal authorities and voluntary bodies.
The aim of the follow-up of the intervention and control cohorts was to identify women who developed breast or other cancers, those who died and those who migrated outside our study area.
Two cancer registries, Manila-PCS and Rizal-DOH,9 covered the study population. Together they serve the whole metropolitan area and the surrounding more rural province of Rizal. The case-finding procedures of both registries were enhanced to reduce time to registration. New abstract forms including detailed information on extent of disease, tumour size, spread and nodal involvement were adopted. All registered cases of breast cancer were followed-up to 2001 to assess their vital status. Hospital records were first reviewed, and treating doctors and the cases' families were contacted to complement this information.
In a pilot study we tested the feasibility and reliability of active collection of death certificates (for all causes of death) for linkage with the study cohort. Mortality rates computed from the information thus obtained were however unrealistically low and cancer was over-represented. This activity was therefore abandoned, and only cancer deaths continued to be recorded, as part of the usual routine of the cancer registries. Breast cancer cases and deaths identified during the follow-up period were linked with the master file (interviews and CBE results) using a probabilistic record linkage software ‘RECLINK’1 Uncertain matches were sorted out by the registries' directors after consultation of paper documents.
All matched cases were retained as incident if date of diagnosis recorded by the registries occurred after date of interview/examination.
The main outcome measures are the number and cumulative incidence of breast cancers in the cohort of interviewed women. Because only one screening round was carried out, sensitivity, specificity and predictive value were calculated using as gold standard the incident cases identified by the registries in 2 years from screening examination, including those diagnosed at the time of testing. Additional parameters describing the performance of the intervention are presented as absolute and relative frequencies, means and their standard deviations (SD) and 95% confidence limits (CI). Because of the large numbers of subjects involved, statistical testing was avoided when comparing examined and refusers. Confidence limits of proportions are based on the exact binomial distribution. Differences between proportions adjusted for age were tested by the Mantel-Haenzel procedure.
The number of women interviewed and offered CBE was 151,168; 8% of these women refused to be examined. Table I shows some socio-demographic characteristics of the 2 groups as assessed at interview, women interviewed and examined, and women interviewed who refused CBE. The 2 groups were very similar in age, 44.8 ± 8.2 years and 44.7 ± 8.4 respectively, and were also of similar age at menarche, between 13.6 and 13.4 years, but differed for other variables. Refusers were 1 year older at their first full-term pregnancy and of higher socio-economic status than compliers (as shown by the proportion of women who attended college [18% vs. 12%] and the proportion illiterate [6% vs. 18%]), had a significantly greater income (means were Pesos 1,556/month/cohabitants vs. 10,800), were more often nulliparous (17% vs. 10%) and less likely to have had 5 or more children (25% vs. 33%).
Table I. Comparison of Characteristics of Women Examined and Interviewed Women Who Refused Examination
|Age in years (mean ± SD)||44.8 ± 8.2||44.7 ± 8.4|
|Attended college/university (%)||12.3||17.7|
|Monthly income (pesos) mean ± SD||5744 ± 5590||10806 ± 12023|
|Income/No. of cohabitants (pesos)||1556 ± 1713||2748 ± 3292|
|Mean age at menarche||13.6 ± 1.7||13.4 ± 1.5|
|Mean age at first full-term pregnancy||23.0 ± 4.5||24.1 ± 4.5|
|Women with 5 or more children (%)||32.6||25.3|
A total of 3,479 women (2.5% of those examined) were judged to have a lump and were referred to the project clinics (Table II). Of these, 1,293 (37.2%) received further investigation, and complete diagnostic follow-up was achieved for 1,220 women, 35% of those positive on screening. A total of 1,475 women (42.4%) actively refused further investigation, even with a home visit, and 784 of the non-compliers (22.6%) were not traced, and were either reported by the neighbours, or assumed, to have moved away or died.
Table II. Results of the Single Round of Screening, and Clinical outcome after 2 years of Follow-Up
|Number of women interviewed:||151,168|| |
|Number of women examined:||138,392 (91.5%)|| |
|Number positive on screening:||3,479 (2.5%)|| |
|Completed diagnostic follow-up||1,220||34|
| at project clinics||556||21|
| at project clinics||73||1|
| at project clinic after home visit||590||12|
|Refused or follow-up incomplete||1,475|
Among the 1,220 women who completed diagnostic follow-up, 34 malignant cancers were detected; the presence of a lump was not confirmed in 563 (46.1%) and 623 (51.1%) were diagnosed as having benign breast disease.
Because of the poor compliance with follow-up of screen positive women, even with home visits, the active intervention was discontinued after completion of the first screening round in December 1997.
Proportion positive by selected personal characteristics
Among examined women the positivity rate decreased constantly with age from 2.9% in women below 40 years to 1.5% in women aged 60 or more (Table III). The percentage of women detected positive was higher in those with <3 pregnancies (3.3% vs. 2.2%). The positivity rate was not consistently associated with the level of education and was higher in women reporting low income. It ranged from 1.1 to 6.0% in the 12 municipalities. Rates above the average were recorded in the more affluent areas of Makati (4.0%), Mandaluyong (6.0%) and Malabon (3.9%). The high rate of positives among women with missing information, in particular on education level, is an interviewer effect. We observed an inverse association between total number of interviews and examinations carried out per nurse and both their referral rate, and the rate of missing answers in their interviews. In other words, the less experienced the poorer the performance.
Table III. Proportion Positive Women per 10,000 Examined, by Selected Personal Characteristics
|Age|| || || |
| < 40||1,356||46,896||2.9|
|Education|| || || |
| Max primary||925||59,803||1.5|
| Max secondary||997||50,221||2.0|
|Full-term Pregnancies|| || || |
| < 3||1,302||39,777||3.3|
|Monthly income per No. of cohabitants|| || || |
After exclusion of cases whose incidence date preceded date of recruitment, there were 133 breast cancer cases, incident within 2 years of enrollment linked with records of women in the intervention cohort (Table IV). The cumulative incidence of breast cancer was 9.6/10,000. Eighty of these cases had been judged negative on CBE, corresponding to a cumulative incidence of 5.4 new cases per 10,000 screen-negative women.
Table IV. Breast Cancer Cases that Occurred among Screened Women, by stage at Diagnosis and Screening Outcome
|Age < 45 years|
| Screen-detected||5 (50)||–||5 (100)||–||5|
| Other casesa||3 (10)||7||16 (62)||3 (12)b||26|
|Age 45–54 years|
| Screen-detected||6 (35)||2||10||–||12|
| Other casesa||5 (12)||6||22 (61)||8 (22)b||36|
|Age 55+ years|
| Screen-detected||4 (60)||–||2 (100)||–||2|
| Other casesa||7 (24)||4||12 (55)||6 (27)b||22|
| Screen-detected||15 (44)||2 (11)||17 (90)||(0)c||19|
| Other casesa||15 (15)||17 (20)||50 (60)||17 (20)c||84|
Fifty-three cases were detected among the 3,479 women who were screen-positive (152.3/10,000), 38 of which were detected through the screening process among women reporting for the follow-up (diagnostic) examination (311.5/10,000). Fifteen (15) cases occurred among those women who did not complete the diagnostic process (66.4/10,000): of these, s6ix were refusers (40.6/10,000) and 9 were lost to follow-up (114.8/10,000). Thirty of 38 screen-positive cases were diagnosed within 12 months of the first examination, and 8 were diagnosed later. Of these late cases of cancer, 4 were found in women who were positive on CBE, but in whom the lesion was judged to be non-malignant at diagnostic follow up (2 by CBE carried out by the specialist doctor and 2 by fine needle biopsy).
Of 15 cases identified among refusers 11 occurred within a year and 4 later. The 80 cases diagnosed among screen-negative women were almost equally distributed between the 2 periods.
If we generously allow that every positive examination in a woman who eventually proved to have cancer (within 2 years of the test) is a true positive, then the test sensitivity for annual CBE was 53.2% and the positive predictive value was 1.2%. Specificity was virtually 100%. The test sensitivity decreased to 39.8% (53/133) for one CBE carried out in 2 years. Only 34 cases were actually diagnosed through the intervention reducing test sensitivity to 25.6% and positive predictive value to 1.0% (34/3,479).
Table V shows the distribution by age of clinical extent of the disease in the 34 cases diagnosed by the screening process and in the other 99 cases identified in the examined cohort 80 cases screen-negative, 15 lost to diagnostic follow-up or refusers and 4 diagnosed as having benign disease. None of the screen-detected cases had distant metastasis at presentation whereas 17% of the others had metastatic disease (p = 0.032 2-sided test of the difference between the 2 proportions). The proportion of advanced cases increased with increasing age from 12% below 45 years to 27% at age 55+ years (test for trend, p = 0.037). Of the staged cases classified benign originally, 2 of 3 presented with distant metastasis.
Table V. Breast Cancer Cases (BC) Identified in the Intervention Cohort (Total 138,392) in 2 Years of Follow-Up, by Screening Outcome and Time Since CBE
|All women examined||138,392||77||56||133||9.6|
|Screen-positives by screening outcome:|
| Refusers and lost:||2,259||11||4||15||66.4|
| Malignant breast cancer||34||30||4||34|| |
| No mass or benign breast disease||1,186||–||4||4||33.7|
Breast cancer is a growing problem in developing countries. Increases in incidence and mortality are widespread and often more marked in younger generations of women.1, 5, 6 In populations of South-East Asia increases range from 1 to 3.6%.15, 16, 17, 18 Mortality of cancer cases and breast cancer cases in particular is unnecessarily high.8 Known risk factors are linked to reproductive history and lifestyle and are hardly modifiable, rather they are likely to become more prevalent with economic development. In these circumstances, interest has tended to focus upon early diagnosis and treatment, as a means of reducing at least mortality.19, 20, 21
The efficacy of breast self-examination has been formally tested in a randomised trial in Shanghai, China.22 No significant reduction of breast cancer mortality in the intervention group was detected after 10 years of follow-up and the distributions of stage at diagnosis in screen and control groups were very similar. The small size of the lesions diagnosed in the control subjects in this trial (47 % ≤2 cm diameter, 48% node negative), however, suggests a high level of health-awareness in this special subset of the Shanghai population, and may give little scope for improvement in outcome through early detection by BSE.
Clinical breast examination carried out by a trained examiner has many attractions. In programmes where it is combined with mammography, CBE finds fewer lesions but does detect some that had been missed by mammography. In general the differential is less for younger women.23 In the CNBSS II trial of women ages 50–59, there was no significant difference in the efficacy of CBE alone compared to CBE combined with mammography.24 CBE has been introduced as a single screening modality in Japan. There is some suggestion that, where coverage by such screening is high, breast cancer mortality rates have declined more than in other areas,25 although a case-control study was inconclusive.26 Manpower requirements for a screening programme based on CBE would be expensive but in many developing countries these are generally easier to mobilize, than the technology required for mammography. Based on these arguments, it has even been suggested that CBE would be a more cost-effective alternative to screening women at high risk, in low-income countries.27
The trial in Manila was designed to assess whether a meaningful reduction in mortality from breast cancer could be achieved in a developing country using an inexpensive procedure and locally available resources, that is physical examination of the breast carried out by nurses and midwives. The mortality reduction that was aimed for, among the women actually screened and followed-up, was 25%, a smaller effect than that of mammography that had been demonstrated in randomised controlled trials (RCT) settings to reduce mortality by about 30–35% among screened women, and probably the minimum mortality reduction that would be worthwhile in any future programme. The Manila area was selected for the trial for several reasons, the relatively high incidence of breast cancer; the availability of treatment facilities (surgery, radiotherapy and systemic therapy are provided by both the public and private service); and the possibility of recruiting a large number of qualified nurses to act as examiners.
The unexpected result that jeopardised the whole intervention was the unforeseen reticence of women found with abnormalities and informed of the implications to their life, to pursue diagnosis and treatment. These problems had in fact been noted during a pilot phase. In the main study they were addressed from the beginning by provision of free transport and consultation. This tactic was not sufficient, however, and a programme of diagnostic home visits was introduced. Even this failed to raise compliance with diagnostic follow-up beyond 35%. The reasons can only be speculative at this stage but deserve ad hoc studies. One may think that women did not understand the implications of undergoing CBE. This is a relatively educated population, however, and highly exposed to media messages. The observation that women who refused to enter the trial were of higher socio-economical level reinforces the notion that is not a lack of information the cause of non-action. Alternative hypothesis are also possible. Lack of trust in the health system and in one's chances to be cured may discourage action. Such attitude is not in contrast with accepting screening examination. In a recent survey in the United States, 60% of interviewed people reported they would wish to be examined for a cancer for which there is no hope of cure.28 It is known that women attend for breast cancer screening in anticipation of a negative finding,29 and screening is not a stressful procedure for those with a negative mammography.30, 31 Receipt of an abnormal result, however, is associated with considerable psychiatric morbidity,32, 33, 34 and this may have played a role in the low level of compliance. The decision not to undergo investigation was a positive one in most instances, and not related to logistical or financial barriers. It is known that patient's decision making is not always apparently rational. Misinformation, denial, overconfidence, distrust and confusion may all play a role.35
The second major limitation highlighted by our study is the modest sensitivity of the screening test in the setting in which it was applied. Sensitivity and specificity of CBE have been measured in randomised trial of mammography and screening programmes, relative to new cases diagnosed within 12 months detected by either CBE or mammography and including interval cases. In these conditions the average sensitivity and specificity were 54% and 94% as estimated in the meta-analysis by Barton.37 The sensitivity was higher (68%) in the Canadian trial in the control arm who received only CBE.38 Sensitivity estimates are difficult to compare due to varying definitions of the reference gold standard in different studies. In our study the reference set included only cases that surfaced clinically within 12 or 24 months, the majority of which were relatively advanced (Table V). We would expect therefore an even lower estimate of the sensitivity had the cohort also been screened by mammography. We also showed that less experienced nurses performed more poorly. Though predictable this observation reinforces the need to form personnel devoted to this activity.
Our results reflect what might realistically be expected from CBE as a screening modality when applied by nursing personnel formally trained in the procedure but necessarily inexperienced. Clearly, if CBE is to be at all useful, a much greater effort in training and quality control of performance than was possible in the Manila trial will be required. But it is unlikely that this can be obtained from staff in primary health centres normally dealing with more common diseases. One could envisage a new professional profile of health workers who specialise in the diagnosis and follow-up of cancer and are made available regularly in health centres. Despite the modest performance observed of the clinical procedure we could document an improvement of stage at presentation among examined women. This justifies pursuing further studies based on CBE.
We show that in the urban population of Manila serious logistic as well as psychological barriers to seeking medical attention for breast cancer persist. An occasional contact with unknown health workers has a minimal impact on health-beliefs and behaviours. In addition, the sensitivity of CBE carried out by trained but inexperienced personnel is low. Yet, early diagnosis remains a high priority to improve the lamentable stage distribution that leads to premature death of a large number of cases.
Cancer awareness needs to be reinforced taking inspiration from the experience of high-risk countries36 but bearing in mind the specific context where other diseases will continue to be of greater importance and governmental expenditure in health care is unlikely to increase substantially. Alternative organizational settings need to be devised and tested. Access to early diagnosis could be improved for instance by promoting detection of BC among health operators in first level primary care services that interact with the population on a daily basis. Cancer centres should be created to provide experienced personnel and appropriate diagnosis and treatment. Rotating breast clinics could then be organised on a regular bases within health centres for primary care, to teach and encourage BSE and to provide opportunistic CBE. The regular presence of specialised personnel could also help to raise awareness and trust.
In recent years research on means to improve cancer control when resources are limited has focused on the evaluation of low-cost screening procedures and our study is an example in this direction. The outcome of the Manila trial is a reminder however, that culturally-related health beliefs are a major obstacle to early diagnosis and that awareness and access need to be addressed in first place.
We would like to thank Mrs. A. Bautista, Manila and Mr. N. Mitton, Lyon for their devoted assistance in data management and the staff of the WHO Regional Office for the Western Pacific for their continuing support in running the project.
RECLINK is a record linkage software developed at unit of Descriptive Epidemiology, International Agency for Research on Cancer, Lyon. The software performs probabilistic linkage between records from different sources using selected personal identifiers (names, date of birth, sex, address).