MAMMOGRAPHY FOR YOUNGER WOMEN

A REVIEW of the EVIDENCE

AUTHOR

IAN D CAMPBELL

MB ChB FRACS

      Breast Surgeon, Clinical Director,

    BreastCare Centre, Waikato Hospital

Senior Lecturer, Waikato Academic Division, Auckland School of Medicine

    Correspondence to: Mr I D Campbell

          Dept of Surgery, Health Waikato PB 3200, Hamilton

                                   MAMMOGRAPHY FOR YOUNGER WOMEN

                                   - A DIFFERENT LOOK AT THE EVIDENCE -

INTRODUCTION:

The Ministry of Health, epidemiologists,  clinicians and the vast majority of women's organisations in New Zealand are uniformly in agreement that screening mammography is very much indicated for women over 50 years of age.  The Ministry of Health Policy on Mammography for Younger Women has to date been opposed to screening of 40-49 year old women. ¹

The issue of screening for younger women 40-49 years of age has been a highly controversial subject world wide and not the least in New Zealand, as evidenced by the multiple letters to the New Zealand Medical Journal on this issue.^2-11  Views are polarised between two extremes ;

(1)  the view that screening mammography in this age group has no justification  (2)  the view that the current evidence provides very strong justification for nation-wide annual screening mammography of 40-49 year olds.

The issue is especially important in New Zealand where we have the dubious distinction of the highest death rate in the world from breast cancer in women under 50 years of age.¹²

A recently published meta-analysis of the results of the randomised trials of screening mammography for younger women has shown a statistically significant 23% reduction in deaths from breast cancer by offering screening and resolves much of the controversy for many epidemiologists and clinicians.¹³

The Latest Evidence

Eight randomised controlled trials ( RCT's) of routine mammographic screening for breast cancer have been performed.  The meta-analysis¹³ combines the data from these trials with periods ranging from 7-18 years of follow up.   Over 170,000 women 40-49 years are included and the data represent over 1,800,000  women - years of follow up.  The data for each of the individual studies is summarised in Table 1.  The data has been analysed both with and without inclusion of the Canadian National Breast Screening Study¹⁴(CNBSS-1) as with previous meta-analyses. ^15,16

Excluding CNBSS-1 data yields a relative  risk of death from breast cancer of 0.77 for the group offered screening versus the control group not offered screening (95% confidence intervals  0.62 - 0.95 ) - a statistically significant 23% reduction in mortality. Inclusion of CNBSS-1 yields a relative risk of death from breast cancer of 0.84 (95% confidence intervals 0.69 - 1.02 ),  a non significant 16% reduction in mortality by offering screening.   These are remarkable results given the numerous deficiencies(by modern screening programme standards) in design and execution of every one of these studies.¹⁷

Of the seven RCT's excluding CNBSS-1, the results are fairly consistent.  Five of the seven studies show a trend towards a mortality reduction for screening.  Of the two that show no trend, Ostergotland had a disproportionally high number of deaths in women who were randomised to screening but refused or had cancers found prior to undergoing mammography.  Without these deaths the Ostergotland Study shows a relative risk for breast cancer death of 0.81 in favour of regular screening.¹⁸

The Stockholm study, the second study to show no trend, screened the control group after only two screening rounds and had the longest screen interval at 28 months between mammograms.  The results therefore represent a 3 - 5 year difference in offering  screening between the intervention and control group and it is not surprising that little difference has emerged.

Both the Malmo and Edinburgh Trials began screening at age 45 so it could be argued that the RCT evidence remains less certain for 40-44 year old women.  Unfortunately there are barely enough women in the RCT's for the whole group to have the statistical power to demonstrate an effect¹⁹ and certainly not enough women or events (breast cancer deaths) in the 40-44 year old age group to enable meaningful analysis of this subgroup alone. Indeed if the women with cancer in the control group experience an 80% 5 year survival, it has been estimated²⁰ that more than 250,000 women would be needed in each arm of a RCT to prove a benefit for younger women at an earlier follow-up than the current RCT's. It is doubtful these numbers will ever be achieved in a further screening mammography trial.

Problems with the Canadian Studies^{14, 21}

There are a number of reasons for excluding the CNBSS-1 data.

(1)       CNBSS-1 utilised clinical breast examination prior to randomisation into study and control groups, and the control group received ongoing screening by physical examination.   The study therefore compares mammographic screening with regular clinical examination.  In the other RCT's the control groups are not offered any form of screening.

(2)       The study recruited volunteers rather than utilising a defined population group.

(3)       The randomisation process in CNBSS-1 was non blinded and somehow a  significantly greater number of women²⁰ with advanced breast cancers at the outset of the study were "randomised" to the mammographic screening group compared to the control group.

(4)       The quality of the mammography in this study has been heavily criticised.

From the first two points above, it is clear that the design of the CNBSS-1 is different from the other RCT's.   Most randomised studies blind the investigator and the patient to the  randomisation process specifically to avoid any possibility of selection bias.  For the CNBSS's the investigators were not blinded and of women with particularly poor prognosis breast cancers (four or more involved axillary lymph nodes) 19 were assigned to the screening group and five to the control group.  Of the 19 assigned to the screening group, 17 had palpable breast cancers.   The  magnitude of this difference suggests that it is  unlikely that this difference has arisen by chance and has raised the unproven doubt that the pre-randomisation clinical examination has influenced the assignment of women to the screening or control groups.  This suggestion has been supported by the fact that women with breast cancer in the control group have an  " unprecedented survival of over 90% at five years "²⁰ although contemporary data show women aged 40-49 years in Canada and the USA have an average 5 year survival rate of less than 80%. 

In fact for the CNBSS-1 to have shown a benefit for screening mammography would have required virtually no breast cancer deaths among the screened women.  The randomisation problem achieves particular importance when it is appreciated that the trend towards an increased death rate in the screening population (38 deaths in this group vs 28 deaths in the control group) is more than entirely accounted for by the excess of advanced breast cancers assigned to the screening group at the start of the study.

Problems with the way the CNBSS's were set up have impacted heavily on the quality of mammography.

(1)       In many centres existing mammography equipment was used although in a number of centres this was far from state of the art even at the time of commencement of study in 1980. (1972 equipment used in one centre).²²

            External reviews of technical quality were carried out^{23, 24} and internal audit was carried out by a reference radiologist.²⁵

(2)       Neither the radiographers who took the mammograms nor the radiologists who interpreted the films had received special training.²²

(3)               Mediolateral oblique projections were not used for the first 5 years of the 8 year study.  A considerable area of breast tissue is not visualised on straight mediolateral views. 

This process is in stark contrast with the Quality Standards in each of these areas required of BreastScreen Aotoeroa and indeed any good modern mammography screening programme.

Over 50% of the mammograms obtained during the first four years of the trial, were technically inadequate and concern was expressed as to the sharpness, contrast and overall quality of the mammograms.

As a result of the above factors, 42% of the cancers that were missed at screening and became palpable in the interval between screenings were visible on the previous mammogram and had been missed by the interpreting radiologist.  Of the cancers diagnosed in the screening programme, 25% could be identified one year earlier having been missed by the interpreting radiologist.  The number of cancers which have never shown on mammograms because of poor quality mammography is not known.²⁰

In the HIP Study, decreased breast cancer mortality among women aged 40-49 years on entry only became apparent seven years after entry.  Every subsequent randomised trial has confirmed that a greater length of follow up is necessary for demonstration of benefit for younger women than for women aged 50 years and over.²⁶ The RCT's also show diverging mortality curves with time for the screened vs control populations, ^27,28 in spite of the fact that screening was either stopped after a period of years or the control group screened (after 3 - 6 years).

When first  published, the Canadian data have only seven years follow-up and a previous meta analysis of all the RCT's at seven years follow up shows no benefit for mammography in younger women at this time. ¹⁵

Reasons why the RCT's may underestimate benefit.

Compliance and Contamination :

To avoid selection bias randomised trials compare a study group of women offered screening with a control group not offered screening.  The groups are analysed on an intention  to screen basis  ie; according to which group the women were randomised and not according to whether  screening actually took place.  Some 25% (10-39%¹³) of women offered screening in the randomised studies never had it.  Some 25% of women not offered screening had mammography at some stage during the course of these studies.  The exact percentages vary from study to study.  The result of this contamination of the control group and dilution of the screening group is that randomised trials underestimate the true effect of actually screening women vs not screening women.  What women (and clinicians and other concerned individuals) want to know is :

 " What is the benefit to me if I actually have a mammogram ? "

Statistical methods have been developed to account for the problems of compliance and contamination in randomised trials.   Utilising these it has been estimated that the  true mortality benefit of having a mammogram is some 40-50% higher than the randomised trials indicate.²⁹

Improvement in mammography :

These studies represent the learning curve  in the development of mammographic interpretation and technique and a period of time during which there were major advances in equipment technology.  Today (20-25 years later) we have much better equipment and knowledge; Radiologists and mammographers undergo detailed training on mammography interpretation and technique, we utilise grid techniques, better film and film processors and two-view mammography - all of which improve sensitivity for small cancers.³⁰  It is quite unscientific to discount this improvement³¹ based on a lack of improvement in detection rate for small breast cancers from 1977 to 1982 compared with a period 10 years later from 1989-92 (data from the two-counties study).³²  Many other factors besides quality of mammography may influence detection rates of breast cancer over a ten year period.  It is exactly because of other potential biases that studies using historical controls as evidence of effect are generally regarded as poor quality evidence. It could just as well be argued that the improved small cancer detection rates seen in the New Zealand Pilot Programme³³ compared with the RCT data, is proof of the advances in knowledge, equipment and technique over the last 15 years.  This argument may be true but many other factors may also be playing a role.

The difference in the quality of mammograms taken today compared with the mammograms of 15 years ago is very obvious even to a surgeon such as myself,.

Better evidence is available from a number of sources.³⁴  Indeed the two-counties study³² provides other evidence supporting improved sensitivity of modern equipment and techniques - a lower rate of interval  cancers (cancers arising in the interval between screening rounds and not detected at the previous screen) in the more recent series and a higher detection rate of DCIS.

For the Edinburgh Trial³⁵ both improved cancer detection rates and decreased interval cancer rates were noted in later rounds concomitant with a significant improvement in mammographic technique and attributed by the investigators to improved  technical quality.

Even the CNBSS'S have demonstrated this trend.  Baines et al²³ showed that 51% of mammograms performed in 1980-82 were acceptable rising steadily to 85% acceptable for 1985-87.  Coincident with this improvement, interval cancer rates dropped progressively over years 1-4 from 60% to 20% and the  non palpable cancer detection rate doubled for women aged 40-49.³⁶ 

Even with modern day equipment small improvements such as use of higher density film³⁷ may improve the detection rate of small cancers by as much as 50%.   

Quality improvements such as single versus two-view mammography and double reading of films can and have been rigorously tested. Two-view mammography increases the sensitivity of mammography for detection of cancer by 3-11%^38-41  and this improvement is especially of value for women with dense breasts (more common in younger women) and for detection of small cancers - the ones we need to detect to save lives.  Single view mammography  only was used in over half the RCT's rather than two-view. (Table 1)

Double reading by two independent radiologists has also been studied and  improves the sensitivity of mammography⁴²  by up to 15%.⁴³  This is now standard practice in BSA, but was not routine in the original randomised trials.

Screen Intervals :

The interval  between screens was up to 2½ years in a number of the RCT’s.

(Table 1) As the growth rate for breast cancers in younger women is considerably faster on average than that for older women^18,44 screening at a one year interval for younger women may improve the mortality benefits by 50-100%^18.  Further support for a better screening interval is apparent from the RCT's.  The 2 studies with the greatest trend towards a benefit ( Malmo and Gothenberg) are also the two studies (apart from CNBSS-1 for reasons above and HIP because this was conducted in the 1960’s)  with the shortest screening interval¹³. 

Screening period :

The control group in four of the five Swedish studies were all offered screening some 4 to 6 years after the study commenced.   The results for these studies therefore represent an approximate 5 year period of offering screening vs not offering screening²⁸.   Had the control group not been offered screening we would expect that with time any differences between the control and study groups would have been considerably more dramatic.

Age group analysis :

The full effects of screening by age remain far from completely sorted out.  The data have thus far been arbitrarily analysed in the age brackets 40-49 years and over 50 years.   Women's breasts do not suddenly change at 50 years of age Although they do gradually become less dense with age only 26% of women 40-49 years of age have mammographically dense glandular tissue³².  We should be looking especially closely at the 45-49 year old age group because these women have a breast cancer incidence very similar to the 50-54 group⁴⁵, and the latest NSW, reveal more actual cancers registered for 45-49 year old women than for the 50-54 age group (Prof J Forbes - personal communication).  Furthermore, saving the life of a woman between 45 and 49 years of age contributes an average 10 extra women-years of life compared with saving the life of a woman between 50 -64 years of age.

The Ministry of Health have previously attempted to dismiss the potential benefits of screening younger women by pointing out that only a minority of breast cancer deaths occur in women under 50 years of age.  This is a gross misuse of statistics.  It ignores the point that in order to save the lives of many women who die of breast cancer in their 50’s and 60’s these breast cancers need to be diagnosed many years earlier while these women are in their 40’s.  At this time these breast cancers are at a much earlier stage or even at a precancerous stage such as ductal carcinoma in situ (DCIS).  If DCIS can be removed, invasive cancer may be prevented which otherwise might take many years before leading to death from cancer if left undiagnosed and untreated.

It has also been argued that because it takes a longer time for the benefits of mammography to emerge for younger women than for women 50 and over, the benefits may be due to younger women turning 50 during the course of the trials and benefiting once they reach this age.  In fact, of the breast cancers diagnosed by screening younger women in the randomised trials, 75% were diagnosed in women under 50 years of age (personal communication – Stephen Duffy, Principle Statistician for the Two Counties Study and Swedish Screening Trials).  If the benefit was solely due to screening women in their 50’s (ie there was no benefit from detecting the 75% of cancers found while women were in their 40’s) we therefore might expect the screening trials for younger women to be only 25% as effective as for older women.  Actually the benefits seen are very similar in magnitude to those for women in their 50’s and 60’s.  Other factors must explain the difference in time to benefit. 

Two possible explanations are:  that women in their 40’s have the best outcome, when

matched for stage, of any age group.  It therefore takes longer for screening to produce a benefit, because nonscreened women of this age group who get breast cancer live longer on average than older women.  Another possible explanation is that a higher proportion of breast cancers are diagnosed at the DCIS stage.  DCIS, if it progresses to invasive cancer and death, takes on average more years to do so than a cancer that is already invasive .

Morbidity :

The RCT's examine only the mortality issue and ignore morbidity.  While mortality benefits are the ultimate goal, morbidity from breast cancer treatment is also an important goal.  There is no doubt that the average size of cancer detected by screening mammography is smaller in all age groups³².  Small breast cancers are much easier to treat by breast conserving surgery with a good cosmetic result and, pending the result of current trials, are also likely to be amenable to sentinel node based management of the axilla.  This technique removes the need for axillary dissection with it’s attendant risk of arm lymphoedema for many women with small breast cancers.  The technique is not appropriate for women with larger breast cancers.  Many women will also not require treatment with chemotherapy if diagnosed at an earlier stage.  These factors represent very significant advantages to these women.

The Adverse effects of Screening Mammography

The adverse effects can be summarised as: the cost; the psychological impact; the discomfort and inconvenience of the procedure; the recall of normal women; breast biopsies of some women with benign problems; possible risks from radiation, and; the false reassurance of a false negative result.

Cost: The cost of screening 40-49 year old women has been estimated at $AUS14,656⁴⁶ - $US26,200⁴⁷ per year of life saved - very much in the mid range of health care costs for current routine health care interventions  ( eg; costs per year of life saved for drug treatment of moderate to severe hypertension: $US11 -73,000⁴⁸; coronary artery bypass surgery: $US7,000; haemodialysis for chronic renal failure: $US35,000⁴⁹. These estimates of cost benefit may overestimate the cost compared with the marginal increase in cost of extending the age group of an existing programme for 50 - 64 year old women.   In this situation much of the capital outlay for equipment and infrastructure has already been spent and extending the age range primarily increases the costs of consumables and staff with a much lesser impact on requirement for additional equipment, buildings etc.

The psychological impact of breast cancer screening has yet to be determined.  However studies have shown that women are less anxious about their risk of death from breast cancer having gone through the screening process than women who have never been invited to screening mammography⁵⁰. Certainly the process of screening and of recall and biopsy for some women may be stressful and highlights the need for any screening service to be properly run with rapid assessment, information and support for women concerned so that such stresses can be minimised.

Pain and discomfort have been assessed in the Waikato Pilot Programme A randomised survey of almost 1,000 screening participants ( unpublished data )  has shown that some 6% of women 50-64 years of age find mammography painful and  a further 10% find it very uncomfortable.  Similar results have been shown in a British study where most women rated mammography similar to having their blood pressure taken and less uncomfortable than a venepuncture⁵¹. These are the results of dedicated mammography units and staff and again highlight the need for such units, along with public education to counter the widespread misconception that most women find mammography painful.

Recall of normal women: Some six normal women will be recalled for more detailed assessment including further mammographic workup ± ultrasound ± needle biopsy for each 100 women screened, and 1-2 benign biopsies (most of these needle biopsies rather than open surgical procedures) will be performed for each cancer detected^52,53.  In other words, while the number of cancers detected per thousand younger women screened is lower at 3-4/1,000, the recall rate and the positive predictive value for surgical intervention are well within established guidelines for older women⁵³.  These are results from  modern quality breast clinics in the Australian and U.K. settings and are considerably more specific than figures quoted from the North American environment⁵⁴.  These are first screening round figures and drop for subsequent rounds.

Radiation risk: The radiation dose from two view mammography is approximately 0.25 rads.  There is no known risk for induction of breast cancer from this dose but certainly much higher doses of 100 rads or more as in Hiroshima victims and women undergoing multiple chest fluoroscopies in the 1920's and 30's for TB have been shown to cause breast cancer after a latent period of 4-10 years.  Extrapolating risks from these much higher doses it has been predicted that five breast cancer deaths may occur per million younger women receiving a single screen²⁶.  If these possible radiation induced cancers were detected by screening at an early stage this theoretical death rate might be decreased by 20-40%.  On the other hand, a single screen for women in this age group would be expected to save between 150 and 300 deaths per million women assuming  mortality reductions of 20-40%.  

False reassurance: For all the potential benefits of screening mammography in any age group, mammography is an imperfect tool.  Some 4-10% of breast cancers in older women and maybe as many as 20% of breast cancers in younger women will not show up on mammography.  This needs to be made very clear to women and their GP's to avoid the false reassurance of a false negative result. It is therefore important that women continue to practise regular breast awareness see their family practitioners for any new breast symptom.  Breast self examination for those who practise regularly, does reduce average size and stage of detection of breast cancer (Prof  R Blamey, personal communication of the Nottingham Trial Results).

Discussion:

Many countries, including Australia, parts of Sweden, Iceland and Finland already have national screening programmes under way including 40-49 year old women. After careful consideration of the evidence, many other reputable groups including that US National Cancer Institute and American Cancer Society and the 2002 review by the United States Preventative Services Task Force⁵⁹, all recommend mammography screening for women in their 40’s. 

To tell women 40-49 years of age that screening is no good but on turning 50 years of age it should suddenly become highly effective is a confusing message.  Not only is nothing being achieved for the younger women but the message may have a negative impact on compliance for older women where we are all agreed, breast cancer screening is highly worthwhile.   A broader age range provides a more consistent and easier to administer policy.

This review has concentrated predominately on RCT evidence to avoid controversies and biases over non-randomised studies.  It should be noted however, that there is a large body of indirect evidence that is also supportive for screening younger women.  There is no doubt screening mammography detects breast cancers at an earlier stage in younger women, ^32,56,57  and that survival rates for earlier stage breast cancers and screen detected breast cancers in younger women are just as good if not better than for older women (these data are however subject to several possible biases: selection bias, lead and length time bias).

If a National Programme for younger women gets under way, it is essential that it is of the highest quality.  The CNBSS's^14,21 are useful for one thing: these studies clearly demonstrate that poor quality mammography will at best achieve nothing, and at worst may cause potential harm. 

If a National Programme does not embrace a wider age range than 50-64 years of age, private radiology clinics certainly will and are promoting screening in this age range, and are currently not subject to any adequate quality assurance for breast cancer screening. To provide high quality population based screening requires a national quality controlled approach together with adequate funding.

CONCLUSION:

The controversy over screening mammography for younger women is now to a considerable extent resolved. It is clear that from the metaanalyses of RCT's^13,28 that offering a population screening mammography reduces the death rate from breast cancer for all ages at least from 45-70 years and probably from 40-75 years.^13,18  With modern day screening equipment and knowledge, appropriate quality standards, two view mammography, an appropriate screen interval and for those women who attend, we will achieve a considerably greater reduction in breast cancer mortality than the RCT's indicate.  As breast cancer is the most common cancer in women and breast cancers diagnosed in this age range account for many of the years of suffering and life lost from breast cancer it is imperative that the National Screening Programme in New Zealand is made available to women 40-49 years of age.

TABLE 1:  SUMMARY OF RANDOMISED CONTROLLED TRIAL RESULTS FOR WOMEN AGED 40-49 YEARS

                                                                                                                                                 

Screening Study  (Dates)	Age of entry	Regimen	Frequency	Yrs F/U	Women years of F/U		Breast cancer deaths		RR (95% CI)
					Invited	Control	Invited	Control

References

1.         The Government's policy on mammograms for younger women.  Prescriber Update 1994; 5 : 2-5.  Ed. Margaret Owen.  Published by the Therapeutics Section, Ministry of Health,  P O Box 5013, Wellington, NZ.

2.            Campbell I, Dowle C, Harman J et al.  Mammography for younger women (letter) NZ Med J 1994 ; 107 : 137.

3.            Vallance S.  Mammography for younger women (letter) NZ Med J 1994; 107 : 137.

4.            Williams L.  Mammography screening under 50 (letter) NZ Med J 1994; 107 : 207.

5.         Coney S.  Mammography screening under the age of 50 (letter) NZ Med J 1994 107 : 230-1.

6.         Blue J.  Mammography Screening (letter)  NZ Med J 1994; July : 276-7.

7.         Cox B.  Mammography screening under the age of 50 (letter) NZ Med J 1994; 107 : 340.

8.         Coney S.  Mammography screening (letter) NZ Med J 1994 107 : 361 - 2.

9.            Robinson B.  Mammography screening (letter)  NZ Med J 1994;  107 : 362.

10.            Hitchcock GC.  Breast screening under 50 (letter)  NZ Med J 1994; 107 : 411.

11.            Elwood JM.  Breast cancer screening in younger women ; the need for evidence based medicine  NZ Med J 1995; 1-8 : 239-241.

12.       Peto R, Lopez A D, Boreham J,  et al.   Mortality from smoking in developed countries 1950-2000.  Oxford: Oxford University Press 1994.

13.       Smart C R, Hendrick R E, Rutledge J H  and Smith R A.  Benefit of Mammography Screening in Women ages 40-49 years.  Cancer 1995; 75; 1619 -26.

14.       Miller A,^(a) Baines C J, To T, Wall C.  Canadian Breast Screening Study: 1. Breast cancer detection and death rates among women aged 40 to 49 years.  Canad Med Assoc J 1992; 147 : 1459-76.

15.            Elwood J M,  Cox B, Richardson A K.  The effectiveness of breast cancer screening by mammography in younger women.  J Curr Clin Trials (on line) 1993; 1 : 32 - 227.

16.       Wald A J, Chamberlain J, and Hackshaw A, on behalf of the Evaluation Committee. European Society of Mastology Consensus Conference on Breast Cancer Screening, Paris, 4-5 February 1993 : Report of the Evaluation committee.  Clinical Oncology 1994; 6 : 261-268.

17.            Sickles EA, Kopans DB.  Deficiencies in the analysis of breast cancer screening data (Editorial). J Natl Cancer Inst. 1993; 1621-4.

18.       Tabar L, Fargerberg G, Chen H, et el.  Efficacy of Breast Cancer Screening by age.  Cancer 1995; 75 : 2507 -17.

19.            Sondik E J.  Screening for breast cancer: response. J Natl Cancer Inst  1994;86:560-561.

20.            Kopans D B and Feig S A.  The Canadian National Breast Cancer Screening Study: A critical review.  AJR 1993; 161: 755-760.

21.       Miller A^(b), Baines C J, To T, Wall C.  Canadian Breast Screening Study: 2. Breast cancer detection and death rates among women aged 50 to 59 years.  Canad Med Assoc J 1992; 147 : 1477-88.

22.            Burhenne L J W  & Burhenne H J.  The Canadian Breast Screening Study :  A Canadian critique.  AJR 1993 ; 161: 761 - 763.

23.            Baines CJ, Miller AB, Kopans DB, et el Canadian National Breast Screening Study : assessment of technical quality with external review.   AJR 1990 ; 155 : 743 - 747.

24.            Kopans DB.  The Canadian Screening Programme : a different perspective

            (commentary).   AJR  1990 : 155: 748 - 749.

25.            Baines CJ, McFarlane DV, Miller AB. The role of the reference radiologist: estimates of interobserver agreement and potential delay in cancer detection in the National Breast Screening Study.   Invest Radiol 1990; 25: 971 - 976.

26.       Feig S A.   Mammographic Screening of Women aged 40-49 years: is it justified?  Obstetrics and Gynaecology Clinics of North America 1994;  21  (4)  : 587 - 606.

27.            Shapiro S, Venet W, Strax P, et al. Periodic Screening for Breast cancer ; the Health Insurance Plan Project and its sequelae, 1963 - 1986.  Baltimore, John Hopkins University Press, 1988.

28.            Nystrom L, Rutqvist LE, Wall S et al.  Breast cancer screening with mammography: An overview of the Swedish randomized trials.  Lancet 1993; 341: 973-78.

29.            Glasziou P P.  Meta-analysis adjusting for compliance: The example of screening for breast cancer.  J Clin. Epidemiol. 1992; 45: 1251-1256.

30.       Boyd N F, Jong R A, Yaffe M J et al.  A critical appraisal of the Canadian National Breast Cancer Screening Study.  Radiology 1993; 189: 661-663.

31.            Elwood J M.  Breast Cancer Screening in younger women; the need for evidence based medicine.  NZ Med J 1995; 108 : 239 -241.

32.       Tabar L, Duffy S W, Burhenne L W.  New Swedish Breast Cancer Detection Results for Women Aged 40-49.  Cancer 1993; 72: 1437-1448.

33.            Elwood J M.  Breast Cancer screening in New Zealand: a decision point.  NZ Med J 1994; 107 : 348-9).

34.       Feig S A. Estimation of currently attainable benefit from mammographic screening of women aged 40-49 years.   Cancer 1995;  75:  2412-2419.

35.            Roberts M M, Alexander F E, Anderson T J, Chetty U, Donnan P T, Forrest P, et al.  Edinburgh trial of screening for breast cancer:  mortality at seven years.  Lancet  1990; 335: 241-6.

36.            Moskowitz M. Author of Canadian Breast Cancer Study retracts warnings (response).   J Natl Cancer Inst 1992;84 :1368-71.

37.       Young K C, Wallis M G, Rasdale M L.  Mammographic film density and detection of small breast cancers.  Clin Radiol 1994; 49:461-5).

38.       Muir B B, Kirkpatrick A E, Roberts M M, Duffy S W.  Oblique-view mammography: adequacy for screening.  Radiology 1984; 151:39-41.

39.            Anderson I, Hildell J, Muhlow A, Pettersson H.  Number of projections in mammography: influence on detection of breast disease.  AJR201978; 130:349-51.

40.            Thurfjell G, Taube A, Tabar L.  One-versus two-view mammography screening, a prospective population based study.  Acta Radioligica 1994; 35:340-4.

41.            Sickles E A, Weber W N, Galvin H B,  et al.  Baseline screening mammography: one vs two views per breast.  AJR 1986; 147:1149-53.

42.       NHS Breast Screening Programme.  "Quality Assurance Guidelines for Radiologists" Screen Publications - Dr Muir Gray [Ed.] National Breast Cancer Screening Education Programme, Dept of Community Medicine, Radcliffe Infirmary, Oxford.

43.            Thurfjell E L, Lernevall K A, Taube A A S.  Benefit of independent double reading in a population-based mammography screening program.  Radiology 1994;  191:241-244.