Comparative evaluation of digital mammography and film mammography: systematic review and meta-analysis

ABSTRACT CONTEXT AND OBJECTIVE: Mammography is the best method for breast-cancer screening and is capable of reducing mortality rates. Studies that have assessed the clinical impact of mammography have been carried out using film mammography. Digital mammography has been proposed as a substitute for film mammography given the benefits inherent to digital technology. The aim of this study was to compare the performance of digital and film mammography. DESIGN: Systematic review and meta-analysis. METHOD: The Medline, Scopus, Embase and Lilacs databases were searched looking for paired studies, cohorts and randomized controlled trials published up to 2009 that compared the performance of digital and film mammography, with regard to cancer detection, recall rates and tumor characteristics. The reference lists of included studies were checked for any relevant citations. RESULTS: A total of 11 studies involving 190,322 digital and 638,348 film mammography images were included. The cancer detection rates were significantly higher for digital mammography than for film mammography (risk relative, RR = 1.17; 95% confidence interval, CI = 1.06-1.29; I² = 19%). The advantage of digital mammography seemed greatest among patients between 50 and 60 years of age. There were no significant differences between the two methods regarding patient recall rates or the characteristics of the tumors detected. CONCLUSION: The cancer detection rates using digital mammography are slightly higher than the rates using film mammography. There are no significant differences in recall rates between film and digital mammography. The characteristics of the tumors are similar in patients undergoing the two methods.


INTRODUCTION
Breast cancer is the second most frequent malignant neoplasm among the female population. 1 In Brazil, it has been estimated that there are 50.71cases for every 100,000 women, second only to non-melanoma skin neoplasms.The specific mortality rate is 11.4 per 100,000 women, which means that, each year, more than 10,000 women die from this disease. 2There is evidence in the medical literature that suggests that periodic mammography screening is the most effective method for early diagnosis of breast cancer, with significant reductions in the specific mortality rates caused by this disease.[5][6] One recent historic landmark in mammography was the United States Food and Drug Administration's (FDA's) approval of the first full-field digital equipment for breast-cancer screening, in January 2000. 7Initially, digital technology was merely used to orient mammography-guided intervention procedures, in which detectors allowed physicians to obtain images from a small field of view.Full-field digital mammography is intended to replace film mammography in screening and diagnosing breast cancer.Digital technology brings with it a series of implicit benefits, which include storage of images in digital databases, without an ensuing loss of quality, and the ability to transmit images over long distances.One of the factors that most compromises film mammography is the fact that the images need to be developed and fixed chemically, and the image rejection rate due to processing errors can surpass 20%. 8The need to repeat these examinations increases costs and exposes patients to higher doses of ionizing radiation.Digital mammography does away with the chemical processing of images and, by enabling correction of brightness and contrast, can potentially reduce the rate of image rejection due to technical errors.
Other benefits inherent to digital technology are: the ability to amplify images on a monitor without the need to subject the patient to further X-ray exposure for magnified imaging; the ability to subsequently manipulate image contrast and brightness, and to use filters (software that selects or excludes certain gray-scale tones); and the ability to make computeraided diagnoses using specific software that recognizes image patterns in lesions.However, all studies that demonstrate the effectiveness of mammography for reducing breast-cancer mortality have been based on traditional, film mammography. 3[11] To date, none of the studies that have compared the performance of digital mammography and film mammography have assessed the impact of digital mammography specifically in terms of mortality.This type of study may take many years or even decades.[14]

OBJECTIVES
The objective of this systematic review of the scientific literature was to compare the performance of digital mammography and film mammography in terms of cancer detection rates, patient recall rates and characteristics of the tumors detected.

Type of study
Systematic review of the scientific literature and meta-analysis.

Types of studies included
Paired studies, cohort studies and clinical trials comparing film mammography and digital mammography that were published up to September 2009.

Types of participants
Women age 40 or over who were enrolled in breast cancer screening programs or who complained of specific ailments, and for whom mammography was recommended.

Types of endpoints
The following endpoints were assessed: A. Cancer detection rate; B. Patient recall rate; C. Characteristics of the tumors detected.
The cancer detection rate is the ratio between the number of cancer cases confirmed through biopsy (histopathological analysis) and the number of cases detected by each method, in all patients screened.The recall rate is the proportion of the patients with images that are sufficiently suspect to require the patients to be called in again for further screening, additional propaedeutic investigation and possible biopsy (histopathological analysis).
In order to compare the characteristics of the tumors found, we evaluated the ratio of invasive tumors to in situ tumors found using each method.

Subgroup analysis
The following subgroup analyses were conducted:

Inclusion criteria
All studies that compared film and digital mammography in similar populations in terms of the aforementioned endpoints were included in this study.

Search strategy
The search strategy involved searching four electronic databases (Medline via PubMed, Embase, Lilacs and Scopus) for articles on the topics of digital and film mammography that had been published up to September 2009.The bibliographic references of the studies included were checked in order to search for additional potentially relevant citations.The search strategy was sensitive to text and abstract wording, was unfiltered and used the follow-

Study locations
Two reviewers (MRT and WI or FV and WI) independently assessed the titles and summaries of all the resulting citations identified in the electronic search.Studies that potentially met the inclusion criteria were then read in full.Divergent opinions were resolved by reaching a consensus.Studies for which no consensus was reached were considered potentially eligible.

Data extraction
All studies with inclusion potential were separated for a full reading, critical assessment and data extraction, and this was done independently by two reviewers (DCS and WI).
A specific form was created for extracting the data from each study, and the following data were gathered: general information on each study (author and publication year), type of study, type of patients, study location, total number of cases, number of recall cases, total number of cases with a cancer diagnosis, positive predictive value and total numbers of in situ carcinomas and invasive carcinomas.Where included, these same data were stratified according to age group.
The data extracted by each researcher were inserted into individual spreadsheets that were subsequently compared.Divergent opinions were resolved by reaching a consensus.

Statistical analysis
We expressed the differences in cancer detection rates, patient recall rates and in situ and invasive cancer rates as relative risk (RR) rates using the statistical random-effects model for dichotomous data.
Risk ratios and the respective 95% confidence intervals were calculated.The data were compared using forest plots.Review Manager 5.0.20, which is a computer software freely distributed by the Cochrane Collaboration, was used for the calculations and to generate the plots.
The heterogeneity of the estimated effects among the studies included was analyzed using the heterogeneity or inconsistency test (I²).I² values less than 30% were considered to be indicators of low heterogeneity; values between 30-50% were considered to be indicators of moderate heterogeneity; and values above 50% were considered to be indicators of high heterogeneity.

A. Studies included
The electronic search identified 1,644 bibliographic citations: 1,203 references in Medline, 214 in Embase, 198 in Scopus and 29 in Lilacs.After reading the titles and abstracts, and after eliminating duplicates, 26 articles were chosen for a full reading and critical assessment.Of these, 15 were excluded because they did not meet the inclusion criteria or because they met the exclusion criteria.
Table 1 12,15-28 summarizes the reasons for excluding these 15 studies.1,618 studies excluded because they clearly did not compare digital mammography with lm mammography 15 studies excluded for the reasons described in Table 1 Studies included n = 11

Author
Reason for exclusion Cole et al. 15 Assessment of cases with previously known cancer diagnoses Hendrick et al. 28 Retrospective study using data from Pisano's 2005 Digital Mammographic Imaging Screening Trial (DMIST) 14 study Venta et al. 17 Did not assess clinical data of interest for this systematic review Lewin et al. 18 Study conducted with part of the population of the complete study, 30 which was published the following year and has been included in the systematic review Nishikawa et al. 19 Retrospective study using data from Pisano et al. 14 DMIST study Onishi et al. 20 Study of population with prior surgical recommendation Pisano et al. 16 Retrospective study using data from Pisano et al. 14 DMIST study Ranganathan et al. 21id not report data for calculating cancer detection rates and recall rates Seo et al. 22 Assessment using previously known mammary lesions Skaane et al. 23 Study that assessed endpoints other than those listed in the objective of this review, using data from the Skaane et al. 31 and Skaane et al. 29 Oslo I 12 and Oslo II 13 studies Skaane et al. 12 First published paper from the study data included in the Skaane et al. 31 Oslo I 12 study Skaane et al. 24 Study that assessed endpoints other than the objective of this review, using data from the Skaane et al. 31 and Skaane et al. 29 Oslo I 12 and Oslo II 13 studies Tosteson et al. 25 Study that assessed endpoints other than those listed in the objective of this review (cost effectiveness) Yamada et al. 26 Study that assesses endpoints other than those listed in the objective of this review (technical parameters) Yamada et al. 27 Known cancer cases included Table 1.Main reasons for excluding studies from the systematic review 12,[15][16][17][18][19][20][21][22][23][24][25][26][27][28] Upon completion of the search strategy and recovery, 11 studies were ultimately included in the systematic review (Fig-

B. Effectiveness of digital mammography compared with film mammography for detecting breast cancer
The results showed homogeneity in terms of the cancer detection rate.The cancer detection rate was significantly higher among patients who underwent digital mammography.
Based on the combination of data from the 11 studies included in this systematic review, the average relative-risk estimate for cancer detection among patients who underwent digital mammography was 1.17 (95% confidence interval, CI = 1.06-1.29;I² = 19%), in relation to film mammography.
On the other hand, taking into consideration only the paired studies, there was no significant difference in cancer detection rates between the two methods: average RR of 0.94 (95% CI = 0.77-1.15;I² = 0%).
Figure 2 shows the forest plot with the same information.

C. Recall rate for digital mammography compared with film mammography
There was great heterogeneity among the studies with regard to the patient recall rate (I² = 96%), even when they were analyzed according to study design (for cohort studies, I² = 95%; for paired studies, I² = 93%).

D. Characteristics of the tumors found using digital mammography, compared with film mammography
The characteristics of the tumors were seen to be similar from the two methods, in studies that provided this information.The relative risk of the proportion of invasive tumors in relation to the total number of tumors was 0.97 (95% CI = 0.91-1.04;I ² = 42%).

e. Breast cancer detection rates in different age groups, comparing digital and film mammography
Subgroup analysis according to age group, in the cohort studies that included this information, 32,35,38 revealed that digital mammography was better than film mammography for detecting   95% CI = 1.05-1.44;I² = 0%).No significant differences were identified in groups older than 60 years of age (RR = 1.14; 95% CI = 0.95-1.38;I² = 0%) (Figure 6 and Figure 7).
The RCT assessed patients between the ages of 45 and 49 years as well as patients older than 50 years.This trial found that digital mammography was more effective only in the older age group: RR = 1.58; 95% CI = 1.02 -2.46; versus RR = 1.55; 95% CI = 0.67-3.58,respectively (Figure 8).
Moreover, despite the data available from the RCT did not allow its inclusion in the meta-analysis, Pisano, in the DMIST study 14 , reported that digital mammography showed greater accuracy for perimenopausal and premenopausal patients and in patients younger than 50 years with higher mammary density.

DISCUSSION
While mammography is a diagnostic test, studies that address the spectrum of patients of interest in clinical practice (women in breast cancer screening programs) are technically and ethically limited in terms of traditional methodology.The main limitation is the lack of a reference standard (a "gold standard") that is acceptable for negative tests.While a biopsy can be taken of the suspect lesion to confirm or reject the presence of cancer, if mammography images on the entire mass of both breasts come up negative, such patients cannot be subjected to histopathological study.Thus, studies lack information on the number of false negative cases.Without this information, the sensitivity, specificity and negative predictive value cannot be reported.Even the positive tests confirmed by biopsy do not assess the contralateral breast or other regions of the breast with the suspect lesion.
Thus, without being able to create a 2 x 2 table, there are serious limitations with regard to obtaining data regarding sensitivity, specificity and negative predictive value.On the other hand, positive predictive values can be obtained.Nonetheless, the criteria for determining that a test is positive or suspect are extremely variable and do not follow specific standards in different clinics, as seen by the extreme heterogeneity of recall rates found in the studies included in this meta-analysis.
The most appropriate means of assessing the effectiveness of mammography for screening for breast cancer is to interpret it like an intervention, such that the primary endpoint is the reduction in mortality rates.Since digital mammography is a relatively     new method, it may take decades until there are sufficient data to compare this endpoint with findings from film mammography.
The primary focus of the present systematic review was a surrogate outcome, i.e. the cancer detection rate.
Following the FDA's approval of digital mammography in January 2000, 7 several studies comparing this method and film mammography were published.More recently, two systematic reviews on the matter were published.One, in 2007, included data from oral presentations given in congresses, in addition to published studies; 39 and the other, in 2009, also included the results from a cohort study. 35Before the present review had been completed, other studies not included in earlier reviews were published, including a large cohort study with more than 188,000 women, of whom 35,000 had undergone digital mammography.Thus, updating the reviews is justified in the name of strengthening the degree of evidence available for subsequent decision-making.
It can be seen that, in older studies published up to 2005, there were no significant differences in cancer detection rate  between the two types of mammography.However, the majority of the studies published from 2007 onwards have reported a significantly higher detection rate from digital mammography, compared with film mammography.One hypothesis for explaining this difference may relate to small technological advances in the equipment and to the learning curve for radiologists who use digital technology.
The subgroup assessment according to age groups showed that the superiority of digital mammography in terms of the cancer detection rate was more evident in the 50-60 year old group.This difference decreased and became insignificant in populations aged 60 years or over.The assessment of the population between 45 and 50 years of age in the RCT also did not demonstrate any significant difference.One possible explanation for this is that mammary density is higher in populations of intermediate age than in women aged 60 or older, and that digital resources, like the use of filters and contrast manipulation, probably influence cancer detection.In women with greater lipid replacement in the mammary parenchyma, which is to be expected in older age groups, these image manipulation resources do not provide additional value.On the other hand, in younger populations, the density of the mammary parenchyma is probably too high for both types of mammography, thus leading to similar cancer detection rates.
There was significant heterogeneity in the study results in terms of patient recall rates, which can be defined as the need for additional investigations to define the diagnosis of cancer.We believe that particular differences in the criteria used to recall patients, which were not standardized, were responsible for this difference in rates.
With regard to the advantages associated with digital technology, the process of replacing film mammography with digital mammography is slow.This is partially explained by the still-high initial costs associated with digital equipment and partially by a lack of trust in the capacity of new technology to detect breast cancer with the same accuracy as the already-revered method of film mammography.
The DMIST study contained an estimate that the cost of digital mammography systems was 1.5 to 4 times the cost of film mammography systems. 14However, some authors believe that the replacement of film systems with digital systems is inevitable and that the path to its acceptance is one from which there is no turning back. 40Moreover, it is well-known that the natural tendency is for the price of new technology to drop over time.
The results from this systematic review showed that there is a small difference in cancer detection rates in favor of digital mammography.However, the most important finding was that there is no evidence to support any claim that digital mammography is inferior to film mammography in terms of cancer detection rates.
The possibility of long-distance transmission, which enables assessment by specialists located thousands of miles away from the examination center, is an important element of the digital system to consider in developing countries or in places with a reduced number of radiologists experienced in mammography.

CONCLUSION
Digital mammography presents cancer detection rates that are slightly higher than the rates shown by film mammography.There are no significant differences in patient recall rates between digital and film mammography.The characteristics of the tumors found are similar in patients who undergo digital and film mammography.
A. Cancer detection rate in patients 60 years old or younger; B. Cancer detection rate in patients older than 60 years of age; C. Cancer detection rate in patients between 45 and 49 years old.
presenting any of the characteristics listed below were excluded: A. Studies that assessed the detection rate for just one of the methods.B. Studies from which it was impossible to extract cancer detection data and recall data relating to one of the mammography methods.C. Studies in which cases of previously reported cancer(s) were included in the study sample.D. Studies that assessed merely one of the technical parameters.E. Studies that did not include an abstract in English, Portuguese or Spanish in the databases consulted.F. Studies that examined subpopulations of larger studies previously included in this review.Location This meta-analysis was conducted at the Brazilian Cochrane Center at the Universidade Federal de São Paulo -Escola Paulista de Medicina (Unifesp-EPM), within the Emergency Medicine and Evidence-Based Medicine Program of the Department of Medicine.

Figure 1 .
Figure 1.Flowchart for the process of study inclusion in the review.

Figures 4 and 5
Figures 4 and 5 present the same information in the form of a forest plot graph.

Figure 2 .
Figure 2. Breast cancer detection rates from digital versus film mammography.

Figure 3 .
Figure 3. Patient recall rates from digital versus film mammography.

Figure 5 .
Figure 5.In situ carcinoma detection rates from digital versus film mammography.

Figure 6 .
Figure 6.Breast cancer detection rates comparing digital mammography with film mammography in subgroups of patients between 50 and 60 years old.

Figure 7 .
Figure 7. Breast cancer detection rates comparing digital mammography with film mammography in subgroups of patients aged 60 or over.

Table 2 .
Main characteristics of the studies included in the systematic review

Digital mammography Film mammography Weight Risk Ratio Year Risk Ratio Study or Subgroup Events Total Events Total M-H, Random, 95% CI M-H, Random, 95% CI 1..1 Cohorts
Invasive carcinoma detection rates from digital versus film mammography

.5.5 50 to 60 y.o.
Breast cancer detection rates comparing digital mammography with film mammography in subgroups of patients between 45 and 49 years old and patients between 50 and 69 years old.