Abstract

Background

Studies have shown that having mammography technologists review screening mammograms in addition to radiologist review may increase the number of breast cancers that are detected at screening mammography. We prospectively examined the effects on screening performance of adding independent double reading of screening mammograms by technologists to standard double reading by radiologists.

Methods

Twenty-one screening mammography technologists and eight certified screening radiologists participated in this study. From January 1, 2003, to January 1, 2005, all 61251 screening mammograms obtained at two mammography screening units in The Netherlands were independently read (although the second reader was not blinded to the first reader’s interpretation) by two technologists and, in turn, by two radiologists. Radiologists were blinded to the referral opinion of the technologists. During a 2-year follow-up period, we collected clinical data, breast imaging reports, biopsy results, and breast surgery reports of all women with a positive screening result (i.e., those that required additional imaging) from any reader. The distributions of categorical variables between subgroups were compared using chi-square or Fisher's exact tests. Differences in referral and detection by radiologists and technologists were assessed using McNemar's test. All statistical tests were two-sided.

Results

The radiologists referred 905 women (referral rate = 1.48%, 95% confidence interval [CI] = 1.38% to 1.57%), of whom 323 had breast cancer, corresponding to a positive predictive value of referral (PPV) of 35.7% (95% CI = 32.6% to 38.8%). Review of 446 additional technologist-positive readings led to another 80 referrals, resulting in the detection of 22 additional cancers. These extra referrals increased the initial referral rate from 1.48% to 1.61% (difference = 0.13%; 95% CI = 0.10% to 0.16%) and the cancer detection rate (CDR) from 5.27 to 5.63 cancers per 1000 women screened (difference = 0.36 cancers per 1000 women screened; 95% CI = 0.24 to 0.55). With technologist double reading only, 829 women would have been referred (referral rate = 1.35%, 95% CI = 1.26% to 1.45%); among these women, 286 cancers were diagnosed (PPV = 34.5%, 95% CI = 31.3% to 37.7%; CDR = 4.67 cancers per 1000 women screened, 95% CI = 4.13 to 5.21). Referral of all 1351 radiologist- and/or technologist-positive readings would have led to 362 cancers found at screening. The cancer detection rate for radiologist double reading would have increased from 5.27 to 5.91 cancers per 1000 women screened (relative increase = 12.1%, 95% CI = 8.8% to 16.5%; difference = 0.64 cancers per 1000 women screened, 95% CI = 0.47 to 0.87).

Conclusion

A referral strategy that includes all technologist-positive readings, which would have increased the cancer detection rate while maintaining a low referral rate, should be considered.

CONTEXT AND CAVEATS
Prior knowledge

Mammographic technologists who are taught to differentiate normal from abnormal mammograms have been shown to play a valuable role in the double reading of mammograms alongside radiologists, but this role has not been studied in a setting in which independent double reading of mammograms by screening radiologists is the standard of care.

Study design

Observational study conducted at two mammography screening units in The Netherlands where mammograms were independently read by two technologists and two radiologists.

Contribution

Adding independent double reading of mammograms by technologists to independent double reading by radiologists was effective in detecting additional cases of breast cancer while decreasing screening specificity only slightly.

Implications

All technologist-positive readings should be considered for referral.

Limitations

Reading conditions differed between technologists and radiologists. The technologists attended quality assurance sessions, which may have influenced their reading decisions. Assumptions were made concerning the screening outcomes of the technologist-positive readings that, in reality, had not been referred. The sensitivity of breast cancer screening at the 2-year follow-up was biased by new interval cancers that developed in the second year after screening.

From 1989 through 1997, the Dutch Nationwide Breast Cancer Screening Program was gradually implemented in The Netherlands ( 1 ). The program offered biennial screening mammography to all women aged 50–75 years. In this program, screening mammograms are performed by specialized mammography technologists, and the resulting films are independently double read by certified screening radiologists. Dutch screening radiologists receive certification in screening mammography after completing 3 weeks of training in film reading at the National Expert and Training Centre for Breast Cancer Screening (NETCB; Nijmegen, The Netherlands). This training involves tutorials conducted by highly experienced breast screening radiologists, daily reading of screening films, and attendance at assessment clinics and pathology meetings. The training material includes extensive files of abnormal screening mammograms and up-to-date literature on screening mammography. All specialized mammography technologists in The Netherlands also receive extensive training at the NETCB before their employment as screening technologists. Central to this training are instruction on mammography technique and positioning and evaluation of the images for technical quality. In addition, technologists also receive instruction in breast anatomy and mammographic features of benign and malignant breast conditions. Technologists receive further training at the NETCB on a yearly basis and are required to attend mammography symposia and conferences every 1–2 years.

For many years, investigators have explored the possibility of using nonradiologists to screen mammograms for breast cancer ( 2–7 ). Since the initiation of mammography screening in 1995 in the Eindhoven and De Kempen regions in the south of The Netherlands, screening technologists have been encouraged to look for mammographic abnormalities. The technologists attend quality assurance sessions once every 3 weeks, where they bring to the attention of a supervising breast radiologist any mammographic abnormalities they have identified that may require additional workup. Examinations for which the screening radiologists requested additional workup are also reviewed at these meetings, mammographic abnormalities are compared to the pathology outcome, and false-negative cases (i.e., interval cancers) are discussed. As a result of this intensive and continuous education, the screening technologists obtain considerable experience in reading mammograms.

Technologists who are taught to differentiate normal from abnormal mammograms have been shown to play a valuable role in the double reading of mammograms alongside radiologists ( 2–7 ), but this role has never been studied in a setting in which independent double reading of mammograms by screening radiologists is the standard of care. We conducted a prospective study to examine the effect of adding independent double reading of mammograms by technologists to independent double reading by radiologists on parameters of mammography performance, including referral rate, cancer detection rate (CDR), and sensitivity and specificity of breast cancer screening. The data were analyzed with respect to different reading strategies, including double reading by technologists only, double reading by radiologists only, and referral of all positive readings by radiologists or technologists.

Subjects and Methods

Study Population

We included in this study all 61251 women aged 50–75 years who underwent screening mammography at one of two specialized conventional mammography screening units (one fixed and one mobile) in the southern breast cancer screening region of The Netherlands (Bevolkings Onderzoek Borstkanker Zuid) between January 1, 2003, and January 1, 2005. The Dutch Breast Cancer Screening Program targets asymptomatic women; women who are experiencing breast symptoms (e.g., a palpable abnormality, skin retraction, or nipple retraction) are instructed in the invitation letter to screening mammography to consult their general practitioner rather than waiting for their next screening examination. Details of the Dutch Nationwide Breast Cancer Screening Program, which offers biennial screening mammography to women aged 50–75 years, are described elsewhere ( 1 , 8 ). Before screening mammography was performed, all women completed a short questionnaire with questions about any previous breast surgery or breast malignancy, their family history of breast cancer, and their use of hormonal replacement therapy. (No data about ethnic background or socioeconomic status were collected.) Only the questionnaire data for women with a screening mammogram that required additional diagnostic assessment in our study were analyzed.

Written informed consent regarding patient identification and exchange of patient data was obtained from all women participating in the breast cancer screening program. Institutional review board approval was not required for this study.

Mammogram Readers and Referral Procedure

Technologist Double Reading.

All 21 technologists who worked at the two screening units participated in this study. Their experience in screening mammography ranged from 1 to 124 months (mean = 69 months; median = 74 months). The two technologists who were on duty to perform the screening examinations independently double-read (although the second reader was not blinded to the reading outcome of the first reader) each mammogram at the screening site immediately after the examination was completed. The mammogram viewing conditions for technologists, such as room lighting and background masking, were not controlled during the study. At the time of each subsequent screening, the technologists could view previous screening mammograms when they assessed the ones they took. The technologists were asked to decide for each mammogram whether additional workup was required (i.e., whether the mammogram was positive). For mammograms with discrepant readings, the technologists were asked to try to reach consensus. However, if one technologist persisted in the opinion that additional workup was indicated, the mammogram was considered to be positive. For each positive mammogram, the technologists recorded the woman's name and date of birth, the date of screening, and the mammographic findings on a special form that was developed for this study. The mammographic findings were classified according to one of five categories of abnormal findings: suspicious high density (e.g., spiculated density or density with indistinct borders), suspicious microcalcifications (e.g., pleomorphic, branching, or amorphous/indistinct microcalcifications), high density in combination with microcalcifications, architectural distortion, or asymmetry.

Radiologist Double Reading.

All eight radiologists who worked at the two screening units participated in this study. Each radiologist read more than 7000 screening mammograms annually. Their experience in screening mammography ranged from 39 to 95 months (mean = 79 months; median = 94 months). The radiologists read all mammograms from each day's mammographic examinations on a light box in a room devoid of daylight. Each of two radiologists who were scheduled to assess the screening examinations independently read, on average, 140 mammograms per day over a period of 60–75 minutes. The two radiologists read the films at different times (i.e., the second radiologist was not in the room at the same time that the first radiologist was reading the film). The second reader was not blinded to the first reader's interpretation and was aware of the first reader's interpretation. In case of a divergent reading, the second reader took the mammogram from the reading unit to the hospital where all screening radiologists work and tried to reach consensus with the first reader on the same day the mammogram was independently read by the two screening radiologists. If the two screening radiologists could not agree and one persisted that further examination was indicated, then the woman was recalled. The radiologists were blinded to the referral opinion of the technologists but had access to all relevant clinical information (e.g., the woman's history of previous breast surgery or whether she had palpable breast abnormalities) indicated on the standard form completed by the technologists. Prior screening mammograms were always available for comparison, and the radiologists used the same five categories used by the technologists to classify positive mammograms. Like the technologists, the radiologists tried to reach consensus in cases where they initially did not agree about referral.

Referral.

A woman was referred for additional workup (primary referral) if the mammogram was considered to be positive by both screening radiologists after initial independent double reading or, in the case of discrepant readings, if at least one radiologist considered referral necessary after consensus meeting. During the 3-weekly quality assurance sessions, mammograms that the technologists had considered to be positive but that had not been referred by the radiologists were reviewed by two screening radiologists (preferably those who had performed the original assessment). The radiologists were informed about the mammographic abnormalities detected by the technologists. A woman was referred if, on review, at least one of the radiologists considered workup to be necessary (secondary referral).

Screening Follow-up

The follow-up period for all screened women included the time through the next screening round (the screening interval was approximately 2 years). For all women with a screening mammogram that was considered to be positive by at least one of the radiologists or technologists, we collected data on diagnostic procedures undertaken, breast cancer diagnosis, histopathology, and TNM (tumor–node–metastasis) classification ( 9 ) to identify screen-detected cancers and interval cancers. Interval cancers are breast cancers that are diagnosed in women after a screening examination yields negative results (defined as no recommendation for referral). We investigated whether the mammographic abnormalities on diagnostic films for interval cancers in cases that the technologists would have referred corresponded to the abnormalities registered by the technologists at screening. We excluded breast malignancies other than primary breast cancers from this analysis and considered lobular carcinoma in situ to be a benign lesion. For women with bilateral disease, we included the cancer with the highest stage in this analysis. Multiple foci of cancer in one breast were counted as one cancer.

Quality Assurance

Throughout the study period, the screening radiologists reviewed breast cancer cases that were detected after secondary referral, as well as interval cancer cases. Every 3 weeks, technologists attended quality assurance and learning sessions. Together with a supervising breast radiologist, they reviewed breast cancer cases that had been detected by radiologists only.

Statistical Analysis

The primary outcome measures were the referral rate, the cancer detection rate (defined as the number of cancers detected per 1000 women screened), the sensitivity and specificity of mammography screening, the positive predictive value of referral, and TNM tumor stage. To calculate the screening performance characteristics, that is, the sensitivity, specificity, and positive predictive value of referral, we had to make some assumptions, in particular, for the technologist-positive readings that, in reality, had not been referred upon review. Breast cancer cases in the screened population included all breast cancers that had been diagnosed after 1 year and after 2 years of follow-up, that is, all screen-detected cancers and all interval cancers, and cancers detected at the subsequent screening examination in women who had not been (secondarily) referred after a technologist-positive reading. The true-positive test results with independent technologist double reading and with recall of all technologist- or radiologist-positive screens included the interval cancers and cancers detected at the subsequent screening examination in women who had not been (secondarily) referred after a technologist-positive reading and the corresponding screen-detected breast cancers. To assess whether the attendance at repeated quality assurance sessions that was required of technologists was associated with screening performance parameters for the technologists, we analyzed whether the screening results of technologists changed over time. All data were entered into an Excel spreadsheet (Microsoft, Redmond, WA). Statistical analyses were performed using Statistical Package for Social Sciences software (version 13.0.1; SPSS Inc, Chicago, IL). The chi-square test or Fisher's exact test was used to compare the distribution of categorical variables between subgroups defined by method of referral. McNemar's test was used to assess the differences in referral and detection by radiologists and technologists, using the numbers of discordant pairs. Poisson analysis was used to calculate confidence intervals (CIs) around the increase of referral and detection rates with secondary referral as a proportion of the original rates. Statistical tests were two-sided. P value less than .05 was considered to be statistically significant.

Results

Independent Radiologist Double Reading Employing Secondary Referral

A total of 61251 mammograms were read in 2003 and 2004: 7038 (11.5%) were initial (prevalent) mammograms and 54213 (88.5%) were subsequent (incident) mammograms. Radiologists had fewer discordant readings (0.23% [141/61251], 95% CI = 0.19% to 0.27%) than technologists (0.44% [268/61251], 95% CI = 0.39% to 0.49%) ( Fig. 1 ). A total of 985 women were referred for further workup, corresponding to a referral rate of 1.61% (95% CI = 1.51% to 1.71%). Of those women, 905 were referred based on a positive reading by the radiologists (primary referral) and 80 were referred based on radiologist review of a positive reading by technologists (secondary referral). The characteristics of screen-positive women and their mammograms are presented in Table 1 .

Fig. 1

Mammography screening outcome at the 2-year follow-up. Mammograms were independently read (although the second reader was not blinded to the reading outcome of the first reader) by two technologists and, in turn, by two radiologists. Radiologists were blinded to the referral opinion of the technologists. SDC = screen-detected cancer; IC = interval cancer; SDC next = cancer detected at subsequent screening; positive reading = screening examination requiring additional imaging; initial concordance = reading agreement between two readers following independent double reading; with consensus = reading agreement between two readers following independent double reading and consensus meeting; without consensus = no reading agreement between two readers following independent double reading and consensus meeting; Secondary referral = referral based on radiologist review of a technologist positive reading.

Fig. 1

Mammography screening outcome at the 2-year follow-up. Mammograms were independently read (although the second reader was not blinded to the reading outcome of the first reader) by two technologists and, in turn, by two radiologists. Radiologists were blinded to the referral opinion of the technologists. SDC = screen-detected cancer; IC = interval cancer; SDC next = cancer detected at subsequent screening; positive reading = screening examination requiring additional imaging; initial concordance = reading agreement between two readers following independent double reading; with consensus = reading agreement between two readers following independent double reading and consensus meeting; without consensus = no reading agreement between two readers following independent double reading and consensus meeting; Secondary referral = referral based on radiologist review of a technologist positive reading.

Table 1

Characteristics of women with a screening examination that required additional evaluation (“screen-positive” examination) *

 Referral in actual screening situation   
 Primary referral  Secondary referral (upon review)
 
No referral in actual screening situation
 
Total
 
  Considered to be screen-positive by radiologists only
 
Considered to be screen-positive by technologists and radiologists
 
Considered to be screen-positive by technologists only
 
Considered to be screen-positive by technologists only
 
Considered to be screen-positive by technologists and/or radiologists
 
Characteristic N = 522 N = 383 N = 80 N = 366 N = 1351 
Mean age, y (SD) 59.2 (7.8) 61.6 (7.3) 61.6 (7.2) 61.1 (7.6) 60.5 (7.6) 
Initial screens, No. (%) 145 (27.8) 56 (14.6) 9 (11.3) 58 (15.8) 268 (19.8) 
History of benign breast surgery, No. (%) 80 (15.3) 49 (12.8) 12 (15.0) 55 (15.0) 196 (14.5) 
History of breast cancer, No. (%) 3 (0.6) 3 (0.8) 0 (0) 2 (0.5) 8 (0.6) 
Family history of breast cancer † , No. (%)  63 (12.1) 56 (14.6) 7 (8.8) 46 (12.6) 172 (12.7) 
Use of hormone replacement therapy, No. (%) 53 (10.2) 41 (10.7) 7 (8.8) 28 (7.7) 129 (9.5) 
Breast symptoms, No. (%) 2 (0.4) 4 (1.0) 0 (0) 3 (0.8) 9 (0.7) 
Mammographic abnormality, No. (%)      
    High density 365 (69.9) 248 (64.8) 48 (60.0) 253 (69.1) 914 (67.7) 
    Microcalcifications 100 (19.2) 82 (21.4) 30 (37.5) 86 (23.5) 298 (22.1) 
    High density with microcalcifications 23 (4.4) 45 (11.7) 0 (0) 16 (4.4) 84 (6.2) 
    Architectural distortion 20 (3.8) 8 (2.1) 1 (1.3) 8 (2.2) 37 (2.7) 
    Breast parenchyma asymmetry 14 (2.7) 0 (0) 1 (1.3) 3 (0.8) 18 (1.3) 
 Referral in actual screening situation   
 Primary referral  Secondary referral (upon review)
 
No referral in actual screening situation
 
Total
 
  Considered to be screen-positive by radiologists only
 
Considered to be screen-positive by technologists and radiologists
 
Considered to be screen-positive by technologists only
 
Considered to be screen-positive by technologists only
 
Considered to be screen-positive by technologists and/or radiologists
 
Characteristic N = 522 N = 383 N = 80 N = 366 N = 1351 
Mean age, y (SD) 59.2 (7.8) 61.6 (7.3) 61.6 (7.2) 61.1 (7.6) 60.5 (7.6) 
Initial screens, No. (%) 145 (27.8) 56 (14.6) 9 (11.3) 58 (15.8) 268 (19.8) 
History of benign breast surgery, No. (%) 80 (15.3) 49 (12.8) 12 (15.0) 55 (15.0) 196 (14.5) 
History of breast cancer, No. (%) 3 (0.6) 3 (0.8) 0 (0) 2 (0.5) 8 (0.6) 
Family history of breast cancer † , No. (%)  63 (12.1) 56 (14.6) 7 (8.8) 46 (12.6) 172 (12.7) 
Use of hormone replacement therapy, No. (%) 53 (10.2) 41 (10.7) 7 (8.8) 28 (7.7) 129 (9.5) 
Breast symptoms, No. (%) 2 (0.4) 4 (1.0) 0 (0) 3 (0.8) 9 (0.7) 
Mammographic abnormality, No. (%)      
    High density 365 (69.9) 248 (64.8) 48 (60.0) 253 (69.1) 914 (67.7) 
    Microcalcifications 100 (19.2) 82 (21.4) 30 (37.5) 86 (23.5) 298 (22.1) 
    High density with microcalcifications 23 (4.4) 45 (11.7) 0 (0) 16 (4.4) 84 (6.2) 
    Architectural distortion 20 (3.8) 8 (2.1) 1 (1.3) 8 (2.2) 37 (2.7) 
    Breast parenchyma asymmetry 14 (2.7) 0 (0) 1 (1.3) 3 (0.8) 18 (1.3) 
*

SD = standard deviation.

At least one first-degree relative (mother, sister, daughter) with a diagnosis of breast cancer before the age of 50 years or at least two second-degree relatives with breast cancer.

A total of 345 screen-detected cancers were diagnosed after 2 years of follow-up ( Figs. 1 and 2 ), corresponding to a breast cancer detection rate of 5.63 breast cancers per 1000 women screened (95% CI = 5.04 to 6.23 cancers per 1000 women screened) ( Table 2 ). Of the 345 screen-detected cancers, 247 were diagnosed in women who were referred by both the radiologists and the technologists, 76 were diagnosed in women who were referred by the radiologists only, and 22 were diagnosed in women who were referred after radiologist review of technologist-positive screens ( Fig. 2 ). There were statistically significant differences among these subgroups with respect to both the proportion of breast cancers that were ductal carcinoma in situ ( P = .005, Fisher's exact test) and the lymph node status of the invasive cancers ( P <.001, Fisher's exact test) ( Table 3 ).

Fig. 2

Mammographic abnormalities in breast cancers detected by technologists and radiologists. SDC = screen-detected cancer; IC = interval cancer; SDC next = cancer detected at subsequent screening.

Fig. 2

Mammographic abnormalities in breast cancers detected by technologists and radiologists. SDC = screen-detected cancer; IC = interval cancer; SDC next = cancer detected at subsequent screening.

Table 2

Performance parameters for the four reading strategies *

Parameter Independent technologist double reading Independent radiologist double reading Reassessment of technologist-positive readings by radiologists after independent radiologist double reading Referral of all technologist-positive readings and radiologist-positive readings 
Referral rate, % (95% CI) 1.35 (1.26 to 1.45) 1.48 (1.38 to 1.57) 1.61 (1.51 to 1.71) 2.21 (2.09 to 2.32) 
1-y follow-up     
    CDR, per 1000 women, (95% CI) 4.60 (4.07 to 5.14) 5.16 (4.59 to 5.73) 5.52 (4.93 to 6.11) 5.80 (5.19 to 6.40) 
    Sensitivity, % (95% CI) 72.3 (67.9 to 76.7) 83.6 (79.9 to 87.3) 89.4 (86.3 to 92.5) 91.0 (88.2 to 93.9) 
    Specificity, % (95% CI) 99.1 (99.0 to 99.2) 99.0 (99.0 to 99.1) 98.9 (98.9 to 99.0) 98.4 (98.3 to 98.5) 
    PPV, % (95% CI) 34.0 (30.8 to 37.2) 34.9 (31.8 to 38.0) 34.3 (31.3 to 37.3) 26.3 (23.9 to 28.6) 
2-y follow-up     
    CDR, per 1000 women (95% CI) 4.67 (4.13 to 5.21) 5.27 (4.70 to 5.85) 5.63 (5.04 to 6.23) 5.91 (5.30 to 6.52) 
    Sensitivity, % (95% CI) 61.5 (57.1 to 65.9) 69.5 (65.3 to 73.6) 74.2 (70.2 to 78.2) 77.8 (74.1 to 81.6) 
    Specificity, % (95% CI) 99.1 (99.0 to 99.2) 99.0 (99.0 to 99.1) 98.9 (98.9 to 99.0) 98.4 (98.3 to 98.5) 
    PPV, % (95% CI) 34.5 (31.3 to 37.7) 35.7 (32.6 to 38.8) 35.0 (32.0 to 38.0) 26.8 (24.4 to 29.2) 
Parameter Independent technologist double reading Independent radiologist double reading Reassessment of technologist-positive readings by radiologists after independent radiologist double reading Referral of all technologist-positive readings and radiologist-positive readings 
Referral rate, % (95% CI) 1.35 (1.26 to 1.45) 1.48 (1.38 to 1.57) 1.61 (1.51 to 1.71) 2.21 (2.09 to 2.32) 
1-y follow-up     
    CDR, per 1000 women, (95% CI) 4.60 (4.07 to 5.14) 5.16 (4.59 to 5.73) 5.52 (4.93 to 6.11) 5.80 (5.19 to 6.40) 
    Sensitivity, % (95% CI) 72.3 (67.9 to 76.7) 83.6 (79.9 to 87.3) 89.4 (86.3 to 92.5) 91.0 (88.2 to 93.9) 
    Specificity, % (95% CI) 99.1 (99.0 to 99.2) 99.0 (99.0 to 99.1) 98.9 (98.9 to 99.0) 98.4 (98.3 to 98.5) 
    PPV, % (95% CI) 34.0 (30.8 to 37.2) 34.9 (31.8 to 38.0) 34.3 (31.3 to 37.3) 26.3 (23.9 to 28.6) 
2-y follow-up     
    CDR, per 1000 women (95% CI) 4.67 (4.13 to 5.21) 5.27 (4.70 to 5.85) 5.63 (5.04 to 6.23) 5.91 (5.30 to 6.52) 
    Sensitivity, % (95% CI) 61.5 (57.1 to 65.9) 69.5 (65.3 to 73.6) 74.2 (70.2 to 78.2) 77.8 (74.1 to 81.6) 
    Specificity, % (95% CI) 99.1 (99.0 to 99.2) 99.0 (99.0 to 99.1) 98.9 (98.9 to 99.0) 98.4 (98.3 to 98.5) 
    PPV, % (95% CI) 34.5 (31.3 to 37.7) 35.7 (32.6 to 38.8) 35.0 (32.0 to 38.0) 26.8 (24.4 to 29.2) 
*

CI = confidence interval; CDR = cancer detection rate; PPV = positive predictive value of referral.

Table 3

Breast cancers and tumor characteristics in women with a screening examination that required additional evaluation (“screen-positive” examination) *

 Referred in actual screening situation  
 Primary referral  Secondary referral (upon review)
 
Not referred in actual screening situation
 
  Considered to be screen positive by radiologists only
 
Considered to be screen positive by technologists and radiologists
 
Considered to be screen positive by technologists only
 
Considered to be screen positive by technologists only
 
Screening outcome N = 522 N = 383 N = 80 N = 366 
Breast cancers † , No.  76 247 22 17 
Positive predictive value, % (95% CI) 14.6 (11.5 to 17.6) 64.5 (59.7 to 69.3) 27.5 (17.7 to 37.3) 4.6 (2.5 to 6.8) 
Type of breast cancer † , No. (%)      
    DCIS 24 (31.6) 35 (14.2) 5 (22.7) 1 (8.3) 
    Invasive ‡ 52 (68.4) 212 (85.8) 17 (77.3) 11 (91.7) 
       T1a–c 45 (86.5) 166 (78.3) 16 (94.1) 4 (36.4) 
       T2 6 (11.5) 46 (21.7) 0 (0) 6 (54.5) 
       Unknown 1 (1.9) 0 (0) 1 (5.9) 1 (9.1) 
Lymph node status of invasive cancers, No. (%)     
    Positive 14 (26.9) 55 (25.9) 1 (5.9) 8 (72.8) 
    Negative 36 (69.2) 156 (73.6) 13 (76.5) 3 (27.3) 
    Unknown 2 (3.8) 1 (0.5) 3 (17.6) 0 (0) 
 Referred in actual screening situation  
 Primary referral  Secondary referral (upon review)
 
Not referred in actual screening situation
 
  Considered to be screen positive by radiologists only
 
Considered to be screen positive by technologists and radiologists
 
Considered to be screen positive by technologists only
 
Considered to be screen positive by technologists only
 
Screening outcome N = 522 N = 383 N = 80 N = 366 
Breast cancers † , No.  76 247 22 17 
Positive predictive value, % (95% CI) 14.6 (11.5 to 17.6) 64.5 (59.7 to 69.3) 27.5 (17.7 to 37.3) 4.6 (2.5 to 6.8) 
Type of breast cancer † , No. (%)      
    DCIS 24 (31.6) 35 (14.2) 5 (22.7) 1 (8.3) 
    Invasive ‡ 52 (68.4) 212 (85.8) 17 (77.3) 11 (91.7) 
       T1a–c 45 (86.5) 166 (78.3) 16 (94.1) 4 (36.4) 
       T2 6 (11.5) 46 (21.7) 0 (0) 6 (54.5) 
       Unknown 1 (1.9) 0 (0) 1 (5.9) 1 (9.1) 
Lymph node status of invasive cancers, No. (%)     
    Positive 14 (26.9) 55 (25.9) 1 (5.9) 8 (72.8) 
    Negative 36 (69.2) 156 (73.6) 13 (76.5) 3 (27.3) 
    Unknown 2 (3.8) 1 (0.5) 3 (17.6) 0 (0) 
*

CI = confidence interval; DCIS = ductal carcinoma in situ.

Breast cancers in the first three groups are screen-detected cancers. Breast cancers in the last group include 12 interval cancers and 5 cancers detected at the subsequent screening round.

Tumor–node–metastasis classification (9).

Compared with radiologist double reading only, for which the cancer detection rate was 5.27 cancers per 1000 women screened, secondary referral resulted in a 6.8% increase (95% CI = 4.5% to 10.3%; Poisson regression) in the breast cancer detection rate at the 2-year follow-up (i.e., to 5.63 cancers per 1000 women screened; Table 2 ).

Independent Double Reading by Technologists Versus by Radiologists

In total, the technologists would have referred 829 women for further workup without radiologist review, corresponding to a referral rate of 1.35% (95% CI = 1.26% to 1.45%) ( Fig. 1 ). At the 2-year follow-up, 286 breast cancers had been diagnosed among these 829 women, including 12 interval cancers and five mammographic abnormalities that proved to be cancers at the subsequent screening examination. Referral of all 829 technologist-positive double readings would have resulted in a cancer detection rate of 4.60 per 1000 women screened (95% CI = 4.07 to 5.14 per 1000 women screened), with a sensitivity of 72.3% (95% CI = 67.9% to 76.7%), a specificity of 99.1% (95% CI = 99.0% to 99.2%), and a positive predictive value of referral of 34.0% (95% CI = 30.8% to 37.2%) at 1-year follow-up ( Table 2 ; Supplementary Table 1, available online). At 2-year follow-up, the cancer detection rate would have been 4.67 per 1000 women screened (95% CI = 4.13 to 5.21 per 1000 women screened), a sensitivity of 61.5% (95% CI = 57.1% to 65.9%), a specificity of 99.1% (95% CI = 99.0% to 99.2%), and a positive predictive value of referral of 34.5% (95% CI = 31.3% to 37.7%). The referral rates for technologist double reading were statistically significantly lower in 2003 than in 2004 (1.25% versus 1.45%, difference = 0.20%, 95% CI = 0.02% to 0.39%; P = .03, chi-square test), but cancer detection rates (4.68 per 1000 women screened in 2003 versus 4.66 per 1000 women screened in 2004; P = 1.0) and sensitivity of screening (63.1% in 2003 versus 60.1% in 2004; P = .5) did not differ statistically significantly (Supplementary Table 2, available online).

The referral rate after primary radiologist double reading was statistically significantly higher than the referral rate for independent technologist double reading would have been (1.48% versus 1.35%; difference = 0.13%; 95% CI = 0.02% to 0.22%; P = .02, McNemar's test). Radiologist double reading yielded more breast cancers than technologist double reading at the 1-year follow-up (316 cancers versus 282 cancers) and at the 2-year follow-up (323 cancers versus 286 cancers), resulting in a statistically significantly higher cancer detection rate at the 2-year follow-up (5.27 versus 4.67 cancers per 1000 women screened, difference = 0.6 cancers per 1000 women screened, 95% CI = 0.3 to 0.9 cancers per 1000 women screened; P <.001, McNemar's test) ( Table 2 ). The sensitivity, specificity, and positive predictive value of referral for independent radiologist double reading at the 1-year follow-up were 83.6% (95% CI = 79.9% to 87.3%), 99.0% (95% CI = 99.0% to 99.1%), and 34.9% (95% CI = 31.8% to 38.0%), respectively, and at 2-year follow-up were 69.5% (95% CI = 65.3% to 73.6%), 99.0% (95% CI = 99.0% to 99.1%), and 35.7% (95% CI = 32.6% to 38.8%), respectively ( Table 2 ). Of the 362 women who had been diagnosed with breast cancer during the 2 years of follow-up, including the 345 women with screen-detected cancers, the 12 women with interval cancers, and five women with cancers detected at the subsequent screening round, 76 women (21.0%) would not have been referred by the technologists and 39 women (10.8%) would not have been referred by the radiologists ( Fig. 2 ).

Recall of All Technologist- or Radiologist-Positive Screens

The referral rate after recall of all screens for which radiologists and/or technologists required additional imaging was higher than the referral rate after recall of screens that were positive by independent radiologist double reading only (2.21% versus 1.48%, difference = 0.73%, 95% CI = 0.66% to 0.80%; Poisson regression) ( Table 2 ). In addition, initial detection of the 22 cancers that were detected through secondary recall, the 12 interval cancers, and the five mammographic abnormalities that proved to be malignant at the subsequent screening would have increased the cancer detection rate at the 2-year follow-up by 12.1% (95% CI = 8.8% to 16.5%), that is, from 5.27 to 5.91 cancers per 1000 women screened. The positive predictive value of referral would have decreased from 35.7% to 26.8% (difference = −8.9%, 95% CI = −9.9% to −7.8%, P <.001; Poisson regression). The median time to diagnosis for the 12 interval cancers and the five cancers detected at subsequent screening among the 366 women for whom the technologists required additional imaging but who were not referred upon radiologist review was 466 days (range = 83–727 days).

Discussion

Independent double reading by mammography technologists in the southern breast screening region in The Netherlands has been shown to result in a breast cancer detection rate that is comparable to the nationwide breast cancer detection rate for independent radiologist double reading ( 10 ). By contrast, in this study, the technologists detected fewer cancers than the screening radiologists. However, the primary aim of our study was not to investigate whether technologists can read mammograms as well as radiologists but to examine the possible benefit of additional technologist double reading on screening results. We found that adding technologist double reading to standard double reading by radiologists increased the cancer detection rate while maintaining a low referral rate.

The role of technologists as film readers in mammography screening has been explored in several studies. Pauli et al. ( 4 ) found that in a setting where mammograms are typically read by one radiologist, adding a reading by a technologist may increase the number of breast cancers detected at screening mammography. Several experimental studies ( 2 , 3 ) have shown that the screening results of formally trained technologists can approach those of radiologists in a single reading setting. Tonita et al. ( 5 ) implemented single technologist reading in addition to single radiologist reading in a population-based breast cancer screening program and suggested that this approach would be a cost-effective alternative to radiologist double reading. In countries with a shortage of trained and experienced radiologists, such as the United States and the United Kingdom, radiologist double reading may not be feasible. In The Netherlands, where there is no such shortage, independent double radiologist reading is the standard of care.

In contrast to other studies ( 2 , 4 ), we found that the technologists in our study detected statistically significantly fewer cancers than the radiologists in our study. There are several possible explanations for this observation. First, the cancer detection rate for the radiologists in our study, 5.27 cancers per 1000 women screened, was higher than the nationwide average rate of 4.90 cancers per 1000 women screened ( 10 ). Second, the radiologists in our study all had at least 3 years of breast cancer screening experience, whereas some of the technologists in our study had had little screening practice before their participation in this study. Third, the film reading conditions were better for the radiologists than for the technologists because they performed the assessments in a room devoid of daylight. On the other hand, the technologists evaluated the mammograms one at a time, whereas the radiologists consecutively assessed daily productions of at least 140 examinations. Finally, the technologists and the radiologists received different feedback on their decisions. That is, the technologists attended quality assurance sessions every 3 weeks, whereas the radiologists reviewed a substantial part of their recall decisions via a multidisciplinary approach at the hospital where they worked.

Film reading training at the NETCB differed between technologists and radiologists. It is possible that the breast cancer detection rate for technologists would improve if they were to receive the same film reading training as radiologists. However, such training could reduce the benefit of having technologists read mammograms if the technologists “learned” a recall strategy that was identical to that of radiologists. The benefit of having discrepant reading results between radiologists and technologists, which in our series constituted a clinically important subset of screening mammograms with an 8.7% prevalence of breast cancer (39 cancers in 446 discrepant readings), might be lost.

Secondary referral after review of technologist-positive readings proved to be a successful strategy for detecting more breast cancers: the breast cancer detection rate increased by 6.8% compared with the rate for double reading by radiologists alone, whereas screening specificity decreased only slightly. Although the technologists detected fewer high densities with suspicious microcalcifications than the radiologists, there were a remarkably high number of microcalcification abnormalities among the malignant cases at secondary recall. We speculate that differences in reading training at the NETCB between radiologists and technologists could account for this finding.

A screening strategy of referral of all technologist-positive readings in addition to all radiologist-positive readings would have resulted in a 12.1% relative increase in the cancer detection rate, i.e., from 5.27 to 5.91 cancers per 1000 women screened. This increase would have been associated with a higher referral rate of 2.21% (from 1.48%) and a higher sensitivity of 77.8% (from 69.5%); specificity decreased slightly to 98.4% (from 99.0%), whereas the positive predictive value of referral would have decreased from 35.7% to 26.8%. A referral rate of 2.21% still is considerably lower than the referral rates observed in the United States and the United Kingdom, whereas a positive predictive value of referral of 26.8% is considerably higher than that found in US and U.K. screening programs ( 11 ). Moreover, a 2.21% referral rate would still be cost-effective in the Dutch screening program ( 12 , 13 ).

Additional independent double reading of mammograms by technologists is likely to be a cost-effective approach in a screening program that is characterized by low referral rates such as that in The Netherlands. In the Dutch screening setting, a doubling or even tripling of the referral rate would still be cost-effective if it resulted in a 2%–5% increase in the cancer detection rate ( 13 ). The extra costs of adding independent double reading of mammograms by technologists would include the extra wages for technologists and the costs associated with having more women who require additional diagnostic assessment. The benefits of additional technologist double reading would involve increased detection of early-stage breast cancers, which require less invasive treatments. The prevalence of breast cancer in women who were referred after re-evaluation of screen examinations for which only technologists considered workup necessary exceeded 25% (22 of 80 women). Our finding that breast cancer was present in 4.6% of the technologist positive readings (17 of 366) for which radiologists persisted that there was no indication for further workup suggests that radiologists should consider referral of all technologist-positive readings rather than referral of only selected readings.

Our study has several limitations that must be considered when interpreting the results. First, this study was an observational study. Differences in reading performance between technologists and radiologists may be due to differences in reading conditions, including lighting and background masking. However, even if the reading conditions had been comparable for technologists and radiologists, differences in performance would still persist because of differences in training and reading skills. Second, it could be argued that the quality assurance sessions, which were introduced several years before our study started and lasted throughout the study period, may have influenced the reading decisions of the technologists, possibly leading to closer results between technologists and radiologists at the time of our study. Third, we had to make some assumptions when calculating the performance parameters for the different reading strategies, in particular, assumptions concerning the screening outcomes of the technologist-positive readings that, in reality, had not been referred. Fourth, the lower sensitivity of breast cancer screening at the 2-year follow-up compared with that at the 1-year follow-up was biased by new interval cancers that developed in the second year after screening. van Dijk et al. ( 14 ) showed that a considerable percentage of interval cancers in a biennial screening program appear de novo between two screening rounds. We calculated screening outcome parameters for both 1 year and 2 years of follow-up. Considering the 2-year screening interval in The Netherlands, 2 years of follow-up will provide full information about interval cancers, any delay in breast cancer diagnosis after a positive screen, and the total costs of follow-up ( 15 ). In addition, review of late interval cancers is part of the quality assurance and evaluation of the Dutch breast cancer screening program.

In conclusion, we found that adding independent double reading of mammograms by technologists to independent double reading by radiologists was effective in detecting additional cases of breast cancer. Our results indicate that all technologist-positive readings should be considered for referral because this subset of screening mammograms shows a high prevalence of breast cancer.

Funding

The authors received no external funding for this work.

References

(1)
Fracheboud
J
de Koning
HJ
Boer
R
Groenewoud
JH
Verbeek
ALM
Broeders
MJM
, et al.  . 
Nationwide breast cancer screening programme fully implemented in the Netherlands
Breast
 , 
2001
, vol. 
10
 (pg. 
6
-
11
)
(2)
Bassett
LW
Hollatz-Brown
AJ
Bastani
R
Pearce
JG
Hirji
K
Chen
L
Effects of a program to train radiologic technologists to identify abnormalities on mammograms
Radiology
 , 
1995
, vol. 
194
 (pg. 
189
-
92
)
(3)
Pauli
R
Hammond
S
Cooke
J
Ansell
J
Radiographers as film readers in screening mammography: an assessment of competence under test and screening conditions
Br J Radiol
 , 
1996
, vol. 
69
 (pg. 
10
-
4
)
(4)
Pauli
R
Hammond
S
Cooke
J
Ansell
J
Comparison of radiographer/radiologist double film reading with single reading in breast cancer screening
J Med Screen
 , 
1996
, vol. 
3
 (pg. 
18
-
22
)
(5)
Tonita
JM
Hillis
JP
Lim
C
Medical radiologic technologist review: effects on a population-based breast cancer screening program
Radiology
 , 
1999
, vol. 
211
 (pg. 
529
-
33
)
(6)
Sumkin
JH
Klaman
HM
Graham
M
Ruskauff
T
Gennari
RC
King
JL
, et al.  . 
Prescreening mammography by technologists: a preliminary assessment
AJR Am J Roentgenol
 , 
2003
, vol. 
180
 (pg. 
253
-
6
)
(7)
Wivell
G
Denton
ERE
Eve
CB
Inglis
JC
Harvey
I
Can radiographers read screening mammograms?
Clin Radiol
 , 
2003
, vol. 
58
 (pg. 
63
-
7
)
(8)
Duijm
LEM
Groenewoud
JH
Hendriks
JHCL
de Koning
HJ
Independent double reading of screening mammograms in the Netherlands: effect of arbitration following reader disagreements
Radiology
 , 
2004
, vol. 
231
 (pg. 
564
-
70
)
(9)
UICC(International Union against Cancer)
TNM classification of malignant tumours
 , 
1987
4th rev. ed
Berlin (Germany)
Springer Verlag
(10)
NETB (National Evaluation Team for Breast Cancer Screening)
National evaluation of breast cancer screening in the Netherlands (XI) [in Dutch, summary in English]
 , 
2005
Rotterdam (The Netherlands)
Department of Public Health, Erasmus Medical Centre
(11)
Smith-Bindman
R
Chu
PW
Miglioretti
DL
Sickles
EA
Blanks
R
Ballard-Barbash
R
, et al.  . 
Comparison of screening mammography in the United States and the United Kingdom
JAMA
 , 
2003
, vol. 
290
 (pg. 
2129
-
37
)
(12)
Otten
JD
Karssemeijer
N
Hendriks
JHCL
Groenewoud
JH
Fracheboud
J
Verbeek
ALM
, et al.  . 
Effect of recall rate on earlier screen detection of breast cancers based on the Dutch peformance indicators
J Natl Cancer Inst
 , 
2005
, vol. 
97
 (pg. 
748
-
54
)
(13)
Groenewoud
JH
Otten
JDM
Fracheboud
J
Draisma
G
van Ineveld
BM
Holland
R
, et al.  . 
Cost-effectiveness of different reading and referral strategies in mammography screening in the Netherlands
Breast Cancer Res Treat
 , 
2007
, vol. 
102
 (pg. 
211
-
8
)
(14)
van Dijck
JA
Verbeek
ALM
Hendriks
JHCL
Holland
R
The current detectability of breast cancer in a mammographic screening program. A review of the previous mammograms of interval and screen-detected cancers
Cancer
 , 
1993
, vol. 
72
 (pg. 
1933
-
8
)
(15)
Duijm
LEM
Groenewoud
JH
Jansen
FH
Fracheboud
J
van Beek
M
de Koning
HJ
Mammography screening in the Netherlands: delay in the diagnosis of breast cancer after breast cancer screening
Br J Cancer
 , 
2004
, vol. 
91
 (pg. 
1795
-
9
)
None of the authors of this study have a conflict of interest.
We thank the technologists and radiologists who participated in this study, as well as the staff of secretarial office for their support in data acquisition. We thank C. W. N. Looman for his statistical advice.