Disclosure Comparison of Digital Breast Research grant : GE - - PowerPoint PPT Presentation

6/7/2017 1 Comparison of Digital Breast Tomosynthesis and Ultrasound in Diagnosis of Breast Cancer: A Prospective Performance Study

Jung Min Chang Seoul National University Hospital

Disclosure

• Research grant : GE Electronic

Contents

• Screening Mammography in dense breasts
• Diagnostic performance of screening breast

US

• Diagnostic performance of DBT
• Results of comparison studies of DBT vs. US

Screening Mammography

• MG remains mainstay of imaging modality
• Conventional MG in dense breasts has known

limitations – Early studies showed sensitivity decreasing to 30% to 48% in dense breasts compared with 80% to 98% in fatty breasts – More recent study showing 57.1% sensitivity in dense breasts and up to 92.7% in fatty breasts

• 20%-30% of breast cancers not detected in MG

6/7/2017 2

Dense Breast Issue

• Tumors obscured by superimposed breast tissue
• Imperfect sensitivity of MG has led to the use of

adjunctive imaging, including US, DBT, MRI

• In a study of 335 screening US–detected cancers,

81% (272 out of 335) were not seen at MG, even in retrospect *

* Bae MS et al. Radiology 2014

F/46 IDC F/49 DCIS F/47 IDC F/37 IDC

Whole Breast US Screening Whole Breast US

• Well tolerated by patients
• Performance

– Supplemental CDR: 3-4/1000

• Incidence vs. Prevalence, patient characteristics

– Using advanced technology s/a Doppler, Elastography – Better characterization for lesion in the breast

6/7/2017 3

F/47, screening

IDC, GR 2, HR+

J-Start

Japan Strategic Anti-cancer Randomized Trial

• 2007-2011
• 72,998 asymptomatic women, ages 40-49y
• 2 screening rounds/ 2 years
• Intervention group

– MG + HH WBUS – n = 36,752

• Control group

– MG only – n = 35,965

Ohuchi N, et al. Lancet 2016

CDR, Recall and PPV of MG vs. MG+US screening in women aged 40-49 years (J-START)

Ohuchi N, et al. Lancet 2016

Intervention group (n=36,752) Control group (n=35,965)

CDR 184 (5/1000) 117 (3.2/1000) Recall 4647(12.6%) 3153 (8.8%) Biopsy 1665 (4.5%) 655 (1.8%) Interval cancer 18 (0.05%) 35 (0.10%)

More frequently clinical stage 0 & I in the intervention group than in the control group (144 [71·3%] vs 79 [52·0%], p=0·0194).

6/7/2017 4

Characteristics of screening US-detected cancers

• Screening US-detected cancers*

– Small invasive cancer without lymph node metastasis – Luminal A subtype >> HER-2 subtype – BI-RADS assessment category 4A

* Bae MS, et al. Cancer Sci 2011 Oct

Breast Cancers Detected at Screening US: Survival Rates

501 women (median age, 47; 425 invasive) Stage 0 (15.2%), I (70.3%), II (13%) and III (1.6%). No deaths but 15 (3.0%) recurrences confined to the ipsilateral or contralateral breast at a median follow-up of 7 years (range, 5-12 yr). The 5-year survival and recurrence-free survival (RFS) was 100% and 98% (96.8-99.2%).

Kim SY, et al. Radiology 2017

Annual Screening US to MG

• Increased cancer detection rate sustainable?

In elevated breast cancer risk

CDR (per 1000) MG alone MG+US Diff (MG+US vs. MG)

No./ Total of Women Estimate (95% CI) No./ Total of Women Estimate (95% CI) Estimate (95% CI) P value Screen 1 20/2659 7.5 (4.6-11.6) 34/2659 12.8 (8.9-17.8) 5.3 (2.1-8.4) <.001 Screen 2,3 39/4814 8.1 (5.8-11.1) 57/4814 11.8 (9.0-15.3) 3.7 (2.1-5.8) <.001 Berg et al. JAMA 2012

Screening with Handheld US

Operator dependency High false positive rate Standardization Time-consuming

• Widely available
• Low cost (portable)
• No radiation
• No compression

6/7/2017 5

ABUS screening

• Reproducible, Diminished operator dependence
• Better demonstration of the breast anatomy and

documenting the lesion orientation

• CAD detection and diagnosis of breast lesions
• Time-efficient for radiologists
• Additional 3.6 cancers per 1000 women screened

when ABUS was added to mammography in dense breasts or high risk *

• Additional 1.9 cancers per 1000 women screened in

dense breast women **

* Kelly et al. European Radiology 2010 ** Brem et al. Radiology 2015

Digital Breast Tomosynthesis Digital Breast Tomosynthesis

• Emerging technique, allows the breasts to be

viewed in 3D information reducing superimposition

• f breast tissue
• Requires the same equipment as two-dimensional

(2D) MG, with only a few additional seconds required for single-view acquisition.

Supplemental role of DBT

• Widely used
• Additional cancer detection by using DBT after

negative mammography: 1- 2.7 / 1000 screens

• Decreased recall rate in the range of 1.6 - 4.5 %

6/7/2017 6

Comparison of Digital Mammography Alone and Digital Mammography Plus Tomosynthesis in a Population-based Screening Program

• 12,631 women (2010.11-2011.12)

Skaane P et al. Radiology 2014

Breast cancer screening with DBT

• Larger multicenter study on breast cancer screening

using DBT in combination with DM.

• About 500,000 patients
• Largest volume study reflecting all diversity of practice

Friedewald SM et al. JAMA. 2014

DM (n=281,187)

(start date 2010.3-2011.10 through DBT implementation)

DM+DBT (n=173,663)

(start date 2011.3-2012.10 through 2012.12.31)

• Cancer detection rate
• RR for additional imaging
• PPV for recall and biopsy

Breast cancer screening with DBT

Per 1000 screens DM DM+DBT Difference Recall rate 107 91 –16 (P < .001) Biopsy 18.1 19.3 1.3 (P = .004) Cancer detection 4.2 5.4 1.2 (P < .001) Invasive cancer 2.9 4.1 1.2 (P < .001). DCIS 1.4 1.4 PPV for recall 4.3 6.4 2.1 (P < .001) PPV for biopsy 24.2 29.2 5.0 (P < .001). Friedewald SM et al. JAMA. 2014;311(24): 2499-2507

Annual DBT to MG

• Improved outcomes (CDR , RR ) observed

after initial implementation of DBT screening are sustainable over time?

ES McDonald et al. JAMA oncol. 2016 Characteristic

DM DBT 1 P value DBT 2 P value DBT 3 P value

Recall n/1000 104 88 <.0001 90 .0005 92 .0025 Biopsy n/1000 18 20 .167 19 .606 19 .597 Cancer n/1000 4.6 5.5 .370 5.8 .196 6.1 .110

6/7/2017 7

F/51, pain in left breast

2D FFDM

Lesion Characterization with DBT

Rt: Invasive ductal carcinoma, Lt: Fibroadenoma Capture slices of DBT

Detection of MG occult architectural distortion on DBT screening: Initial clinical experience

• Retrospective review of BI-RADS category 0 reports from 9982

screening DM examinations with adjunct DBT were searched for the term "architectural distortion" and were reviewed in consensus by three radiologists.

• ADs were classified by whether they were seen better on DM or

DBT, were seen equally well on both, or were occult on either modality.

• The electronic medical record was reviewed to identify

additional imaging studies, biopsy results, and surgical excision pathology results.

Partyka et al. AJR 2014

Detection of MG occult architectural distortion on DBT screening: Initial clinical experience

• Review identified 26 cases of AD

– 19 (73%) of which were seen only on the DBT images. – 6 seen better on DBT than DM. – On diagnostic workup, 9 were assigned to BI-RADS category 4 or 5.

• 2 IDC, 2 DCIS, 3 Radial scars, 2 atypia.
• CDR of DBT in MG occult AD was 21% (4/19).
• US not visible AD: more likely radial sclerosing lesion than

carcinoma

Partyka et al. AJR 2014

6/7/2017 8

Asx, F/51

Lobular carcinoma in situ

2014 Jan MG 2015 Jan MG 2015 Jan, Tomo 2015 July, Tomo

Asx, 50/F

Infiltrating ductal carcinoma, histologic grade I

Marker insertion after US guided vacuum biopsy

IDC, Grade I

Digital Breast Tomosynthesis

• Limitation

– US after DBT still detect a few cancers – The factors affecting DBT performances has not been well investigated » Tumor factors : size, subtype, tumor shape, margin, etc.. » Patient factors: age, breast density, etc..

6/7/2017 9

Factors affecting Lesion Visibility on DBT

• Analysis on 106 non calcified T1 stage invasive

cancers

– Median size: 0.8cm (range 0.4 - 2cm) – IDC (n=97), ILC (n=5), others (n=4) – ER positive tumors (n=86, 81.1%), HER2-enriched (n=10, 9.4%), TN (n=10, 9.4%)

• Visibility score assessment by 2 radiologists
• Factors affecting visibility scores
• Diagnostic performance evaluation according to

the factors affecting visibility scores using normal control data

Chang et al. RSNA 2016 presentation

Visibility scores on DBT

Poorly Visible Fairly visible Definitely visible FFDM only 56 13 37 FFDM+DBT 22 11 73

Chang et al. RSNA 2016 presentation

Variables correlating with Visibility scores

Imaging variables Total (n=106) Visibility Score on DBT+DM Poorly visible (n = 22) Fairly visible (n = 11) Definitely visible (n = 73) P-value Breast density Grade a

22 (20.8) 0 (0) 1 (9.1) 21 (28.8) <.001

Grade b

18 (17.0) 0 (0) 1 (9.1) 17 (23.3)

Grade c

44 (41.5) 5 (22.7) 9 (81.8) 30 (41.1)

Grade d

22 (20.8) 17 (77.3) 0 (0) 5 (6.8)

Lesion density Iso

37 (34.9) 14 (63.6) 3 (27.3) 20 (27.4) .006

Hyper

69 (65.1) 8 (36.4) 8 (72.7) 53 (72.6)

Lesion type Mass

13 (12.3) 3 (13.7) 3 (27.3) 7 (9.6) .014

Asymmetry

72 (67.9) 18 (81.8) 8 (72.7) 46 (63.1)

Architectural distortion

21 (19.8) 1 (4.5) 0 (0) 20 (27.4)

Chang et al. RSNA 2016 presentation

Visibility scores on DBT

Visibility scores

• n DM alone
• n DM+DBT

P value Breast density a,b,c (not extremely dense) 1.3±0.2 2.0±0.3 P = .002 Breast density d (extremely dense) 1.2±0.3 1.4±0.6 P = .382 Variable* Multivariate Analysis† Odds ratio 95% CI P-value Breast density: grade d 0.02 0.04 – 0.09 <.001 Lesion density: isodense 0.29 0.07 – 1.15 .203 Lesion type: architectural distorsion 2.73 0.58-12.78 .078 Chang et al. RSNA 2016 presentation

6/7/2017 10

DBT Performances according to Breast density

• 106 cancers combined with 159 controls

with same breast density

Diagnostic performance (%) Extremely dense breast (n=55) (22 tumors + 33 case controls) Not extremely dense breasts (n=210) (84 tumors + 126 case controls) FFDM alone DBT+FFDM P-value FFDM alone DBT+FFDM P-value Sensitivity 59.1 (13/22) 63.6 (14/22) .642 90.5 (76/84) 95.2 (80/84) .451 Specificity 75.8 (25/33) 84.8 (28/33) .463 81.7 (103/126) 98.4 (124/126) .002 PPV 61.9 (13/21) 79.2 (19/24) .078 76.8 (76/99) 97.6 (80/82) <.001 NPV 73.5 (25/34) 90.3 (28/31) .072 92.8 (103/111) 96.9 (124/128) .589 Chang et al. RSNA 2016 presentation F/40, asymptomatic, Gr c density, IDC, Gr 3, ER 80%/PR 30%/HER2- Mass with irregular shape and spiculated margin (Visibility score 3) F/40, asymptomatic, Gr d density, IDC, 0.5cm, Gr 2, ER>95%/PR>95%/HER2- Non-visible (Visibility score 0)

Breast density and DBT

• Extremely dense breast was significantly associated with

poor visibility and low diagnostic performance of DBT combined with DM in patients with noncalcified T1 invasive breast cancers.

• In patients with extremely dense breasts, DBT combined

with DM may not have added value compared to DM alone for the diagnosis of noncalcified T1 breast cancers.

6/7/2017 11

MG only? MG + Tomo? and then US or not? MG + US?

No Tomo?

MG+Tomo+US

Additional US or DBT after DM: which one is the best combination?

• Retrospective study in an enriched sample of 1042 patients

– 84 malignant lesions, 258 benign lesions, 700 normal – The readers categorized the cases as benign (BI-RADS 1 or 2) or malignant (BI-RADS 3–5) for DM and the different combination of techniques. – The sensitivity (SE), specificity (SP) and the ROC curves evaluation

Elizalde et al. Acta Radiol. 2016; 57:13-18.

Sensitivity and Specificity of DM and the combinations with DBT and/or US

DM DM+US DM+DBT DM+US+DBT Sensitivity 69.05% 92.86% 86.90% 98.81% Specificity 88.20% 74.32% 83.50% 74.11%

Elizalde et al. Acta Radiol. 2016; 57:13-18.

Technique AUC P value DM vs. DM+DBT 0.70 vs.0.84 0.0125 DM vs. DM+US 0.70 vs 0.85 0.0044 DM vs. DM+US+DBT 0.70 vs.0.91 0.0000 DM+US vs. DM+DBT 0.85 vs.0.85 0.7332

ASTOUND study

• Adjunctive

screening with Tomo vs. US in MG negative women c dense breast

Tagliafico AS et al, JCO 2016

6/7/2017 12

ASTOUND study

• Prospective, multicenter study
• 3,231 MG-negative screening participants
• 24 breast cancers detected (23 invasive)

– 13 DBT detected BCs (CDR, 4.0 per 1,000 screens) – 23 US detected BCs (CDR, 7.1 per 1,000 screens)

• FP recall did not differ between DBT (FP = 53) and

US (FP = 65), P = .26

Tagliafico AS et al, JCO 2016

SNUH results: Comparison of DM+DBT vs. DM+US

• Prospective study comparing DBT vs. US
• Single institution study
• IRB approved, written informed consent obtained
• ClinicalTrials.gov (trial number: NCT01910103)
• Showing non-inferiority of DBT to US

Kim WH, Chang JM, et al. Breast Cancer Res Treat. 2017

Participants

• Eligibility Criteria

– At least 20 years – Gr c or d density in at least 1 quadrant – Scheduled for MG, US, DBT (screening, diagnostic) – No prior breast imaging in our institution – No history of surgical excision in the breast – No implants, pregnant, lactating

Sample size consideration

• Non-inferiority study design
• Based on our clinical experience, we expected the AUC of US

to lie within the range of 0.90–0.95, and we considered 0.80– 0.85 as an approximate range for clinically acceptable AUC for DBT in a clinical setting

– US appears to exhibit superior performance over DBT in terms of lesion visualization and characterization with supplementary Doppler sonography and elastography – We have over 10 years of experience in breast cancer screening by US, DBT has only been employed at our clinical practice for 1–2 years prior to this prospective study

• Based on our clinical practice and relevant data form literatures,

absolute difference in AUC was set at 0.1.

6/7/2017 13

Imaging Techniques

• FFDM unit with integrated DBT acquisition

– Selenia Dimensions; Hologic, Bedford, MA, USA – Bilateral two views (CC and MLO) in Combo M.

• Bilateral hand-held, whole-breast US

– By 1 of 12 radiologists – Various equipment with at least 12MHz transducers

Study design and Image interpretation

• 778 women enrolled
• Each participant underwent DBT

and US

• 2 different radiologists interpret

image examination without information of other imaging modality

• Rated probability for malignancy

– BI-RADS – Likelihood of malignancy from 0% to 100%

• Patient management was based
• n recommendations from the

integrated examination results by another senior radiologist

Statistical Analysis

• Primary end point: AUC difference with

noninferiority margin=0.1

• Secondary end point

– Sensitivity, Specificity – PPV, NPV – Comparison of Cancer Characteristics

• ROC, McNemar, Fisher-exact tests

Results: Participants & Cancers

• 698 included for analysis

– Mean age, 49 years (ranges, 20 – 88 years) – Asymptomatic (70%), palpable (22%), others (8%) – BI-RADS composition c (77.1%), BI-RADS composition d (22.9%)

• 140 cancers

– 108 IDC, 21 DCIS – Mean size, 1.8cm (ranges, 0.1cm – 6.5cm)

6/7/2017 14

Overall outcome difference: Tomosynthesis - Ultrasound

• ∆=-0.1

Noninferior Inferior BI-RADS: -0.00456, -0.0568 POM: -0.00677, -0.0615

Since the lower boundary of the 95% CI of absolute difference lay above the predefined noninferiority margin (0.1), noninferiority of tomosynthesis was established

SNUH results: Diagnostic Performance DBT vs. US SNUH results: Diagnostic Performance DBT vs. US

DBT US P value

AUC 0.933 (0.912–0.950) 0.964 (0.947–0.976) 0.014 Sensitivity 91.4 (86.8 to 96.0) 96.4 (93.3 to 99.5) .039 Specificity 83.9 (80.9 to 86.9) 70.4 (66.6 to 74.2) <.001 PPV 58.7 (52.2 to 65.2) 45.0 (39.4 to 50.6) <.001 NPV 97.5 (96.1 to 98.9) 98.7 (97.6 to 99.8) .051

SNUH results: Diagnostic Performance DBT vs. US

DBT US Difference P value All participants 0.933 (0.912 - 0.950) 0.964 (0.947- 0.976)

• 0.031 (-0.055 to -0.006)

.014 Breast density Heterogeneously dense (n = 538) 0.949 (0.927 to 0.966) 0.969 (0.951 to 0.982)

• 0.020 (-0.042 to 0.002)

.076 Extremely dense (n = 160) 0.842 (0.776 to 0.895) 0.931 (0.879 to 0.965)

• 0.089 (-0.191 to 0.014)

.091 Presence of symptom Asymptomatic (n = 484) 0.912 (0.883 to 0.936) 0.934 (0.908 to 0.955)

• 0.022 (-0.075 to 0.030)

.403 Symptomatic (n = 214) 0.937 (0.896 to 0.966) 0.973 (0.941 to 0.990)

• 0.036 (-0.067 to -0.005)

.023 Purpose of examination Screening (n = 192) 0.987 (0.959 to 0.998) 0.950 (0.909 to 0.976) 0.037 (-0.029 to 0.104) .270 Diagnostic (n = 506) 0.921 (0.894 to 0.943) 0.958 (0.937 to 0.974)

• 0.037 (-0.063 to -0.011)

.005

6/7/2017 15 Summary of Cancer Characteristics

Not diagnosed at Both Diagnosed at only DBT Diagnosed at only US Diagnosed at Both

• No. of cancers

4 2 8 120 Invasive cancers, No. (%) 2 (50%) 1 (50%) 5 (63%) 101 (84%) Mean size 0.65cm 2.5cm 1.5cm 1.8cm

F/53 Spot magni CC view After US guided needle localization Tomo CC view 0.6cm grade I IDC

6/7/2017 16

F/38, Asx

RUO 1cm IDC & DCIS(2cm) Gr I, ER positive

SNUH Results: DM+DBT vs. DM+US

• DBT has overall comparable

performances to US as an adjunct to MG

• Higher CDR with US
• US > DBT: in extremely dense breasts

Utility of Whole Breast Screening Ultrasound in Women Undergoing Digital Breast Tomosynthesis

• Preliminary results (ClinicalTrials.gov identifier:

NCT02174406)

• 837 women (mean age 56 years), 77% dense breasts
• MG+ DBT and US performed at same visit

– Interpreted by 2 radiologists – Technologist based US program

Sung et al. 2016 RSNA presentation

• 2D MG recalls decreased after DBT
• DBT and WBUS recalls were added

Sung et al. 2016 RSNA presentation

6/7/2017 17

Sung et al. 2016 RSNA presentation Sung et al. 2016 RSNA presentation

Utility of Whole Breast Screening Ultrasound in Women Undergoing Digital Breast Tomosynthesis

• 7 Cancer detected (8.3/1000)

– 5 (DM, DBT, US), 1 (DBT, US), 1 (DBT)

• Both DBT and WBUS increase the cancer detection rate

compared to 2D MG

• DBT has added benefit of decreased MG recall rate and shows

fewer false positive biopsy recommendations and lower BI- RADS 3 rate compared to US

• Clinical value of US reduced in women undergoing DBT

compared to 2D MG

Sung et al. 2016 RSNA presentation

Comparison of cancers detected by Screening US and DBT

• Retrospective review
• 7146 screening US, DBT(2013.1-2016.6)
• 39 cancers detected

– Mean size 1.7cm (range 0.3 - 4.5cm) – LN positive 8% – 69% mass, 13% calcifications, 8% AD – 70% invasive cancer (30/39)

Destounis S et al. 2017 ARRS presentation

6/7/2017 18 Comparison of cancers detected by Screening US and DBT

• 39 cancers detected (mean size 1.7cm)

– DBT+US: 18/39 – 1.5cm average size at excision (0.3-4cm)

Destounis S et al. 2017 ARRS presentation

Comparison of cancers detected by Screening US and DBT

• US only :17/39

– Majority: invasive cancer (13/17) – Mean size at excision: 1.3cm (0.6-2cm)

• DBT only : 4/39

– All calcifications – 3 DCIS, 1 IDC – Mean size at excision: 2.7cm (1.5-4.5cm)

• Smaller average size of US only detected lesions
• US overall performed better than DBT for invasive

cancer detection

Destounis S et al. 2017 ARRS presentation

Preoperative assessment of Breast cancer

• Accuracy of preoperative assessment of breast cancer per lesion base
• DM, DBT, US, MRI, combination
• Results

– Sensitivity

• DBT > DM (90.7% vs. 85.2%)
• DM + DBT + US: 97.7%, vs. MRI (98.8%)

– Overall accuracy

• MRI : 92.3%, DM + DBT + US : 93.7%

– Breast density affected sensitivity of DM and DBT, not MRI

• Conclusion

– little gain in sensitivity and no gain in overall accuracy, by performing MRI for patients who have been evaluated with DM with DBT and US

Mariscotti et al. Anticancer Res 2014

DBT to Characterize MRI-Detected Additional Lesions Unidentified at Targeted US in Newly Diagnosed Breast Cancer Patients

• DBT helpful to identify MR detected lesion not identified
• n US?
• In 520 patients, MR identified 164 (in 114 women, 22%)

additional enhancing lesions – Targeted US identified 114/164 (69.5%) of these, whereas 50/164 (30.5%) remained unidentified – DBT identified 32/50 of these cases, increasing the

• verall characterization of MR detected additional

findings to 89.0% (146/164)

Mariscotti et al. Eur Radiol 2015

6/7/2017 19

DBT to Characterize MRI-Detected Additional Lesions Unidentified at Targeted US in Newly Diagnosed Breast Cancer Patients

• DBT identified lesions
• Significantly more likely to be malignant than benign

(P = 0.04)

• Conclusion

– DBT improves the characterization of additional MR findings not identified at targeted breast US in preoperative breast cancer staging – DBT found lesions are more significant lesions

Mariscotti et al. Eur Radiol 2015

Additional MRI detected Suspicious Lesions in Known Patients with Breast Cancer: Comparison of Second-Look DBT and US

• 55 breast cancer patients (mean 52.5 yrs) with MRI, DBT, US
• 37 additional MR detected suspicious lesions

– 27 were detected on DBT and/or US

• 2 were detected on both DBT and US (IDC, DCIS)
• 1 was detected only on DBT (Radial scar)
• 24 were detected only on US (10 malignancy, 14 benign)

– US detected more additional suspicious lesions than DBT (P < 0.0001).

Ko et al. Ultrasound Quarterly 2017

F/34, IDC MG DBT MR

Summary

• Supplemental US or DBT helps finding more

invasive cancers in dense breasts

• In comparison with US, DBT exhibits non-inferior

diagnostic performance as an adjunctive to 2D digital mammography for diagnosis of cancer in women with dense breasts

6/7/2017 20

Summary

• Overall higher sensitivity with lower specificity is

noted for US compared to DBT in screening population.

• The visibility of tumor on DBT in extremely dense

could be still limited.

• DBT and US can be used for problem solving for

MG and MR detected lesions.

Thank you for your attention