-
Notifications
You must be signed in to change notification settings - Fork 0
/
Meta-Analysis of Magnetic Resonance Imaging in Detecting Residual Breast Cancer After Neoadjuvant Therapy.txt
3659 lines (3101 loc) · 321 KB
/
Meta-Analysis of Magnetic Resonance Imaging in Detecting Residual Breast Cancer After Neoadjuvant Therapy.txt
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
http://jnci.oxfordjournals.org/content/105/5/321.full
<!DOCTYPE html
PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html
xmlns="http://www.w3.org/1999/xhtml"
xml:lang="en"
lang="en">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8" />
<title>Meta-Analysis of Magnetic Resonance Imaging in Detecting Residual Breast Cancer After Neoadjuvant Therapy </title>
<meta name="googlebot" content="NOODP" />
<meta name="HW.ad-path" content="/cgi/content/full/105/5/321" />
<meta content="/jnci/105/5/321.atom" name="HW.identifier" />
<meta name="DC.Format" content="text/html" />
<meta name="DC.Language" content="en" />
<meta content="Meta-Analysis of Magnetic Resonance Imaging in Detecting Residual Breast Cancer After Neoadjuvant Therapy"
name="DC.Title" />
<meta content="10.1093/jnci/djs528" name="DC.Identifier" />
<meta content="2013-03-06" name="DC.Date" />
<meta content="Oxford University Press" name="DC.Publisher" />
<meta content="Michael L. Marinovich" name="DC.Contributor" />
<meta content="Nehmat Houssami" name="DC.Contributor" />
<meta content="Petra Macaskill" name="DC.Contributor" />
<meta content="Francesco Sardanelli" name="DC.Contributor" />
<meta content="Les Irwig" name="DC.Contributor" />
<meta content="Eleftherios P. Mamounas" name="DC.Contributor" />
<meta content="Gunter von Minckwitz" name="DC.Contributor" />
<meta content="Meagan E. Brennan" name="DC.Contributor" />
<meta content="Stefano Ciatto" name="DC.Contributor" />
<meta content="Journal of the National Cancer Institute"
name="citation_journal_title" />
<meta content="JNCI J Natl Cancer Inst" name="citation_journal_abbrev" />
<meta content="0027-8874" name="citation_issn" />
<meta content="1460-2105" name="citation_issn" />
<meta name="citation_author" content="Michael L. Marinovich" />
<meta name="citation_author_institution"
content="
Affiliations of authors:
Screening and Test Evaluation Program (STEP), Sydney School of Public Health, University of Sydney, Sydney, Australia (MLM, NH, PM, LI, MEB); Dipartimento di Scienze Biomediche per la Salute, Università degli Studi di Milano, Milan, Italy (FS); Unità di Radiologia, IRCCS Policlinico San Donato, Milan, Italy (FS); Aultman Health Foundation, Canton, OH (EPM); German Breast Group & Universitäts-Frauenklinik Frankfurt, Neu-Isenburg, Germany (GvM); U.O. Senologia Clinica e Screening Mammografico, Dipartimento di Radiodiagnostica, APSS, Trento, Italy (SC)." />
<meta name="citation_author" content="Nehmat Houssami" />
<meta name="citation_author" content="Petra Macaskill" />
<meta name="citation_author" content="Francesco Sardanelli" />
<meta name="citation_author" content="Les Irwig" />
<meta name="citation_author" content="Eleftherios P. Mamounas" />
<meta name="citation_author" content="Gunter von Minckwitz" />
<meta name="citation_author" content="Meagan E. Brennan" />
<meta name="citation_author" content="Stefano Ciatto" />
<meta name="citation_author_institution" content="
†Died May 2012." />
<meta content="Meta-Analysis of Magnetic Resonance Imaging in Detecting Residual Breast Cancer After Neoadjuvant Therapy"
name="citation_title" />
<meta content="03/06/2013" name="citation_date" />
<meta content="105" name="citation_volume" />
<meta content="5" name="citation_issue" />
<meta content="321" name="citation_firstpage" />
<meta content="333" name="citation_lastpage" />
<meta content="105/5/321" name="citation_id" />
<meta content="105/5/321" name="citation_id_from_sass_path" />
<meta content="jnci;105/5/321" name="citation_mjid" />
<meta content="10.1093/jnci/djs528" name="citation_doi" />
<meta content="http://jnci.oxfordjournals.org/content/105/5/321.abstract"
name="citation_abstract_html_url" />
<meta content="http://jnci.oxfordjournals.org/content/105/5/321.full"
name="citation_fulltext_html_url" />
<meta content="http://jnci.oxfordjournals.org/content/105/5/321.full.pdf"
name="citation_pdf_url" />
<meta content="/content/105/5.cover.gif" name="issue_cover_image" />
<meta content="http://jnci.oxfordjournals.org/content/105/5/321"
name="citation_public_url" />
<meta content="23297042" name="citation_pmid" />
<meta name="citation_section" content="Review" />
<meta name="robots" content="noarchive,nofollow" />
<meta name="googlebot" content="noarchive" />
<link href="/content/105/5/316.short" rel="prev" />
<link href="/content/105/5/334.short" rel="next" />
<link rel="stylesheet" type="text/css" media="all" href="/shared/css/hw-global.css" />
<link rel="stylesheet" type="text/css" media="print" href="/shared/css/hw-print.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/local/css/hw-local-global.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/publisher/css/hw-publisher-global.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/publisher/css/hw-publisher-ac.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/shared/css/hw-page-content.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/shared/css/jquery.fancybox-1.3.4.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/shared/css/hw-global-colexpand.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/publisher/css/hw-publisher-page-content.css" /><script type="text/javascript" id="session-d29380190e1">var callbackToken='5376635C83DF83E';</script><script type="text/javascript" id="session-d29380190e3">
var subCode='oupjournal_sub';
</script><script type="text/javascript">
var gAuthTimeStamp = '2015-10-27T12:13:45.868-07:00';
var gSessionId = 'rI3bM0LvTy35dmOHk2Hxcg';
var gAuthzRequired = 'false';
var gAuthnMethods1 = 'ip ip';
var gAuthnMethods2 = 'ip,ip';
var gAuthnIPs = '18.189.59.169,18.189.59.169';
var gAuthnIndividuals = '';
var gAuthnInstitutions = '10005282,10072538,10328193,100274300,10082403,05094000,10093227,153074108,10080507,21879000,10095822,10081742,10082405,10004595,10019502,10082401,07635000,10082404,06602000';
</script><script type="text/javascript" src="/shared/js/jquery-min.js"></script><script type="text/javascript" src="/shared/js/fingerprint.js"></script><script type="text/javascript" src="/shared/js/hw-shared.js"></script><script type="text/javascript" src="/publisher/js/hw-publisher-site.js"></script><script type="text/javascript" src="/local/js/local-js-vars.js"></script><script type="text/javascript" src="/shared/js/pages/hw-content.js"></script><script type="text/javascript" src="/shared/js/fancybox/jquery.fancybox-1.3.4.js"></script><script type="text/javascript" src="/shared/js/fancybox/jquery.easing-1.3.pack.js"></script><script type="text/javascript"
src="/shared/js/fancybox/jquery.mousewheel-3.0.4.pack.js"></script><script type="text/javascript" src="/shared/js/util/content.jquery.addVariantLink.js"></script><script type="text/javascript" src="/shared/js/util/hw-col-expand.js"></script><script type="text/javascript" src="/publisher/js/hw-publisher-modalwin-vars.js"></script><script type="text/javascript"
src="/publisher/js/hw-publisher-article-dynamic-elements.js"></script><script type="text/javascript" src="/publisher/js/hw-publisher-content.js"></script><script type="text/javascript" src="/shared/js/util/hw-mathjax.js"></script><script type="text/x-mathjax-config">
MathJax.Hub.Config({
tex2jax: {
inlineMath: [["$","$"],["\\(","\\)"]],
processClass: "tex2jax_process|mathjax"
}
});
MathJax.Hub.Queue(function() {
gColTempResize = true;
fixColHeights(1);
gColTempResize = false;
});
</script><script type="text/javascript"
src="http://cdn.mathjax.org/mathjax/latest/MathJax.js?config=TeX-AMS-MML_HTMLorMML"></script><script type="text/javascript">
var siqDOI = encodeURIComponent("10.1093/jnci/djs528");
var siqIsOpenAccess = encodeURIComponent("");
var siqPubDate = encodeURIComponent("20130306");
if (siqDOI.length == 0) {
siqDOI = "UNKNOWN";
}
if (gAuthnIndividuals.length != 0) {
if (gAuthnInstitutions.length != 0) {
authnEntity = encodeURIComponent(gAuthnIndividuals
+ ',' + gAuthnInstitutions);
} else {
authnEntity = encodeURIComponent(gAuthnIndividuals);
}
} else {
authnEntity = encodeURIComponent(gAuthnInstitutions);
}
var commonString =
'authSessionId=' + gSessionId + String.fromCharCode(0x26)
+ 'authzRequired=' + gAuthzRequired + String.fromCharCode(0x26)
+ 'authentication_method=' + encodeURIComponent(gAuthnMethods2) + String.fromCharCode(0x26)
+ 'authnIPs=' + gAuthnIPs + String.fromCharCode(0x26)
+ 'authnInstitutions=' + authnEntity;
var gPageId = "pageid-content";
var gVariant = "full-text";
// Not completely done
var eventType = "full-text";
var accessType;
if (siqIsOpenAccess == 'true') {
accessType = 'SOA';
} else {
accessType = 'subscription';
}
var NTPT_PGEXTRA =
commonString + String.fromCharCode(0x26)
+ 'event_type=' + eventType + String.fromCharCode(0x26)
+ 'publication_date=' + siqPubDate + String.fromCharCode(0x26)
+ 'access_type=' + accessType + String.fromCharCode(0x26)
+ 'doi=' + siqDOI ;
// alert("NTPT_PGEXTRA is " + NTPT_PGEXTRA);
</script><link rel="stylesheet" type="text/css" media="all" href="/resource/css/hw20.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/resource/css/h20-oup-header.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/resource/css/h20-oup-footer.css" />
<link rel="stylesheet" type="text/css" media="all"
href="/resource/css/h20-oup-sidebars.css" />
<link type="text/css" media="all" rel="stylesheet" href="/site/resource/jnci_h20.css" />
<link rel="stylesheet" type="text/css" href="/resource/css/print.css" media="print" /><script type="text/javascript" src="/resource/js/main.js"></script><link rel="stylesheet" type="text/css" media="all"
href="http://oi-underbar.ifactory.com/underbar/css/pf_oiunderbar.css" />
<link rel="stylesheet" type="text/css" media="screen"
href="http://gab.cookie.oup.com/aws-cookie/jquery.fancybox-1.3.4_1.css" /><script type="text/javascript" src="/resource/js/oup_ad_size.js"></script></head>
<body class="general_page journal jnci">
<div class="hw-gen-page pagetype-content hw-pub-id-article" id="pageid-content"
itemscope="itemscope"
itemtype="http://schema.org/ScholarlyArticle">
<noscript>
<p>We use cookies to enhance your experience on our website. By continuing to use our website, you are agreeing to our use of
cookies. You can change your cookie settings at any time. <a href="http://global.oup.com/cookiepolicy/">Find out more</a></p>
</noscript>
<p class="hide">
<a href="#content">Skip Navigation</a>
</p>
<div id="secondary_nav">
<strong id="page_logo" title="Oxford Journals"><a href="http://www.oxfordjournals.org/"><span>Oxford Journals</span></a></strong>
<ul>
<li id="nav_contact_us" title="Contact Us"><a href="http://www.oxfordjournals.org/contact_us.html"><span>Contact Us</span></a></li>
<li id="nav_my_basket" title="My Basket"><a href="https://secure.oxfordjournals.org/basket.html"><span>My Basket</span></a></li>
<li id="nav_my_account" title="My Account"><a href="http://services.oxfordjournals.org/cgi/tslogin?url=http://www.oxfordjournals.org/service/Register"><span>My
Account</span></a></li>
</ul>
</div>
<div id="header">
<h1 id="page_title" title="JNCI J Natl Cancer Inst"><a href="/"><span>
JNCI J Natl Cancer Inst
</span></a></h1>
</div>
<div id="primary_nav">
<ul>
<li id="nav_about_this_journal" title="About This Journal">
<a href="http://www.oxfordjournals.org/jnci/about.html">
<span>About This Journal</span>
</a>
</li>
<li id="nav_contact_this_journal" title="Contact This Journal"><a href="/cgi/feedback/"><span>Contact This Journal</span></a></li>
<li id="nav_subscriptions" title="Subscriptions">
<a href="http://www.oxfordjournals.org/jnci/access_purchase/buy_online.html">
<span>Subscriptions</span>
</a>
</li>
<li id="nav_current_issue" title="Current"><a href="/content/current"><span>View Current Issue (Volume 107 Issue 10 October 2015)</span></a></li>
<li id="nav_archive" title="Archive"><a href="/content/by/year"><span>Archive</span></a></li>
<li id="nav_search" title="Search"><a href="/search"><span>Search</span></a></li>
</ul>
</div>
<div id="user_nav">
<div class="header-ac-elements">
<div id="authstring">
<ul>
<li class="subscr-ref">Institution: MIT Libraries</li>
<li class="no-left-border">
<a href="/login?uri=http%3A%2F%2Fjnci.oxfordjournals.org%2Fcontent%2F105%2F5%2F321.full">
Sign In as Personal Subscriber
</a>
</li>
</ul>
</div>
</div>
</div>
<ul id="site-breadcrumbs">
<li class="first"><a href="http://services.oxfordjournals.org/cgi/tslogin?url=http%3A%2F%2Fwww.oxfordjournals.org">Oxford Journals</a></li>
<li><span class="breadcrumb_subjects"><a href="http://www.oxfordjournals.org/subject/medicine/"
class="breadcrumb_subject">Medicine & Health</a></span></li>
<li><a href="/">JNCI J Natl Cancer Inst</a></li>
<li><a href="/content/105/5.toc">
<span xmlns="" class="volume-value">Volume 105</span>
<span xmlns="" class="issue-value"> Issue 5</span></a></li>
<li>Pp. <span class="slug-pages">
321-333.
</span></li>
</ul>
<div id="oas_top" class="ad_hidden">
<iframe id="id_advertframe_top"
src="/resource/htmlfiles/advert.html?p=Top&u=jnci.oxfordjournals.org/content/105/5/321.full"
width="0"
height="0"
frameborder="0"
scrolling="no"
class="resize_ad"></iframe>
</div>
<a name="content"></a>
<div id="h20_page"></div>
<div id="content-block">
<div class="article fulltext-view " itemprop="articleBody"><span class="highwire-journal-article-marker-start"></span><h1 id="article-title-1" itemprop="headline">Meta-Analysis of Magnetic Resonance Imaging in Detecting Residual Breast Cancer After Neoadjuvant Therapy</h1>
<div class="contributors">
<ol class="contributor-list" id="contrib-group-1">
<li class="contributor" id="contrib-1" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Michael+L.+Marinovich&sortspec=date&submit=Submit">Michael L. Marinovich</a></span>,
</li>
<li class="contributor" id="contrib-2" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Nehmat+Houssami&sortspec=date&submit=Submit">Nehmat Houssami</a></span>,
</li>
<li class="contributor" id="contrib-3" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Petra+Macaskill&sortspec=date&submit=Submit">Petra Macaskill</a></span>,
</li>
<li class="contributor" id="contrib-4" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Francesco+Sardanelli&sortspec=date&submit=Submit">Francesco Sardanelli</a></span>,
</li>
<li class="contributor" id="contrib-5" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Les+Irwig&sortspec=date&submit=Submit">Les Irwig</a></span>,
</li>
<li class="contributor" id="contrib-6" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Eleftherios+P.+Mamounas&sortspec=date&submit=Submit">Eleftherios P. Mamounas</a></span>,
</li>
<li class="contributor" id="contrib-7" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Gunter+von+Minckwitz&sortspec=date&submit=Submit">Gunter von Minckwitz</a></span>,
</li>
<li class="contributor" id="contrib-8" itemprop="author" itemscope="itemscope"
itemtype="http://schema.org/Person"><span class="name" itemprop="name"><a class="name-search"
href="/search?author1=Meagan+E.+Brennan&sortspec=date&submit=Submit">Meagan E. Brennan</a></span> and
</li>
<li class="last" id="contrib-9"><span class="name"><a class="name-search"
href="/search?author1=Stefano+Ciatto&sortspec=date&submit=Submit">Stefano Ciatto</a></span><a id="xref-aff-2-1" class="xref-aff" href="#aff-2"><sup>†</sup></a></li>
</ol>
<ol class="affiliation-list">
<li class="aff"><a id="aff-1" name="aff-1"></a><address>
<strong>Affiliations of authors:</strong>
Screening and Test Evaluation Program (STEP), Sydney School of Public Health, University of Sydney, Sydney, Australia (MLM,
NH, PM, LI, MEB); Dipartimento di Scienze Biomediche per la Salute, Università degli Studi di Milano, <span class="addr-line">Milan</span>, Italy (FS); Unità di Radiologia, IRCCS Policlinico <span class="addr-line">San Donato, Milan</span>, Italy (FS); Aultman Health Foundation, Canton, OH (EPM); German Breast Group & Universitäts-Frauenklinik Frankfurt, <span class="addr-line">Neu-Isenburg</span>, Germany (GvM); U.O. Senologia Clinica e Screening Mammografico, Dipartimento di Radiodiagnostica, <span class="addr-line">APSS, Trento</span>, Italy (SC).
</address>
</li>
<li class="aff"><a id="aff-2" name="aff-2"></a><address>
†Died May 2012.
</address>
</li>
</ol>
<ol class="corresp-list">
<li class="corresp" id="corresp-1">Correspondence to: Michael Luke Marinovich, MPH, Screening and Test Evaluation Program, Sydney School of Public Health, A27,
Edward Ford Building, University of Sydney, NSW 2006, Australia (e-mail: <span class="em-link"><span class="em-addr">luke.marinovich{at}sydney.edu.au</span></span>).
</li>
</ol>
<ul class="history-list">
<li xmlns:hwp="http://schema.highwire.org/Journal" class="rev-recd"
hwp:start="2012-11-05"><span class="rev-recd-label">Revision received </span>November 5, 2012.
</li>
<li xmlns:hwp="http://schema.highwire.org/Journal" class="accepted"
hwp:start="2012-11-06"><span class="accepted-label">Accepted </span>November 6, 2012.
</li>
</ul>
</div>
<div class="section abstract" id="abstract-1" itemprop="description">
<div class="section-nav">
<div class="nav-placeholder"> </div><a href="#sec-5" title="Methods" class="next-section-link"><span>Next Section</span></a></div>
<h2>Abstract</h2>
<div id="sec-1" class="subsection">
<p id="p-1"><strong>Background</strong> It has been proposed that magnetic resonance imaging (MRI) be used to guide breast cancer surgery by differentiating residual
tumor from pathologic complete response (pCR) after neoadjuvant chemotherapy. This meta-analysis examines MRI accuracy in
detecting residual tumor, investigates variables potentially affecting MRI performance, and compares MRI with other tests.
</p>
</div>
<div id="sec-2" class="subsection methods">
<p id="p-2"><strong>Methods</strong> A systematic literature search was undertaken. Hierarchical summary receiver operating characteristic (HSROC) models were
used to estimate (relative) diagnostic odds ratios ([R]DORs). Summary sensitivity (correct identification of residual tumor),
specificity (correct identification of pCR), and areas under the SROC curves (AUCs) were derived. All statistical tests were
two-sided.
</p>
</div>
<div id="sec-3" class="subsection">
<p id="p-3"><strong>Results</strong> Forty-four studies (2050 patients) were included. The overall AUC of MRI was 0.88. Accuracy was lower for “standard” pCR
definitions (referent category) than “less clearly described” (RDOR = 2.41, 95% confidence interval [CI] = 1.11 to 5.23) or
“near-pCR” definitions (RDOR = 2.60, 95% CI = 0.73 to 9.24; <em>P</em> = .03.) Corresponding AUCs were 0.83, 0.90, and 0.91. Specificity was higher when negative MRI was defined as contrast enhancement
less than or equal to normal tissue (0.83, 95% CI = 0.64 to 0.93) vs no enhancement (0.54, 95% CI = 0.39 to 0.69; <em>P</em> = .02), with comparable sensitivity (0.83, 95% CI = 0.69 to 0.91; vs 0.87, 95% CI = 0.80 to 0.92; <em>P</em> = .45). MRI had higher accuracy than mammography (<em>P</em> = .02); there was only weak evidence that MRI had higher accuracy than clinical examination (<em>P</em> = .10). No difference in MRI and ultrasound accuracy was found (<em>P</em> = .15).
</p>
</div>
<div id="sec-4" class="subsection">
<p id="p-4"><strong>Conclusions</strong> MRI accurately detects residual tumor after neoadjuvant chemotherapy. Accuracy was lower when pCR was more rigorously defined,
and specificity was lower when test negativity thresholds were more stringent; these definitions require standardization.
MRI is more accurate than mammography; however, studies comparing MRI and ultrasound are required.
</p>
</div>
</div>
<p id="p-5">Neoadjuvant chemotherapy (NAC) has a well-established role in the management of breast cancer (<a id="xref-ref-1-1" class="xref-bibr" href="#ref-1">1–4</a>). For women with operable disease at presentation, the primary aim of NAC is the achievement of pathologic complete response
(pCR) prior to surgery (<a id="xref-ref-5-1" class="xref-bibr" href="#ref-5">5</a>,<a id="xref-ref-6-1" class="xref-bibr" href="#ref-6">6</a>), which has been shown to confer improvements in long-term disease-free and overall survival relative to cases in which residual
invasive tumor remains after NAC (<a id="xref-ref-7-1" class="xref-bibr" href="#ref-7">7</a>,<a id="xref-ref-8-1" class="xref-bibr" href="#ref-8">8</a>). Accurate ascertainment of whether pCR has been achieved or, conversely, accurate detection of the presence of residual
tumor is needed to inform surgical planning (<a id="xref-ref-9-1" class="xref-bibr" href="#ref-9">9</a>). Currently, assessment of the presence or absence of residual tumor after NAC informs the extent of subsequent surgery;
however, the avoidance of surgery remains a future goal for patients in whom an absence of residual tumor can be accurately
detected (<a id="xref-ref-10-1" class="xref-bibr" href="#ref-10">10</a>).
</p>
<p id="p-6" class="indent">Various breast imaging modalities have been used to detect whether residual malignancy is present or absent after NAC, of
which magnetic resonance imaging (MRI) has been increasingly used and recommended in recent years (<a id="xref-ref-9-2" class="xref-bibr" href="#ref-9">9</a>,<a id="xref-ref-11-1" class="xref-bibr" href="#ref-11">11</a>). In this systematic review, we examine the evidence on the ability of MRI to identify whether residual malignancy is present
or whether pCR has been achieved at completion of NAC, report estimates of MRI accuracy and comparative accuracy, and investigate
variables potentially affecting MRI accuracy in the NAC setting.
</p>
<div class="section methods" id="sec-5">
<div class="section-nav"><a href="#abstract-1" title="Abstract" class="prev-section-link"><span>Previous Section</span></a><a href="#sec-10" title="Results" class="next-section-link"><span>Next Section</span></a></div>
<h2>Methods</h2>
<div id="sec-6" class="subsection">
<h3>Identification of Studies</h3>
<p id="p-7">A systematic search of the biomedical literature up to February 2011 was undertaken to identify studies assessing the accuracy
of MRI after NAC in differentiating the presence of residual tumor from the absence of disease (ie, pCR). MEDLINE and EMBASE
were searched through EMBASE.com; PREMEDLINE, Database of Abstracts of Reviews of Effects, Heath Technology Assessment, and
Cochrane databases were searched through Ovid. Search terms were selected to link MRI with breast cancer and response to NAC.
Keywords and medical subject headings included “breast cancer,” “nuclear magnetic resonance imaging,” “MRI,” “neoadjuvant,”
and “response.” The full search strategy is available in <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix A</a> (available online). Reference lists were also searched, and content experts were consulted to identify additional studies.
</p>
</div>
<div id="sec-7" class="subsection">
<h3>Review of Studies and Eligibility Criteria</h3>
<p id="p-8">All abstracts were screened for eligibility by one author (M. L. Marinovich), and a sample of 10% was assessed independently
by a second author (N. Houssami) to ensure consistent application of the eligibility criteria. Eligible studies were required
to have enrolled patients with newly diagnosed breast cancer undergoing NAC, with MRI undertaken after NAC to detect the presence
of residual tumor before surgery. Studies must have the counts required to estimate sensitivity (the proportion of patients
with residual tumor correctly classified by MRI as having disease present) and specificity (the proportion of cases with pCR
in whom MRI declared an absence of residual tumor) or sufficient data to allow a 2×2 table to be extracted. Pathologic response
based on surgical excision was the reference standard, but studies were not excluded if alternative reference standards were
used in a minority of patients. Where studies presented comparisons with alternative assessment methods (ultrasound, clinical
examination, mammography), estimates of accuracy were also extracted or derived for these tests. Studies in which MRI was
undertaken only during NAC and studies that enrolled fewer than 10 patients were ineligible. One study (<a id="xref-ref-12-1" class="xref-bibr" href="#ref-12">12</a>) was identified a priori as having used a fixed MRI contrast dose, rather than dosage per unit of body weight, and was therefore
excluded.
</p>
<p id="p-9" class="indent">Potentially eligible citations were reviewed in full to determine eligibility (M. L. Marinovich or N. Houssami). The screening
and inclusion process is summarized in <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix B</a> (PRISMA flowchart) (available online).
</p>
</div>
<div id="sec-8" class="subsection">
<h3>Data Extraction</h3>
<p id="p-10">Data that was related to test accuracy, study design, patient characteristics, tumors, treatment, technical details of MRI,
comparator tests, and the reference standard were extracted independently by two authors (M. L. Marinovich, and either S.
Ciatto, M.E. Brennan, or F. Sardanelli). Study-level definitions of pathologic response and MRI thresholds for the absence
of residual tumor were categorized according to the criteria in <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix C</a> (available online). Quality appraisal was undertaken using the Quality Assessment of Diagnostic Accuracy Studies (QUADAS)
checklist (modified for application to studies of residual tumor detection in this setting) (<a id="xref-ref-13-1" class="xref-bibr" href="#ref-13">13</a>,<a id="xref-ref-14-1" class="xref-bibr" href="#ref-14">14</a>). Disagreements were resolved by discussion and consensus, with arbitration by a third author (N. Houssami) when required.
</p>
</div>
<div id="sec-9" class="subsection">
<h3>Statistical Analysis</h3>
<p id="p-11">Descriptive analyses were conducted using coupled forest plots and scatter plots of study-specific estimates of sensitivity
and specificity. Because studies varied in the criterion used to define a positive test result, the Rutter and Gatsonis hierarchical
summary receiver operating characteristic (HSROC) model (<a id="xref-ref-15-1" class="xref-bibr" href="#ref-15">15</a>) was used to model MRI accuracy in terms of the diagnostic odds ratio (DOR): [sensitivity/(1−sensitivity)]/[(1−specificity)/specificity].
The DOR is the ratio of the odds of MRI being positive when residual tumor is truly present relative to the odds of MRI being
positive when pCR has been achieved. A DOR of 1 means that the test does not discriminate between patients with and without
residual tumor; higher values indicate better test performance (<a id="xref-ref-16-1" class="xref-bibr" href="#ref-16">16</a>). The HSROC model takes into account uncertainty in estimates of sensitivity and specificity within studies, as well as additional
unexplained variation (heterogeneity) between studies, by the inclusion of random study effects for test accuracy and threshold
(a function of the underlying test positivity rate). A shape parameter, fitted in the model as a fixed effect, allows for
asymmetry in the SROC curve (ie, variation in accuracy by test threshold). For a symmetrical SROC curve, the estimated log<sub>e</sub>(DOR) is constant across thresholds. <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix D</a> (available online) shows a detailed specification of the model.
</p>
<p id="p-12" class="indent">HSROC models were fitted using PROC NLMIXED in SAS version 9.2 (SAS Institute, Cary, NC) (<a id="xref-ref-17-1" class="xref-bibr" href="#ref-17">17</a>). The distribution of the random effects for accuracy and threshold was checked for each model to ensure that normality assumptions
were met. For models in which the variation in accuracy between studies was observed to be negligible, test accuracy was modeled
as a fixed effect. Further detail of the model-fitting strategy is described by Macaskill (<a id="xref-ref-18-1" class="xref-bibr" href="#ref-18">18</a>).
</p>
<p id="p-13" class="indent">Covariables were added to the HSROC model to assess whether the shape or position (accuracy) of the SROC curve(s) was associated
with differences in patient, test, treatment, and study characteristics. Covariables were age (median of <50 years vs <em>></em>50 years); histology (proportion that was invasive ductal carcinoma); stage (proportion that was stage I/II); human epidermal
growth factor receptor 2 and estrogen receptor status (proportion that was receptor positive); chemotherapy type (anthracycline-based,
anthracycline/taxane-based, anthracyclines/taxanes alone or combined, other); surgery type (proportion that was mastectomy);
time from MRI to surgery (mean); midpoint of study enrollment period; definition of pCR (see <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix C</a>, available online); prevalence of pCR; comparative vs noncomparative study design; prospective vs retrospective design; and
consecutive vs nonconsecutive patient enrollment. Each covariable was modeled separately, and their contributions to the model
were assessed by the likelihood ratio test (<a id="xref-ref-19-1" class="xref-bibr" href="#ref-19">19</a>). Where subgroups of studies used equivalent MRI contrast enhancement thresholds (ie, no contrast enhancement vs enhancement
less than or equal to normal breast tissue) to define a negative result for residual tumor, summary estimates of sensitivity
and specificity were derived for these thresholds. Ninety-five percent confidence intervals (CIs) for the expected sensitivity
and specificity and the <em>t</em> statistics and corresponding <em>P</em> values for differences between MRI thresholds were derived using the ESTIMATE command in PROC NLMIXED (<a id="xref-ref-18-2" class="xref-bibr" href="#ref-18">18</a>).
</p>
<p id="p-14" class="indent">The HSROC model was also used to compare the test performance of MRI relative to ultrasound, clinical examination, and mammography
for subgroups of studies in which MRI and at least one comparator test were evaluated in the same patients (or in patient
groups that substantially overlapped). Test type was included as a covariable, with separate models used for each comparison.
</p>
<p id="p-15" class="indent">Where there was no evidence of asymmetry in the estimated SROC curves (assessed by the likelihood ratio test), the shape parameter
was set to zero, and the relative DOR (RDOR) was used to compare accuracy for levels of the covariable. Ninety-five percent
confidence intervals for RDORs were derived from the asymptotic standard error of the estimate reported by PROC NLMIXED and
assuming a <em>t</em> distribution, as described previously (<a id="xref-ref-18-3" class="xref-bibr" href="#ref-18">18</a>). The area under the curve (AUC) for each fitted SROC was computed by the method described by Walter (<a id="xref-ref-20-1" class="xref-bibr" href="#ref-20">20</a>) to provide a global measure of accuracy or using numerical integration when curves were asymmetric. The fitted curves were
displayed graphically, superimposed on a scatter-plot in ROC space of study-specific estimates of (sensitivity, 1-specificity)
pairs. Plotted curves were restricted to the range of data points.
</p>
<p id="p-16" class="indent">Differences in QUADAS items between studies were tested using χ<sup>2</sup> or Fisher exact tests, as appropriate. All tests of statistical significance were two-sided; the level chosen for statistical
significance was .05.
</p>
</div>
</div>
<div class="section" id="sec-10">
<div class="section-nav"><a href="#sec-5" title="Methods" class="prev-section-link"><span>Previous Section</span></a><a href="#sec-18" title="Discussion" class="next-section-link"><span>Next Section</span></a></div>
<h2>Results</h2>
<div id="sec-11" class="subsection">
<h3>Study Characteristics</h3>
<p id="p-17">A total of 2107 citations were identified. Forty-four studies (<a id="xref-ref-21-1" class="xref-bibr" href="#ref-21">21–64</a>)were eligible for inclusion in our meta-analysis, reporting data on 2949 patients (n = 2967 cancers) undergoing MRI and/or
comparator tests; MRI data were reported for 2050 patients (n = 2068 cancers). Studies enrolled patients between 1990 and
2008 (median midpoint of recruitment = year 2001) and included a median of 36 patients in the analysis of MRI accuracy (range
= 14–208). Characteristics of included studies are summarized in <a id="xref-table-wrap-1-1" class="xref-table" href="#T1">Table 1</a> and <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendices C</a> (pCR definitions and MRI thresholds) and E (MRI technical characteristics) (available online).
</p>
<div class="table pos-float" id="T1">
<div class="table-inline">
<div class="callout"><span>View this table:</span><ul class="callout-links">
<li><a href="321/T1.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/T1.expansion.html">In a new window</a></li>
</ul>
</div>
</div>
<div class="table-caption"><span class="table-label">Table 1. </span>
<p id="p-18" class="first-child">Summary of cohort, tumor, treatment and reference standard characteristics of included studies<sup>*</sup>
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<p id="p-22" class="indent">Patients enrolled in included studies had predominantly stage II and III cancer, and the majority had invasive ductal carcinoma
(see <a id="xref-table-wrap-1-2" class="xref-table" href="#T1">Table 1</a>). NAC was primarily anthracycline-taxane based, either sequential or in combination. Trastuzumab was used in 11 studies (range
of patients within studies = 1.5%–62.5%). Radiotherapy was given before surgery in two studies (<a id="xref-ref-54-1" class="xref-bibr" href="#ref-54">54</a>,<a id="xref-ref-64-1" class="xref-bibr" href="#ref-64">64</a>). For studies that specified the type of surgery undertaken, a majority of patients underwent breast conservation. Study
quality appraisal is summarized in <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix F</a> (available online).
</p>
</div>
<div id="sec-12" class="subsection">
<h3>MRI Details</h3>
<p id="p-23">The majority of studies used dynamic contrast-enhanced MRI (86.4%) with a 1.5-T magnet (77.3%). Dedicated bilateral breast
coils were used in all studies in which the coil type was reported. All studies that provided detail on contrast employed
gadolinium-based materials, most commonly gadopentetate dimeglumine (50.0%), typically at the standard dosage of 0.1 mmol/kg
body weight (61.4%) (see <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix E</a>, available online). Ten studies (22.7%) (<a id="xref-ref-27-1" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-29-1" class="xref-bibr" href="#ref-29">29–32</a>,<a id="xref-ref-37-1" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-38-1" class="xref-bibr" href="#ref-38">38</a>,<a id="xref-ref-42-1" class="xref-bibr" href="#ref-42">42</a>,<a id="xref-ref-46-1" class="xref-bibr" href="#ref-46">46</a>,<a id="xref-ref-50-1" class="xref-bibr" href="#ref-50">50</a>) considered MRI to be negative (absence of residual tumor) when there was an absence of contrast enhancement; a further six
studies (13.6%) (<a id="xref-ref-23-1" class="xref-bibr" href="#ref-23">23</a>,<a id="xref-ref-24-1" class="xref-bibr" href="#ref-24">24</a>,<a id="xref-ref-26-1" class="xref-bibr" href="#ref-26">26</a>,<a id="xref-ref-40-1" class="xref-bibr" href="#ref-40">40</a>,<a id="xref-ref-49-1" class="xref-bibr" href="#ref-49">49</a>,<a id="xref-ref-57-1" class="xref-bibr" href="#ref-57">57</a>) defined MRI negativity as contrast enhancement less than or equal to normal breast tissue. The remaining studies either
did not report MRI negativity in terms of the degree of contrast enhancement (n = 20, 45.4%) (<a id="xref-ref-21-2" class="xref-bibr" href="#ref-21">21</a>,<a id="xref-ref-22-1" class="xref-bibr" href="#ref-22">22</a>,<a id="xref-ref-25-1" class="xref-bibr" href="#ref-25">25</a>,<a id="xref-ref-33-1" class="xref-bibr" href="#ref-33">33</a>,<a id="xref-ref-34-1" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-36-1" class="xref-bibr" href="#ref-36">36</a>,<a id="xref-ref-39-1" class="xref-bibr" href="#ref-39">39</a>,<a id="xref-ref-41-1" class="xref-bibr" href="#ref-41">41</a>,<a id="xref-ref-43-1" class="xref-bibr" href="#ref-43">43–45</a>,<a id="xref-ref-48-1" class="xref-bibr" href="#ref-48">48</a>,<a id="xref-ref-52-1" class="xref-bibr" href="#ref-52">52</a>,<a id="xref-ref-54-2" class="xref-bibr" href="#ref-54">54</a>,<a id="xref-ref-56-1" class="xref-bibr" href="#ref-56">56</a>,<a id="xref-ref-58-1" class="xref-bibr" href="#ref-58">58</a>,<a id="xref-ref-59-1" class="xref-bibr" href="#ref-59">59</a>,<a id="xref-ref-61-1" class="xref-bibr" href="#ref-61">61–63</a>) or did not specify a threshold (n = 8, 18.2%) (<a id="xref-ref-28-1" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-35-1" class="xref-bibr" href="#ref-35">35</a>,<a id="xref-ref-47-1" class="xref-bibr" href="#ref-47">47</a>,<a id="xref-ref-51-1" class="xref-bibr" href="#ref-51">51</a>,<a id="xref-ref-53-1" class="xref-bibr" href="#ref-53">53</a>,<a id="xref-ref-55-1" class="xref-bibr" href="#ref-55">55</a>,<a id="xref-ref-64-2" class="xref-bibr" href="#ref-64">64</a>) (see <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix C</a>, available online).
</p>
</div>
<div id="sec-13" class="subsection">
<h3>Reference Standard</h3>
<p id="p-24">Pathology from surgical excision was the reference standard for all patients in all but two studies; test results were verified
by localization biopsy in a small number (6.2%) of patients in one study (<a id="xref-ref-36-2" class="xref-bibr" href="#ref-36">36</a>) and by follow-up (41.2%) in another (<a id="xref-ref-64-3" class="xref-bibr" href="#ref-64">64</a>).
</p>
<p id="p-25" class="indent">Definitions of reference standard positivity (presence of residual tumor) and negativity (pCR) varied across studies (see
<a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix C</a>, available online). Twenty studies (<a id="xref-ref-21-3" class="xref-bibr" href="#ref-21">21</a>,<a id="xref-ref-23-2" class="xref-bibr" href="#ref-23">23–25</a>,<a id="xref-ref-27-2" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-30-1" class="xref-bibr" href="#ref-30">30–33</a>,<a id="xref-ref-36-3" class="xref-bibr" href="#ref-36">36</a>,<a id="xref-ref-37-2" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-41-2" class="xref-bibr" href="#ref-41">41</a>,<a id="xref-ref-43-2" class="xref-bibr" href="#ref-43">43</a>,<a id="xref-ref-44-1" class="xref-bibr" href="#ref-44">44</a>,<a id="xref-ref-48-2" class="xref-bibr" href="#ref-48">48</a>,<a id="xref-ref-52-2" class="xref-bibr" href="#ref-52">52</a>,<a id="xref-ref-56-2" class="xref-bibr" href="#ref-56">56</a>,<a id="xref-ref-57-2" class="xref-bibr" href="#ref-57">57</a>,<a id="xref-ref-59-2" class="xref-bibr" href="#ref-59">59</a>,<a id="xref-ref-62-1" class="xref-bibr" href="#ref-62">62</a>) (45.5%) de- fined pCR as the absence of invasive cancer on pathological examination, with or without the presence of ductal
carcinoma in situ (DCIS; ie, residual DCIS was considered negative). Nine of these studies (<a id="xref-ref-27-3" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-30-2" class="xref-bibr" href="#ref-30">30</a>,<a id="xref-ref-32-1" class="xref-bibr" href="#ref-32">32</a>,<a id="xref-ref-36-4" class="xref-bibr" href="#ref-36">36</a>,<a id="xref-ref-37-3" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-41-3" class="xref-bibr" href="#ref-41">41</a>,<a id="xref-ref-44-2" class="xref-bibr" href="#ref-44">44</a>,<a id="xref-ref-56-3" class="xref-bibr" href="#ref-56">56</a>,<a id="xref-ref-57-3" class="xref-bibr" href="#ref-57">57</a>) provided data that allowed DCIS to be classified as either positive or negative for residual disease; primary analyses classified
DCIS as negative on the reference standard, consistent with the Miller–Payne grading system (<a id="xref-ref-65-1" class="xref-bibr" href="#ref-65">65</a>), and the effect of classifying residual DCIS as positive in these studies was explored in sensitivity analyses. In four
additional studies (9.1%) (<a id="xref-ref-34-2" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-45-1" class="xref-bibr" href="#ref-45">45</a>,<a id="xref-ref-50-2" class="xref-bibr" href="#ref-50">50</a>,<a id="xref-ref-53-2" class="xref-bibr" href="#ref-53">53</a>), pCR was defined as the absence of any residual invasive cancer or DCIS (ie, residual DCIS was considered positive). In
12 studies (27.3%) (<a id="xref-ref-26-2" class="xref-bibr" href="#ref-26">26</a>,<a id="xref-ref-38-2" class="xref-bibr" href="#ref-38">38–40</a>,<a id="xref-ref-46-2" class="xref-bibr" href="#ref-46">46</a>,<a id="xref-ref-47-2" class="xref-bibr" href="#ref-47">47</a>,<a id="xref-ref-49-2" class="xref-bibr" href="#ref-49">49</a>,<a id="xref-ref-51-2" class="xref-bibr" href="#ref-51">51</a>,<a id="xref-ref-58-2" class="xref-bibr" href="#ref-58">58</a>,<a id="xref-ref-60-1" class="xref-bibr" href="#ref-60">60</a>,<a id="xref-ref-61-2" class="xref-bibr" href="#ref-61">61</a>,<a id="xref-ref-63-1" class="xref-bibr" href="#ref-63">63</a>), nonspecific definitions that did not describe whether residual DCIS was considered positive or negative (eg, pCR was defined
simply as the absence of residual disease/malignancy or the measurement of residua being zero) were employed. Four studies
(9.1%) (<a id="xref-ref-28-2" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-42-2" class="xref-bibr" href="#ref-42">42</a>,<a id="xref-ref-54-3" class="xref-bibr" href="#ref-54">54</a>,<a id="xref-ref-55-2" class="xref-bibr" href="#ref-55">55</a>) allowed reference standard negativity to include small clusters of microscopic invasive cells or similar definitions of
minimal residual disease (“near-pCR”). A further four studies (9.1%) (<a id="xref-ref-22-2" class="xref-bibr" href="#ref-22">22</a>,<a id="xref-ref-29-2" class="xref-bibr" href="#ref-29">29</a>,<a id="xref-ref-35-2" class="xref-bibr" href="#ref-35">35</a>,<a id="xref-ref-64-4" class="xref-bibr" href="#ref-64">64</a>) did not define reference standard positivity and negativity.
</p>
</div>
<div id="sec-14" class="subsection">
<h3>pCR Rates</h3>
<p id="p-26">Study-specific pCR rates ranged between 2.6% and 54.9%, with a median of 16.0%. The rates are presented in <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix G</a> (available online), stratified by response definition.
</p>
</div>
<div id="sec-15" class="subsection">
<h3>Accuracy of MRI</h3>
<p id="p-27">Study-specific estimates of MRI sensitivity and specificity are presented in <a id="xref-fig-1-1" class="xref-fig" href="#F1">Figure 1</a>. Median sensitivity across studies was 0.92 (interquartile range [IQR] = 0.85–0.97), and median specificity was 0.60 (IQR
= 0.39–0.96). <a id="xref-table-wrap-2-1" class="xref-table" href="#T2">Table 2</a> reports the overall and covariable-specific modeled estimates of MRI accuracy (derived from separate models for each covariable);
in all but one of these models (midpoint of patient enrollment = year 2000 or earlier vs year 2001 or later), the shape parameter
was not statistically significant (ie, SROC curves were symmetrical). Overall, the AUC for MRI based on all 44 studies was
0.88; the SROC curve for all studies is shown in <a id="xref-fig-2-1" class="xref-fig" href="#F2">Figure 2</a>.
</p>
<div class="table pos-float" id="T2">
<div class="table-inline">
<div class="callout"><span>View this table:</span><ul class="callout-links">
<li><a href="321/T2.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/T2.expansion.html">In a new window</a></li>
</ul>
</div>
</div>
<div class="table-caption"><span class="table-label">Table 2. </span>
<p id="p-28" class="first-child">Univariate models of magnetic resonance imaging (MRI) accuracy and comparisons of the accuracy of MRI and clinical examination,
ultrasound, and mammography<sup>*</sup>
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<div id="F1" class="fig pos-float type-figure odd">
<div class="fig-inline"><a href="321/F1.expansion.html"><img alt="Figure 1. " src="321/F1.small.gif" /></a><div class="callout"><span>View larger version:</span><ul class="callout-links">
<li><a href="321/F1.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/F1.expansion.html">In a new window</a></li>
</ul>
<ul class="fig-services">
<li class="ppt-link"><a href="/powerpoint/105/5/321/F1">Download as PowerPoint Slide</a></li>
</ul>
</div>
</div>
<div class="fig-caption"><span class="fig-label">Figure 1. </span>
<p id="p-36" class="first-child">Forest plot of study-specific estimates of magnetic resonance imaging (MRI) sensitivity and specificity. The <strong>black squares</strong> and <strong>horizontal lines</strong> represent the estimate and 95% confidence interval (CI) for each study. The overlap between cohorts reported in the two studies
by Chen et al. (<a id="xref-ref-23-3" class="xref-bibr" href="#ref-23">23</a>,<a id="xref-ref-24-2" class="xref-bibr" href="#ref-24">24</a>) is 14 patients. The two Belli et al. (<a id="xref-ref-39-2" class="xref-bibr" href="#ref-39">39</a>,<a id="xref-ref-40-2" class="xref-bibr" href="#ref-40">40</a>) studies report different study cohorts; there were no overlapping patients. FN = false negative; FP = false positive; TN
= true negative; TP = true positive.
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<div id="F2" class="fig pos-float type-figure odd">
<div class="fig-inline"><a href="321/F2.expansion.html"><img alt="Figure 2. " src="321/F2.small.gif" /></a><div class="callout"><span>View larger version:</span><ul class="callout-links">
<li><a href="321/F2.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/F2.expansion.html">In a new window</a></li>
</ul>
<ul class="fig-services">
<li class="ppt-link"><a href="/powerpoint/105/5/321/F2">Download as PowerPoint Slide</a></li>
</ul>
</div>
</div>
<div class="fig-caption"><span class="fig-label">Figure 2. </span>
<p id="p-37" class="first-child">Summary receiver operating characteristics curve for magnetic resonance imaging from all included studies. The <strong>black squares</strong> represent estimates of sensitivity and specificity for each study. The <strong>relative size</strong> of the black square represents study sample size.
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<p id="p-38" class="indent">MRI accuracy (DOR) differed according to the applied study-level definition of pCR (<em>P</em> = .03). <a id="xref-fig-3-1" class="xref-fig" href="#F3">Figure 3</a> displays SROC curves stratified by pCR definition. Accuracy was lowest in studies that permitted residual DCIS in the definition
of pCR; relative to this referent group, accuracy was higher in studies that excluded residual DCIS from the definition of
pCR (RDOR = 1.31, 95% CI = 0.33 to 5.20), applied a nonspecific definition of reference standard positivity/negativity (RDOR
= 2.41, 95% CI = 1.11 to 5.23), or used a near-pCR definition (RDOR = 2.60, 95% CI = 0.73 to 9.24). Relatively few studies
excluded DCIS from the pCR definition (n = 4) or used a near-pCR outcome (n = 4); hence confidence intervals around RDORs
for these definitions are relatively wide. AUCs for pCR definitions were 0.83 for the absence of invasive disease, with or
without DCIS; 0.86 for the absence of invasive disease and DCIS; 0.90 for nonspecific definitions; and 0.91 for near-pCR.
</p>
<div id="F3" class="fig pos-float type-figure odd">
<div class="fig-inline"><a href="321/F3.expansion.html"><img alt="Figure 3. " src="321/F3.small.gif" /></a><div class="callout"><span>View larger version:</span><ul class="callout-links">
<li><a href="321/F3.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/F3.expansion.html">In a new window</a></li>
</ul>
<ul class="fig-services">
<li class="ppt-link"><a href="/powerpoint/105/5/321/F3">Download as PowerPoint Slide</a></li>
</ul>
</div>
</div>
<div class="fig-caption"><span class="fig-label">Figure 3. </span>
<p id="p-39" class="first-child">Summary receiver operating characteristics curves for magnetic resonance imaging stratified by definition of pathologic complete
response (pCR). The <strong>black circles</strong> and <strong>black curve</strong> represent studies that defined pathologic response as the absence of invasive cancer, with or without the presence of ductal
carcinoma in situ (DCIS). The <strong>red diamonds</strong> and <strong>red curve</strong> represent studies that defined pathologic response as the absence of invasive cancer and DCIS. The <strong>green squares</strong> and <strong>green curve</strong> represent studies that defined pathologic response as no residual disease (not further specified). The <strong>blue triangles</strong> and <strong>blue curve</strong> represent studies that defined pathologic response as near-pCR (minimal residual disease). The <strong>relative size</strong> of the symbols represents study sample size.
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<p id="p-40" class="indent">The midpoint of patient enrollment to each study (year 2000 or earlier vs year 2001 or later) was also associated with MRI
accuracy. However, different SROC curve shapes were observed for earlier and later studies (<em>P</em> = .01) (see <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix H</a>, available online); it is therefore not possible to report single DORs (or a RDOR) as summary measures of accuracy. Earlier
studies reported consistently high sensitivity across the range of specificity values, whereas a trade-off between sensitivity
and specificity at different thresholds was evident in later studies. Studies with a midpoint of patient enrollment of year
2000 or earlier reported higher overall accuracy than those with a midpoint of year 2001 or later (AUC = 0.92 vs 0.83). Comparison
of QUADAS items between earlier vs later studies suggested no major differences in study quality between levels of this covariable
(<em>P</em> > .05 for all QUADAS items). No statistical evidence was found for associations between MRI accuracy and other variables
related to study design, patient characteristics, and treatment characteristics (<a id="xref-table-wrap-2-2" class="xref-table" href="#T2">Table 2</a>).
</p>
<p id="p-41" class="indent">Sensitivity analyses, in which modelling was repeated with changed pCR definitions in nine studies (see “Reference Standard”),
resulted in similar parameter estimates to those in the primary analysis. Sensitivity analyses were also undertaken to exclude
one study (<a id="xref-ref-23-4" class="xref-bibr" href="#ref-23">23</a>) with a patient cohort that overlapped with a second study (<a id="xref-ref-24-3" class="xref-bibr" href="#ref-24">24</a>); additional analyses excluded two studies that did not use a reference standard of pathologic examination in all patients
(<a id="xref-ref-36-5" class="xref-bibr" href="#ref-36">36</a>,<a id="xref-ref-64-5" class="xref-bibr" href="#ref-64">64</a>). Exclusion of these studies did not substantially affect parameter estimates.
</p>
</div>
<div id="sec-16" class="subsection">
<h3>Threshold-Specific Sensitivity and Specificity of MRI</h3>
<p id="p-42">Ten studies used a complete absence of MRI enhancement to identify pCR (threshold 1), and six studies used contrast enhancement
equal to or less than normal breast tissue to define a negative MRI result (threshold 2). The summary estimate of specificity
was higher for threshold 2 (0.83, 95% CI = 0.64 to 0.93) than for threshold 1 (0.54, 95% CI = 0.39 to 0.69; <em>P</em> = .02), with comparable pooled sensitivity (0.83, 95% CI = 0.69 to 0.91; vs 0.87, 95% CI = 0.80 to 0.92; <em>P</em> = .45). Summary estimates are displayed in <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix I</a> (available online).
</p>
</div>
<div id="sec-17" class="subsection">
<h3>Comparisons of the Accuracy of MRI and Other Tests</h3>
<p id="p-43">
<a id="xref-table-wrap-2-3" class="xref-table" href="#T2">Table 2</a> presents comparisons between MRI (referent group) and comparator tests, based on subgroups of studies undertaking MRI and
clinical examination [11 studies (<a id="xref-ref-31-1" class="xref-bibr" href="#ref-31">31</a>,<a id="xref-ref-32-2" class="xref-bibr" href="#ref-32">32</a>,<a id="xref-ref-34-3" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-43-3" class="xref-bibr" href="#ref-43">43</a>,<a id="xref-ref-44-3" class="xref-bibr" href="#ref-44">44</a>,<a id="xref-ref-50-3" class="xref-bibr" href="#ref-50">50</a>,<a id="xref-ref-52-3" class="xref-bibr" href="#ref-52">52</a>,<a id="xref-ref-54-4" class="xref-bibr" href="#ref-54">54</a>,<a id="xref-ref-56-4" class="xref-bibr" href="#ref-56">56</a>, <a id="xref-ref-58-3" class="xref-bibr" href="#ref-58">58</a>,<a id="xref-ref-64-6" class="xref-bibr" href="#ref-64">64</a>)], ultrasound [10 studies (<a id="xref-ref-31-2" class="xref-bibr" href="#ref-31">31</a>,<a id="xref-ref-34-4" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-36-6" class="xref-bibr" href="#ref-36">36</a>,<a id="xref-ref-42-3" class="xref-bibr" href="#ref-42">42</a>,<a id="xref-ref-45-2" class="xref-bibr" href="#ref-45">45</a>,<a id="xref-ref-50-4" class="xref-bibr" href="#ref-50">50</a>,<a id="xref-ref-54-5" class="xref-bibr" href="#ref-54">54</a>,<a id="xref-ref-55-3" class="xref-bibr" href="#ref-55">55</a>,<a id="xref-ref-62-2" class="xref-bibr" href="#ref-62">62</a>,<a id="xref-ref-63-2" class="xref-bibr" href="#ref-63">63</a>)], or mammography [7 studies (<a id="xref-ref-34-5" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-36-7" class="xref-bibr" href="#ref-36">36</a>,<a id="xref-ref-43-4" class="xref-bibr" href="#ref-43">43</a>,<a id="xref-ref-50-5" class="xref-bibr" href="#ref-50">50</a>,<a id="xref-ref-54-6" class="xref-bibr" href="#ref-54">54</a>,<a id="xref-ref-55-4" class="xref-bibr" href="#ref-55">55</a>,<a id="xref-ref-64-7" class="xref-bibr" href="#ref-64">64</a>)]. There was evidence that mammography had lower accuracy than MRI (RDOR = 0.27, 95% CI = 0.07 to 1.02; <em>P</em> = .02) (AUC = 0.89 vs 0.95; SROC presented in <a id="xref-fig-4-1" class="xref-fig" href="#F4">Figure 4</a>). The analysis was repeated after removing one potentially influential study (<a id="xref-ref-36-8" class="xref-bibr" href="#ref-36">36</a>), and the statistically significant difference in accuracy between MRI and mammography remained (RDOR = 0.36, 95% CI = 0.08
to 1.60; <em>P</em> = .04), with comparable AUCs (0.88 vs 0.94).
</p>
<div id="F4" class="fig pos-float type-figure odd">
<div class="fig-inline"><a href="321/F4.expansion.html"><img alt="Figure 4. " src="321/F4.small.gif" /></a><div class="callout"><span>View larger version:</span><ul class="callout-links">
<li><a href="321/F4.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/F4.expansion.html">In a new window</a></li>
</ul>
<ul class="fig-services">
<li class="ppt-link"><a href="/powerpoint/105/5/321/F4">Download as PowerPoint Slide</a></li>
</ul>
</div>
</div>
<div class="fig-caption"><span class="fig-label">Figure 4. </span>
<p id="p-44" class="first-child">Summary receiver operating characteristics curves for magnetic resonance imaging (MRI) vs mammography. The <strong>black squares</strong> and <strong>solid curve</strong> represent MRI. The <strong>black diamonds</strong> and <strong>dashed curve</strong> represent mammography. The <strong>relative size</strong> of the symbols represents study sample size.
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<p id="p-45" class="indent">There was only weak evidence that clinical examination had lower accuracy than MRI (RDOR = 0.53, 95% CI = 0.22 to 1.28; <em>P</em> = .10; AUC = 0.83 vs 0.89; SROC presented in <a id="xref-fig-5-1" class="xref-fig" href="#F5">Figure 5</a>). Accuracy favored MRI in four studies; in the remaining seven studies, MRI was observed to have higher sensitivity but lower
specificity than clinical examination. The lower accuracy observed for ultrasound compared with MRI was not statistically
significant (RDOR = 0.54, 95% CI = 0.20 to 1.44; <em>P</em> = .15; AUC 0.90 vs 0.93; SROC presented in <a id="xref-fig-6-1" class="xref-fig" href="#F6">Figure 6</a>). Differences in sensitivity were generally small in all 10 studies comparing the tests; in three of four studies with relatively
larger differences in specificity, this difference favored MRI.
</p>
<div id="F5" class="fig pos-float type-figure odd">
<div class="fig-inline"><a href="321/F5.expansion.html"><img alt="Figure 5. " src="321/F5.small.gif" /></a><div class="callout"><span>View larger version:</span><ul class="callout-links">
<li><a href="321/F5.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/F5.expansion.html">In a new window</a></li>
</ul>
<ul class="fig-services">
<li class="ppt-link"><a href="/powerpoint/105/5/321/F5">Download as PowerPoint Slide</a></li>
</ul>
</div>
</div>
<div class="fig-caption"><span class="fig-label">Figure 5. </span>
<p id="p-46" class="first-child">Summary receiver operating characteristics curves for magnetic resonance imaging (MRI) vs clinical examination. The <strong>black squares</strong> and <strong>solid curve</strong> represent MRI. The <strong>black diamonds</strong> and <strong>dashed curve</strong> represent clinical examination. The <strong>relative size</strong> of the symbols represents study sample size.
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<div id="F6" class="fig pos-float type-figure odd">
<div class="fig-inline"><a href="321/F6.expansion.html"><img alt="Figure 6. " src="321/F6.small.gif" /></a><div class="callout"><span>View larger version:</span><ul class="callout-links">
<li><a href="321/F6.expansion.html">In this window</a></li>
<li><a class="in-nw" href="321/F6.expansion.html">In a new window</a></li>
</ul>
<ul class="fig-services">
<li class="ppt-link"><a href="/powerpoint/105/5/321/F6">Download as PowerPoint Slide</a></li>
</ul>
</div>
</div>
<div class="fig-caption"><span class="fig-label">Figure 6. </span>
<p id="p-47" class="first-child">Summary receiver operating characteristics curves for magnetic resonance imaging (MRI) vs ultrasound. The <strong>black squares</strong> and <strong>solid curve</strong> represent MRI. The <strong>black diamonds</strong> and <strong>dashed curve</strong> represent ultrasound. The <strong>relative size</strong> of the symbols represents study sample size.
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
</div>
</div>
<div class="section" id="sec-18">
<div class="section-nav"><a href="#sec-10" title="Results" class="prev-section-link"><span>Previous Section</span></a><a href="#sec-19" title="Funding" class="next-section-link"><span>Next Section</span></a></div>
<h2>Discussion</h2>
<p id="p-48">In the neoadjuvant setting, accurate information on whether residual malignancy is present or whether pCR has been achieved
assists in guiding surgical management of breast cancer. We modeled the accuracy of breast MRI, when performed preoperatively
after NAC, through evidence synthesis from 44 studies (MRI data for 2068 cancers). Studies generally showed high sensitivity
(correct detection of residual tumor), with evidence of heterogeneity in the estimates of specificity (correct identification
of pCR) (<a id="xref-fig-1-2" class="xref-fig" href="#F1">Figure 1</a>). Our meta-analysis showed that the capability of MRI for differentiating the presence of residual malignancy from pCR had
an overall AUC of 0.88 and that overall accuracy differed according to definition of pCR and study timeframe.
</p>
<p id="p-49" class="indent">Our meta-analysis adds substantially to earlier work (<a id="xref-ref-66-1" class="xref-bibr" href="#ref-66">66</a>), not only by including a greater number of studies but also by addressing comparative accuracy of MRI and other tests. In
addition, we extensively explored study-level covariables, which allowed us to identify new associations and to provide methodologically
appropriate estimates of sensitivity and specificity according to MRI positivity thresholds. In this meta-analysis, the median
pCR rate was 16.0%, and, although this varied across 44 studies (range = 2.6%–54.9%) and an earlier review based on fewer
studies suggested MRI accuracy was associated with rates of pCR (<a id="xref-ref-66-2" class="xref-bibr" href="#ref-66">66</a>), there was no statistically significant association between pCR rate and MRI accuracy in our models. However, the accuracy
of MRI differed according to pCR definition (<em>P</em> = .03; see <a id="xref-fig-3-2" class="xref-fig" href="#F3">Figure 3</a>). Relative to a referent group of studies using a clearly described “standard” definition (no invasive tumor, with or without
the presence of residual DCIS) the accuracy of MRI was higher in studies using pCR definitions that were not clearly described
(RDOR = 2.41, 95% CI = 1.11 to 5.23). Underlying methodological problems within studies may be associated with a poorly defined
outcome definition, contributing to an overestimation of the accuracy of MRI relative to studies that employed clearly described
standardized definitions of pCR.
</p>
<p id="p-50" class="indent">Compared with a standard pCR definition, RDORs for studies that defined pCR as an absence of both invasive tumor and DCIS
or as near-pCR were 1.31 (95% CI = 0.33 to 5.20) and 2.60 (95% CI = 0.73 to 9.24), respectively. Wide confidence intervals
around these estimates reflect relatively few studies using the latter definitions (n = 4 for each definition); however, an
increase in accuracy when residual DCIS is excluded vs included in the pCR definition is consistent with previous studies
that reported lower MRI sensitivity in detecting DCIS relative to invasive cancer (<a id="xref-ref-67-1" class="xref-bibr" href="#ref-67">67</a>). Similarly, MRI has been observed to have limitations in detecting scattered, microscopic tumor foci after NAC (<a id="xref-ref-11-2" class="xref-bibr" href="#ref-11">11</a>,<a id="xref-ref-68-1" class="xref-bibr" href="#ref-68">68</a>); the estimated RDOR for near-pCR relative to a standard pCR definition may reflect fewer false negatives and a consequent
increase in true negative MRI results when a near-pCR definition is used. Given that near-pCR may plausibly overestimate accuracy
relative to standard pCR definitions and given the impact of residual malignancy on prognosis (<a id="xref-ref-69-1" class="xref-bibr" href="#ref-69">69</a>), the use of near-pCR as an outcome in the preoperative, post-NAC setting is not recommended. This analysis highlights the
importance of standardizing pCR definitions in the NAC setting (<a id="xref-ref-70-1" class="xref-bibr" href="#ref-70">70</a>).
</p>
<p id="p-51" class="indent">One of the strengths of our work is the characterization (<a id="xref-table-wrap-1-3" class="xref-table" href="#T1">Table 1</a>) and evaluation in analysis (<a id="xref-table-wrap-2-4" class="xref-table" href="#T2">Table 2</a>) of a large number of covariables related to study quality, patient characteristics, tumor characteristics, MRI, and treatment.
We found that studies with a midpoint of patient enrollment of year 2000 or earlier reported higher overall AUC (AUC = 0.92)
than those with a midpoint of year 2001 or later (AUC = 0.83), although SROC curve shapes differed for earlier and later studies
(<em>P</em> = .01). Examination of the curves indicated that earlier studies had consistently high sensitivity across the range of specificity
values, whereas a trade-off between sensitivity and specificity at different thresholds was evident in later studies. With
the evolution of MRI technology over time, it may appear counterintuitive that relatively lower accuracy was observed in more
recent studies; however, a meta-analysis of preoperative MRI (<a id="xref-ref-71-1" class="xref-bibr" href="#ref-71">71</a>) also suggested similar findings. No clear differences in study quality between timeframes were observed to account for the
above finding; however, it is possible that earlier studies may have involved radiologists with MRI expertise and that later
studies involved readers with less MRI-dedicated expertise, reflecting broader adoption of MRI in breast imaging practice.
It may also be possible that readers in more recent studies adjusted the implicit threshold used to define MRI positivity/negativity
in response to the relatively lower specificity reported in many earlier studies.
</p>
<p id="p-52" class="indent">In our subgroup analysis of studies that reported contrast enhancement thresholds applied to declare a positive or negative
test, there was no statistically significant difference between thresholds for summary estimates of MRI sensitivity (ie, correct
detection of residual malignancy) (0.83 vs 0.87; <em>P</em> = .45). However, MRI specificity (ie, correct identification of pCR) was statistically significantly greater when contrast
enhancement equal to or less than normal breast tissue was considered negative for residual tumor compared with a complete
absence of contrast uptake (0.83 vs 0.54; <em>P</em> = .02), which reflects the likelihood that enhancement caused by inflammatory or reactive changes post-NAC may be considered
false positive for residual malignancy using the latter threshold. These findings raise concerns about MRI potentially underestimating
the effect of NAC in achieving pCR where a stringent threshold is applied for defining the absence of residual malignancy;
however, when pCR is identified by contrast enhancement equal to or less than normal breast tissue, the relatively higher
specificity of MRI may allow better planning of breast conserving surgery. Standardization of MRI interpretation criteria/thresholds
in this clinical setting is required.
</p>
<p id="p-53" class="indent">Our analysis showed that the accuracy of MRI was statistically significantly higher than that of mammography (<em>P</em> = .02). There was only weak evidence suggesting that MRI also had greater accuracy than clinical examination (<em>P</em> = .10). Differences in summary accuracy estimates for MRI and ultrasound were not statistically significant (<em>P</em> = .15). These subgroup analyses were based on fewer studies because they were limited to studies that directly compared tests
and, therefore, have relatively reduced power to detect differences in test accuracy. This may account for the lack of statistical
differences between MRI and ultrasound accuracy in subgroup analysis; however, the findings may also represent true similarity
in accuracy for MRI and ultrasound. We were unable to compare the accuracy of MRI with a combination of ultrasound and clinical
examination because of a lack of studies that presented these data. The high relative cost of MRI, combined with potential
advantages of clinical examination and ultrasound in terms of accessibility, suggest that a combination of the latter may
be a reasonable alternative testing strategy to MRI in preoperative assessment after NAC. We recommend that future research
aim to compare the combined accuracy of ultrasound and clinical examination with that of MRI in the NAC setting.
</p>
<p id="p-54" class="indent">This study has some limitations. The reporting of information related to methodological quality was highly variable between
studies and individual QUADAS items (see <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix F</a>, available online), and some studies did not adequately describe MRI technical details (see <a href="http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djs528/-/DC1">Supplementary Appendix E</a>, available online). Investigators should be encouraged to fully describe study methodology, MRI technology, and technique
to allow the risk of bias and the generalizability of study findings to be assessed. Furthermore, relatively recent improvements
in MRI technology that may be expected to potentially improve accuracy (eg, multichannel coils; ≥3-T magnets; contrast materials
with high relaxivity; additional sequences allowing for diffusion weighted imaging) were underrepresented in studies included
in this analysis. The effect of these developments on the accuracy of MRI should be the subject of further study.
</p>
<p id="p-55" class="indent">In summary, our meta-analysis has shown good overall accuracy for MRI, although accuracy estimates varied with the definition
of pCR, which highlights the importance of standardizing pCR definitions. Subgroup analysis also suggests that MRI may be
more likely to be false positive for residual malignancy (thereby falsely underestimating the effect of NAC in achieving pCR)
in studies that defined absence of residual malignancy on MRI as contrast enhancement less than or equal to that of normal
breast tissue, rather than an absence of enhancement. In comparative studies, MRI was more accurate than mammography, but
no differences in accuracy were observed between MRI and other less technically complex and costly tests (ultrasound, clinical
examination). However, relatively few studies reported direct comparisons between MRI and other tests, and the comparative
accuracy of MRI and combined ultrasound and clinical examination warrants further investigation in well-designed clinical
trials.
</p>
</div>
<div class="section" id="sec-19">
<div class="section-nav"><a href="#sec-18" title="Discussion" class="prev-section-link"><span>Previous Section</span></a><a href="#notes-1" title="Notes" class="next-section-link"><span>Next Section</span></a></div>
<h2>Funding</h2>
<p id="p-56">This work was partly funded by the National Health and Medical Research Council (NHMRC) (program grant 633003 to the Screening
& Test Evaluation Program). MLM was supported by an NHMRC postgraduate scholarship.
</p>
</div>
<div class="section notes" id="notes-1">
<div class="section-nav"><a href="#sec-19" title="Funding" class="prev-section-link"><span>Previous Section</span></a><a href="#ref-list-1" title="References" class="next-section-link"><span>Next Section</span></a></div>
<h2>Notes</h2>
<p id="p-57">M. L. Marinovich conceived and co-ordinated the study, conducted the literature searches and review of studies, performed
the data extraction and statistical analysis, and drafted the manuscript. N. Houssami conceived the study, advised on literature
searches and study eligibility, contributed to resolution of data extraction, advised on clinical aspects and data interpretation,
and contributed to drafting the manuscript. P. Macaskill conceived the statistical methods used, advised on data analysis
and interpretation, and contributed to drafting the manuscript. F. Sardanelli contributed to data extraction, advised on MRI
technical issues and clinical aspects, and contributed to drafting the manuscript. L. Irwig advised on methodological aspects
and data interpretation and contributed to drafting the manuscript. E. P. Mamounas advised on clinical aspects and contributed
to drafting the manuscript. G. von Minckwitz advised on clinical aspects and contributed to drafting the manuscript. M. E.
Brennan contributed to data extraction and to drafting the manuscript. S. Ciatto contributed to data extraction, advised on
clinical aspects and contributed to drafting the manuscript. All authors read and approved the final manuscript.
</p>
<p id="p-58" class="indent">The funder had no role in the design of the study; the collection, analysis and interpretation of the data; the writing of
the manuscript; and the decision to submit the manuscript for publication.
</p>
</div>
<ul class="copyright-statement">
<li class="fn" id="copyright-statement-1">© The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com.</li>
</ul>
<div class="section ref-list" id="ref-list-1">
<div class="section-nav"><a href="#notes-1" title="Notes" class="prev-section-link"><span>Previous Section</span></a><div class="nav-placeholder"> </div>
</div>
<h2>References</h2>
<ol class="cit-list ref-use-labels">
<li><span class="ref-label">1.</span><a class="rev-xref-ref" href="#xref-ref-1-1" title="View reference 1. in text"
id="ref-1">↵</a>
<div class="cit ref-cit ref-journal" id="cit-105.5.321.1"
data-doi="10.1002/jso.21696">
<div class="cit-metadata">
<ol class="cit-auth-list">
<li><span class="cit-auth"><span class="cit-name-surname">Makhoul</span> <span class="cit-name-given-names">I</span></span>,
</li>
<li><span class="cit-auth"><span class="cit-name-surname">Kiwan</span> <span class="cit-name-given-names">E</span></span></li>
</ol><cite>.
<span class="cit-article-title">Neoadjuvant systemic treatment of breast cancer</span>.
<abbr class="cit-jnl-abbrev">J Surg Oncol</abbr>.
<span class="cit-pub-date">2011</span>;
<span class="cit-vol">103</span>(<span class="cit-issue">4</span>):<span class="cit-fpage">348</span>–<span class="cit-lpage">357</span>.
</cite></div>
<div class="cit-extra"><a href="/external-ref?access_num=10.1002/jso.21696&link_type=DOI"
class="cit-ref-sprinkles cit-ref-sprinkles-webofscience">CrossRef</a><a href="/external-ref?access_num=21337570&link_type=MED"
class="cit-ref-sprinkles cit-ref-sprinkles-medline">Medline</a><a target="_blank"
href="http://scholar.google.com/scholar_lookup?title=Neoadjuvant%20systemic%20treatment%20of%20breast%20cancer&author=I%20Makhoul&author=E%20Kiwan&publication_year=2011&journal=J%20Surg%20Oncol&volume=103&issue=4&pages=348-357">Google Scholar</a></div>
</div>
</li>
<li><span class="ref-label">2.</span>
<div class="cit ref-cit ref-journal no-rev-xref" id="cit-105.5.321.2">
<div class="cit-metadata">
<ol class="cit-auth-list">
<li><span class="cit-auth"><span class="cit-name-surname">Fisher</span> <span class="cit-name-given-names">B</span></span>,
</li>
<li><span class="cit-auth"><span class="cit-name-surname">Bryant</span> <span class="cit-name-given-names">J</span></span>,
</li>
<li><span class="cit-auth"><span class="cit-name-surname">Wolmark</span> <span class="cit-name-given-names">N</span></span>,
</li>
<li><span class="cit-etal">et al.</span></li>
</ol><cite>
<span class="cit-article-title">Effect of preoperative chemotherapy on the outcome of women with operable breast cancer</span>.
<abbr class="cit-jnl-abbrev">J Clin Oncol</abbr>.
<span class="cit-pub-date">1998</span>;
<span class="cit-vol">16</span>(<span class="cit-issue">8</span>):<span class="cit-fpage">2672</span>–<span class="cit-lpage">2685</span>.