-
Notifications
You must be signed in to change notification settings - Fork 0
/
Breast-Feeding and Cancer The Boyd Orr Cohort and a Systematic Review With Meta-Analysis.txt
2357 lines (1858 loc) · 199 KB
/
Breast-Feeding and Cancer The Boyd Orr Cohort and a Systematic Review With Meta-Analysis.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/97/19/1446.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>Breast-Feeding and Cancer: The Boyd Orr Cohort and a Systematic Review With Meta-Analysis </title>
<meta name="googlebot" content="NOODP" />
<meta name="HW.ad-path" content="/cgi/content/full/97/19/1446" />
<meta content="/jnci/97/19/1446.atom" name="HW.identifier" />
<meta name="DC.Format" content="text/html" />
<meta name="DC.Language" content="en" />
<meta content="Breast-Feeding and Cancer: The Boyd Orr Cohort and a Systematic Review With Meta-Analysis"
name="DC.Title" />
<meta content="10.1093/jnci/dji291" name="DC.Identifier" />
<meta content="2005-10-05" name="DC.Date" />
<meta content="Oxford University Press" name="DC.Publisher" />
<meta content="Richard M. Martin" name="DC.Contributor" />
<meta content="Nicos Middleton" name="DC.Contributor" />
<meta content="David Gunnell" name="DC.Contributor" />
<meta content="Christopher G. Owen" name="DC.Contributor" />
<meta content="George Davey Smith" 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="Richard M. Martin" />
<meta name="citation_author" content="Nicos Middleton" />
<meta name="citation_author" content="David Gunnell" />
<meta name="citation_author" content="Christopher G. Owen" />
<meta name="citation_author" content="George Davey Smith" />
<meta content="Breast-Feeding and Cancer: The Boyd Orr Cohort and a Systematic Review With Meta-Analysis"
name="citation_title" />
<meta content="10/05/2005" name="citation_date" />
<meta content="97" name="citation_volume" />
<meta content="19" name="citation_issue" />
<meta content="1446" name="citation_firstpage" />
<meta content="1457" name="citation_lastpage" />
<meta content="97/19/1446" name="citation_id" />
<meta content="97/19/1446" name="citation_id_from_sass_path" />
<meta content="jnci;97/19/1446" name="citation_mjid" />
<meta content="10.1093/jnci/dji291" name="citation_doi" />
<meta content="http://jnci.oxfordjournals.org/content/97/19/1446.abstract"
name="citation_abstract_html_url" />
<meta content="http://jnci.oxfordjournals.org/content/97/19/1446.full"
name="citation_fulltext_html_url" />
<meta content="http://jnci.oxfordjournals.org/content/97/19/1446.full.pdf"
name="citation_pdf_url" />
<meta content="http://jnci.oxfordjournals.org/content/97/19/1446"
name="citation_public_url" />
<meta content="16204694" name="citation_pmid" />
<meta name="citation_section" content="Article" />
<meta name="robots" content="noarchive,nofollow" />
<meta name="googlebot" content="noarchive" />
<link href="/content/97/19/1437.short" rel="prev" />
<link href="/content/97/19/1458.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-d29523030e1">var callbackToken='537479237A06E7C';</script><script type="text/javascript" id="session-d29523030e3">
var subCode='oupjournal_sub';
</script><script type="text/javascript">
var gAuthTimeStamp = '2015-10-27T12:15:56.473-07:00';
var gSessionId = 'jo7nvi9W7tGoavTvTPVKwA';
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/dji291");
var siqIsOpenAccess = encodeURIComponent("");
var siqPubDate = encodeURIComponent("5 October 2005");
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%2F97%2F19%2F1446.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/97/19.toc">
<span xmlns="" class="volume-value">Volume 97</span>
<span xmlns="" class="issue-value"> Issue 19</span></a></li>
<li>Pp. <span class="slug-pages">
1446-1457.
</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/97/19/1446.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">Breast-Feeding and Cancer: The Boyd Orr Cohort and a Systematic Review With Meta-Analysis</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=Richard+M.+Martin&sortspec=date&submit=Submit">Richard M. Martin</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=Nicos+Middleton&sortspec=date&submit=Submit">Nicos Middleton</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=David+Gunnell&sortspec=date&submit=Submit">David Gunnell</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=Christopher+G.+Owen&sortspec=date&submit=Submit">Christopher G. Owen</a></span> and
</li>
<li class="last" id="contrib-5"><span class="name"><a class="name-search"
href="/search?author1=George+Davey+Smith&sortspec=date&submit=Submit">George Davey Smith</a></span></li>
</ol>
<ol class="affiliation-list">
<li class="aff"><a id="aff-1" name="aff-1"></a><address><em>Affiliations of authors:</em> Department of Social Medicine, University of Bristol, Canynge Hall, Bristol, United Kingdom (RMM, NM, DG, GDS); Department
of Community Health Sciences, St. George's Hospital Medical School, London, United Kingdom (CGO)
</address>
</li>
</ol>
<ol class="corresp-list">
<li class="fn" id="corresp-1"><em>Correspondence to:</em> Richard Martin, BM, PhD, Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Road, Bristol BS8
2PR, United Kingdom (e-mail: <span class="em-link"><span class="em-addr">richard.martin{at}bristol.ac.uk</span></span>).
</li>
</ol>
<ul class="history-list">
<li xmlns:hwp="http://schema.highwire.org/Journal" class="received"
hwp:start="2005-02-22"><span class="received-label">Received </span>February 22, 2005.
</li>
<li xmlns:hwp="http://schema.highwire.org/Journal" class="rev-recd"
hwp:start="2005-07-22"><span class="rev-recd-label">Revision received </span>July 22, 2005.
</li>
<li xmlns:hwp="http://schema.highwire.org/Journal" class="accepted"
hwp:start="2005-08-03"><span class="accepted-label">Accepted </span>August 3, 2005.
</li>
</ul>
</div>
<div class="section abstract" id="abstract-1" itemprop="description">
<div class="section-nav">
<div class="nav-placeholder"> </div><a href="#sec-1" title="SUBJECTS AND METHODS" class="next-section-link"><span>Next Section</span></a></div>
<h2>Abstract</h2>
<p id="p-1"><em>Background:</em> Having been breast-fed has been suggested to influence cancer risk in adulthood. We investigated associations between breast-feeding
during infancy and adult cancer incidence and mortality in a cohort study and meta-analyses of published studies. <em>Methods:</em> The Boyd Orr cohort consisted of 4999 subjects who were originally surveyed in 1937–39, when they were 0–19 years of age.
Cancer outcomes from 1948 through 2003 were available for 4379 (88%) subjects, and 3844 had complete data on all covariates.
Associations of breast-feeding with cancer were investigated using proportional hazards models. We also identified 14 studies
on infant feeding and cancer published from 1966 through July 2005, of which 10 could be combined with the Boyd Orr cohort
results in a meta-analysis of breast cancer using random-effect models. <em>Results:</em> In the Boyd Orr cohort, ever having been breast-fed, compared with never having been breast-fed, was not associated with
the incidence of all cancers (hazard ratio [HR] = 1.07, 95% confidence interval [CI] = 0.89 to 1.28) or of any individual
cancer type examined (prostate HR = 1.43, 95% CI = 0.58 to 3.52; breast HR = 1.62, 95% CI = 0.89 to 2.94; colorectal HR =
0.86, 95% CI = 0.45 to 1.63; gastric HR = 1.22, 95% CI = 0.47 to 3.15). In the meta-analysis, there was also no association
between breast-feeding and breast cancer (regardless of menopausal status) (relative risk [RR] = 0.94, 95% CI = 0.85 to 1.04).
However, breast-fed women had a reduced risk of premenopausal breast cancer (RR = 0.88, 95% CI = 0.79 to 0.98) but not of
postmenopausal breast cancer (RR = 1.00, 95% CI = 0.86 to 1.16). <em>Conclusion:</em> Ever having been breast-fed was not associated with overall breast cancer risk, although the meta-analysis revealed a reduced
risk of premenopausal breast cancer in women who had been breast-fed.
</p>
</div>
<p id="p-2">Early-life environmental exposures may influence subsequent cancer risk <em>(<a id="xref-ref-1-1" class="xref-bibr" href="#ref-1">1</a>,<a id="xref-ref-2-1" class="xref-bibr" href="#ref-2">2</a>)</em>. For example, taller individuals are at a 20%–60% increased risk of a range of cancers <em>(<a id="xref-ref-3-1" class="xref-bibr" href="#ref-3">3</a>)</em>, indicating the possible importance of growth-promoting factors in the development of cancer. Increased height may be a marker
for exposure to higher levels of insulin-like growth factor I (IGF-I) in childhood <em>(<a id="xref-ref-4-1" class="xref-bibr" href="#ref-4">4</a>)</em>, and breast, prostate, and colorectal cancers have been positively associated with levels of IGF-I in adulthood in both cohort
and case-control studies <em>(<a id="xref-ref-5-1" class="xref-bibr" href="#ref-5">5</a>)</em>.
</p>
<p id="p-3">Another early-life exposure that can differ among individuals and possibly modify subsequent cancer risk is exposure to breast
milk. Breast-feeding is positively associated with both stature <em>(<a id="xref-ref-6-1" class="xref-bibr" href="#ref-6">6</a>,<a id="xref-ref-7-1" class="xref-bibr" href="#ref-7">7</a>)</em> and circulating IGF-I levels <em>(<a id="xref-ref-8-1" class="xref-bibr" href="#ref-8">8</a>)</em> in later childhood, raising the possibility that breast-feeding may contribute to associations between height/IGF-I and cancer
<em>(<a id="xref-ref-5-2" class="xref-bibr" href="#ref-5">5</a>,<a id="xref-ref-9-1" class="xref-bibr" href="#ref-9">9</a>)</em>. Breast-feeding could, in theory, also affect cancer risk if it is the source of a carcinogenic substance; in the 1930s,
it was hypothesized that an oncogenic virus transmitted in breast milk causes subsequent breast cancer in offspring <em>(<a id="xref-ref-10-1" class="xref-bibr" href="#ref-10">10</a>)</em>. Although the epidemiologic evidence of such an effect was limited <em>(<a id="xref-ref-11-1" class="xref-bibr" href="#ref-11">11</a>)</em>, mothers with a family history of breast cancer were advised not to breast-feed their daughters. Results of subsequent studies
relating having been breast-fed with breast cancer are inconclusive <em>(<a id="xref-ref-12-1" class="xref-bibr" href="#ref-12">12</a>)</em>. Increasing interest in perinatal factors associated with testicular cancer has also led to an analysis of possible associations
of breast-feeding with this tumor <em>(<a id="xref-ref-13-1" class="xref-bibr" href="#ref-13">13</a>)</em>.
</p>
<p id="p-4">To gain a better understanding of possible cancer risks associated with having been breast-fed, we investigated the association
of breast-feeding in infancy with adult cancer risk in a 65-year follow-up of the Boyd Orr cohort <em>(<a id="xref-ref-14-1" class="xref-bibr" href="#ref-14">14</a>)</em>. This cohort is based on the long-term follow-up of the Carnegie (Lord Boyd Orr) study of Family Diet and Health in Pre-War
Britain (1937–1939) <em>(<a id="xref-ref-15-1" class="xref-bibr" href="#ref-15">15</a>)</em>, which was originally designed to investigate “the connection between economic factors and physical welfare” <em>(<a id="xref-ref-16-1" class="xref-bibr" href="#ref-16">16</a>)</em> and which was reconstructed as an historical cohort in 1988 to investigate a range of disease endpoints, particularly coronary
heart disease and cancer, in relation to infant and childhood diet, to the socioeconomic conditions experienced by the children,
and to markers of childhood nutritional status (body mass index [BMI], leg length, and height) <em>(<a id="xref-ref-17-1" class="xref-bibr" href="#ref-17">17</a>)</em>.
</p>
<p id="p-5">Information on the breast-feeding history of cohort members is available, and the members are now at an age (range, 64–85
years) at which cancer is a substantial burden. Because breast-feeding is positively associated with height <em>(<a id="xref-ref-6-2" class="xref-bibr" href="#ref-6">6</a>,<a id="xref-ref-7-2" class="xref-bibr" href="#ref-7">7</a>)</em> and IGF-I <em>(<a id="xref-ref-8-2" class="xref-bibr" href="#ref-8">8</a>)</em> and because both are, in turn, positively associated with breast, prostate, and colorectal cancers <em>(<a id="xref-ref-3-2" class="xref-bibr" href="#ref-3">3</a>,<a id="xref-ref-5-3" class="xref-bibr" href="#ref-5">5</a>)</em>, we hypothesized a priori that breast-feeding may be associated with an increase in the risk of these cancers. We also hypothesized
an inverse relationship between breast-feeding and gastric cancer, because breast-feeding is associated with a lower prevalence
of <em>Helicobacter pylori</em> infection <em>(<a id="xref-ref-18-1" class="xref-bibr" href="#ref-18">18</a>)</em>, which has been implicated in gastric cancer etiology <em>(<a id="xref-ref-19-1" class="xref-bibr" href="#ref-19">19</a>)</em>. We placed the results of the Boyd Orr cohort in context by conducting a systematic review and meta-analysis of the published
literature on the relationship between breast-feeding in infancy and adult cancer.
</p>
<div class="section" id="sec-1">
<div class="section-nav"><a href="#abstract-1" title="Abstract" class="prev-section-link"><span>Previous Section</span></a><a href="#sec-6" title="RESULTS" class="next-section-link"><span>Next Section</span></a></div>
<h2>S<span class="sc">UBJECTS AND</span> M<span class="sc">ETHODS</span></h2>
<h3>Boyd Orr Cohort</h3>
<p id="p-6">The methods used in the Boyd Orr cohort have been described previously <em>(<a id="xref-ref-20-1" class="xref-bibr" href="#ref-20">20</a>)</em>. Briefly, the cohort comprised 4999 children who were aged 0–19 years at enrollment in 1937–1939. The cohort members came
from 1343 families living in 16 urban and rural districts in Britain; they underwent a 1-week assessment of family diet and
health at enrollment (the Carnegie Survey of Family Diet and Health in Pre-War Britain) <em>(<a id="xref-ref-15-2" class="xref-bibr" href="#ref-15">15</a>)</em>. We used the National Health Service Central Register (NHSCR) to trace 4379 (88%) of the original study members <em>(<a id="xref-ref-14-2" class="xref-bibr" href="#ref-14">14</a>)</em>, who have been followed up since the inception of the NHSCR in 1948 through February 28, 2003, and flagged for cancer registration
and mortality. The United Bristol's Hospital Trust Local Research Ethics Committee provided ethical approval for flagging
and tracing the Boyd Orr cohort.
</p>
<p id="p-7">From the original survey data, we obtained information on the method of infant feeding (recalled by the mother an average
of 7 years after birth), age at baseline survey, sex, per capita weekly household food expenditure group (six categories),
birth order, and survey district <em>(<a id="xref-ref-15-3" class="xref-bibr" href="#ref-15">15</a>)</em>. Using the original survey records from 1937–1939, we coded history of infant feeding as either ever breast-fed or never
breast-fed. Duration of breast-feeding was coded as follows: <6 months; 6–11 months; >11 months; or unknown. These cut points
were chosen to assess the effects of both the currently recommended duration of breast-feeding (<6 months) and prolonged breast-feeding
(>11 months). Socioeconomic status of the head of the household was assigned to one of eight categories (social class I, II,
III, IV, or V; unemployed; armed forces; unclassifiable) using the Registrar General's 1931 classification <em>(<a id="xref-ref-20-2" class="xref-bibr" href="#ref-20">20</a>)</em>. Based on single measurements of standing height, leg length, and body weight at the time of the original survey, internally
age- and sex-standardized z-scores for measured childhood height, leg length, and BMI were computed <em>(<a id="xref-ref-21-1" class="xref-bibr" href="#ref-21">21</a>)</em>. Because height measurement in children under 2 years of age tends to be unreliable and because of a large amount of missing
data in children of this age and in the 15 and over age band, z-scores were calculated only for the subset of children aged
between 2 and 14.75 years at the time of the original survey, as in previous reports (n = 1191 women and 1103 men in the current
analysis) <em>(<a id="xref-ref-21-2" class="xref-bibr" href="#ref-21">21</a>)</em>. Further information on diet, health, and lifestyle was obtained for 1648 subjects who completed a questionnaire in 1997–1998
<em>(<a id="xref-ref-17-2" class="xref-bibr" href="#ref-17">17</a>)</em>.
</p>
<p id="p-8">Cancer incidence and cause of death in the Boyd Orr cohort members, based on data obtained from the NHSCR, were defined by
the International Classification of Diseases Ninth (ICD-9) or Tenth (ICD-10) Revision. The outcomes included all cancers (ICD-9,
140–208; ICD-10, C0–C97); breast cancer (ICD-9, 174; ICD-10, C50); colorectal cancer (ICD-9, 153–154, excluding 154.2 and
154.3, which are cancers of the anal canal and anus; ICD-10, C18–20); prostate cancer (ICD-9, 185; ICD-10, C61); gastric cancer
(ICD-9, 151; ICD-10, C16); all cancers thought to be etiologically related to smoking, including cancers of the mouth and
oro-pharynx (ICD-9, 140–149 and 160; ICD-10, C0–C14 and C30–C31), larynx (ICD-9, 161; ICD-10, C32), other sections of the
respiratory tract (ICD-9, 165; ICD-10, C39), pancreas (ICD-9, 157; ICD-10, C25), trachea and lung (ICD-9, 162; ICD-10, C33–C34),
and bladder (ICD-9, 188; ICD-10, C67); and all cancers excluding those thought to be etiologically related to smoking. We
subdivided cancers into those thought to be related to smoking and those thought to be unrelated to smoking because smoking
may confound associations between breast-feeding and smoking-related cancers and because information on smoking was available
for only approximately 1000 subjects with breast-feeding data who were still alive for a questionnaire survey in 1997.
</p>
<p id="p-9">Complete data on all covariates were available for 3855 subjects (1889 males and 1966 females), and these subjects were included
in the analyses relating having been breast-fed with cancer mortality (n = 363 cancer deaths). Eleven subjects with an incident
cancer lacked a date for the outcome, and these subjects were therefore excluded from the total number of subjects included
in the analyses of cancer incidence in relation to having been breast-fed (n = 3844 subjects; 1883 males and 1961 females;
587 incident cancers).
</p>
<h3>Statistical Analysis of the Boyd Orr Cohort</h3>
<p id="p-10">The outcomes were total and site-specific cancer incidence and mortality. Both endpoints were analyzed because cancer mortality
is associated with socioeconomic position and because associations of having been breast-fed with cancer incidence may differ
from associations with cancer mortality. The cancer incidence outcome was derived from the first cancer that was registered
or, if no cancer had been registered prior to death, from the presence of a cancer code anywhere on the death certificate.
Subsequent registered cancers were not included in the analysis because these could be secondary cancers or could have arisen
due to adverse effects of treatment (in any event, only 22 of the 587 individuals with cancer had more than one cancer registration).
For breast cancer, the main outcome was all cancers irrespective of menopausal status at diagnosis. We also separately examined
breast cancers diagnosed in women under 50 years of age (n = 13) and in women 50 years of age and over (n = 61), but it was
not possible to use a clinical definition of menopause because we had only the date of cancer diagnosis. The single case of
male breast cancer was excluded from the analyses.
</p>
<p id="p-11">The association of breast-feeding initiation and duration with cancer outcomes was investigated using Cox proportional hazards
models. Cohort members who were never breast-fed formed the reference group. Follow-up was censored on February 28, 2003.
Subjects who had been traced but with whom contact via the NHSCR had been lost after 1948 (for example, if the subject is
not currently registered with a Health Authority doctor) or who emigrated or died were included in the survival analysis up
to the date of death, emigration, or last contact. Losses to follow-up were 11.3% among those breast-fed and 12.3% among those
never breast-fed (<em>P</em> = .3).
</p>
<p id="p-12">Because age is a strong determinant of mortality risk, and because individuals entered the study over a 19-year range of ages
(0–19 years) and over a 2-year period (1937–1939), we controlled for current age in all models. Because both the prevalence
of breast-feeding <em>(<a id="xref-ref-6-3" class="xref-bibr" href="#ref-6">6</a>)</em> and cancer rates differed substantially between survey areas, all models were stratified by survey district, thus allowing
for district-specific baseline rate parameters. Clustering effects may have arisen because most subjects in the cohort belonged
to families that included other cohort members and therefore shared childhood conditions and possible genetic effects on cancer;
we calculated robust standard errors to allow for a between-family component of variation <em>(<a id="xref-ref-22-1" class="xref-bibr" href="#ref-22">22</a>)</em>.
</p>
<p id="p-13">The analyses controlling for age, survey district, and clustering form the simple models presented in the results. Multivariable
models were then developed that controlled additionally for sex (except for sex-specific cancers), socioeconomic status of
the child's father at the time of the original survey, the child's birth order, and per capita weekly household food expenditure
in childhood. The proportional hazards assumption was investigated both graphically and by formally testing that the log hazard
ratio (HR) was constant over time for each model <em>(<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>)</em>. There was no evidence against the proportional hazards assumption of a constant hazard ratio over time. We assessed whether
the associations with breast-feeding differed according to sex and age at original survey using the likelihood ratio test
in fully adjusted models.
</p>
<p id="p-14">Because differences in growth <em>(<a id="xref-ref-6-4" class="xref-bibr" href="#ref-6">6</a>,<a id="xref-ref-7-3" class="xref-bibr" href="#ref-7">7</a>)</em> and in circulating IGF-I levels <em>(<a id="xref-ref-8-3" class="xref-bibr" href="#ref-8">8</a>)</em> between those breast- and bottle-fed may be on the biologic pathway between mode of infant feeding and subsequent cancer
risk, we examined the effect on breast-feeding-cancer associations of adjustment for age- and sex-standardized z-scores for
measured childhood height and BMI in the subset of the cohort with available anthropometry in childhood. Any attenuation in
effect estimates could indicate that childhood height or adiposity is an intermediary variable or marker for confounding factors
operating in childhood.
</p>
<p id="p-15">Finally, changes in feeding patterns—for example, with respect to exclusivity of breast-feeding or alternatives to breast-feeding—may
have occurred over the range of the years of births of the subjects (1918–1939). Therefore, we also tested for interaction
by year of birth, which was dichotomized for this analysis as before 1930 or in 1930 or later (median year of birth).
</p>
<h3>Systematic Review</h3>
<p id="p-16">The data retrieved for the systematic review were based on a systematic search (completed by R.M.M.) of all published papers,
letters, abstracts, and review articles on infant feeding and cancer using the MEDLINE database from January 1966 through
to June 2004. We used a combined text word and MeSH heading search strategy, with terms for infant feeding combined with terms
for cancer (see <a id="xref-sec-14-1" class="xref-sec" href="#sec-14">Appendix</a>). We also manually searched the reference lists of all studies that fulfilled the inclusion criteria for further relevant
publications. The search began in June 2004 and was repeated weekly through July 2005 using the automated OVID alert system.
</p>
<p id="p-17">Articles were included if they fulfilled the following criteria: 1) infants who had been breast-fed were compared with those
never breast-fed; 2) the outcome was cancer incidence or mortality; and 3) quantitative estimates of the association of having
been breast-fed and cancer outcomes were available or could be derived. Articles that related breast-feeding with cancers
in childhood or adolescence (up to age 19 years) were excluded from the present analysis but are the subject of a separate
report <em>(<a id="xref-ref-25-1" class="xref-bibr" href="#ref-25">25</a>)</em>. R.M.M. extracted the data on two separate occasions to check the consistency of the data extraction.
</p>
<p id="p-18">A meta-analysis of the associations of breast-feeding with all cancers and with specific cancers was conducted that included
the findings from the Boyd Orr study. Cancer in those who had ever or exclusively been breast-fed was compared with that in
those who had never been breast-fed. If results for both ever and exclusive breast-feeding were presented, the exclusive breast-feeding
association was used in the meta-analysis. For some studies, we calculated relative risks (RRs) from reported prevalence rates
of cancer in different infant feeding groups or by using fixed-effects models to combine relative risks given for different
durations of breast-feeding <em>(<a id="xref-ref-13-2" class="xref-bibr" href="#ref-13">13</a>,<a id="xref-ref-26-1" class="xref-bibr" href="#ref-26">26</a>–<a id="xref-ref-28-1" class="xref-bibr" href="#ref-28">28</a>)</em>. To assess the impact of extended breast-feeding, separate meta-analyses comparing any or exclusive breast-feeding of ≥6
months with never breast-feeding were undertaken.
</p>
<p id="p-19">We calculated the I<sup>2</sup> statistic as a quantitative measure of the degree of inconsistency across studies that is not dependent on the number of
studies <em>(<a id="xref-ref-29-1" class="xref-bibr" href="#ref-29">29</a>)</em>. An I<sup>2</sup> value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity. Because it is conceivable
that any associations of cancer with breast-feeding could depend on its duration or exclusivity, the population studied, or
the era in which the subjects were born, effect estimates from the individual studies were pooled using random-effects models
(although results from fixed-effects models were similar; data not shown). Random-effects models are appropriate in this instance
because there was little evidence of publication bias. The Egger regression test was conducted to examine the relationship
between sample size and observed cancer risk by infant feeding group <em>(<a id="xref-ref-30-1" class="xref-bibr" href="#ref-30">30</a>,<a id="xref-ref-31-1" class="xref-bibr" href="#ref-31">31</a>)</em>.
</p>
<h3>Sensitivity Analysis</h3>
<p id="p-20">Because no generally accepted lists of appropriate quality criteria for observational studies are available, we investigated
factors that might explain differences between studies as a way to gain clues to possible sources of systematic bias. Selected
study characteristics, chosen a priori, were thus entered as indicator variables in separate meta-regression analyses <em>(<a id="xref-ref-32-1" class="xref-bibr" href="#ref-32">32</a>)</em> to assess their impact on between-study variation (heterogeneity). Meta-regression analysis offers a conservative test of
the effect of certain exposures on outcome, assessed at study level <em>(<a id="xref-ref-32-2" class="xref-bibr" href="#ref-32">32</a>)</em>. These characteristics were study design (case–control or cohort/nested case–control study, i.e., whether infant feeding
information was obtained retrospectively or prospectively); study size (≤500 cases/>500 cases); reliance on maternal recall
of breast-feeding beyond infancy (with infancy defined as <1 years; coded as yes/no); whether effect estimates in the final
models controlled for socioeconomic factors in childhood or adulthood (yes/no); whether effect estimates in the final models
controlled for reproductive factors (yes/no); whether the study was population or hospital based (yes/no); and whether the
response rate was less than 80% (yes/no). Other study characteristics that may be true effect modifiers of the association
between breast-feeding and cancer were also examined, including the region in which the study was conducted (categorized as
United Kingdom, North America, Europe, or other); the year of birth of the participants (dichotomized as 1970 or earlier versus
after 1970); and whether the prevalence of any breast-feeding was at least 70%.
</p>
<p id="p-21">We used two-sided tests of statistical significance, and the precision of our estimates was based on 95% confidence limits
throughout. No formal statistical approaches to account for multiple hypothesis testing were used, but we have quoted exact
rather than threshold <em>P</em> values. All statistical analyses were performed using Stata 8.0 <em>(<a id="xref-ref-33-1" class="xref-bibr" href="#ref-33">33</a>)</em>.
</p>
</div>
<div class="section" id="sec-6">
<div class="section-nav"><a href="#sec-1" title="SUBJECTS AND METHODS" class="prev-section-link"><span>Previous Section</span></a><a href="#sec-13" title="DISCUSSION" class="next-section-link"><span>Next Section</span></a></div>
<h2>R<span class="sc">ESULTS</span></h2>
<h3>Boyd Orr Cohort</h3>
<p id="p-22">Of the 4999 original survey participants, 3844 (1961 females and 1883 males) included in the analysis of cancer incidence
contributed 185 458 person-years of observation between January 1, 1948, and February 28, 2003. Of these, 2716 participants
(71%) were breast-fed and the median duration of breast-feeding was 9 months (range, 0.5–25 months). The prevalence and duration
of breast-feeding did not differ by sex, socioeconomic status of the father, or decade of birth, but both were positively
associated with household food expenditure <em>(<a id="xref-ref-14-3" class="xref-bibr" href="#ref-14">14</a>)</em>.
</p>
<p id="p-23">Ever having been breast-fed was not associated with all cancers, with smoking-related cancers, or with non–smoking-related
cancers (<a id="xref-table-wrap-1-1" class="xref-table" href="#T1">Table 1</a>). There was weak evidence of an association between breast-feeding in infancy and risk of breast cancer in adulthood (HR
= 1.62, 95% confidence interval [CI] = 0.89 to 2.94; <em>P</em> = .11). This association was the same regardless of the child's age when the mother was interviewed (<em>P</em> for age–outcome interaction = .8). The association also did not differ by the woman's age at cancer diagnosis (i.e., less
than 50 years or 50 years or older; <em>P</em> for interaction = .5). The direction and size of the association with breast cancer mortality (HR = 1.53, 95% CI = 0.61 to
3.83; <em>P</em> = .37) was the same as that for breast cancer incidence. Adjusting for childhood BMI or height did not explain the association
between breast-feeding and breast cancer (data not shown), suggesting that growth in childhood is unlikely to be on the causal
pathway linking breast-feeding with breast cancer.
</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="1446/T1.expansion.html">In this window</a></li>
<li><a class="in-nw" href="1446/T1.expansion.html">In a new window</a></li>
</ul>
</div>
</div>
<div class="table-caption"><span class="table-label"><strong>Table 1.</strong> </span>
<p id="p-24" class="first-child">HRs (with 95% CIs) for cancer incidence and mortality in relation to ever having been breast-fed compared with never having
been breast-fed in the Boyd Orr cohort*
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<p id="p-31">In addition, we found no evidence that ever having been breast-fed was associated with incident prostate, colorectal, or gastric
cancer. As with breast cancer, however, the confidence limits were wide and, therefore, the results are consistent with the
possibility that having been breast-fed is associated with increased or reduced risks of these cancers. There was no evidence
that associations between breast-feeding and cancer outcomes varied by sex or year of birth (<em>P</em> for interaction > .1 for all cancer outcomes). There was no association between increased duration of breast-feeding and
any of the cancers examined (data not shown).
</p>
<h3>Systematic Review</h3>
<p id="p-32">The search strategy (see <a id="xref-sec-14-2" class="xref-sec" href="#sec-14">Appendix</a>) yielded 1415 hits, of which 78 articles met the inclusion criteria outlined in the Subjects and Methods. After detailed
review of these 78 potentially relevant reports, 14 <em>(<a id="xref-ref-13-3" class="xref-bibr" href="#ref-13">13</a>,<a id="xref-ref-26-2" class="xref-bibr" href="#ref-26">26</a>–<a id="xref-ref-28-2" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-1" class="xref-bibr" href="#ref-34">34</a>–<a id="xref-ref-43-1" class="xref-bibr" href="#ref-43">43</a>)</em> were included in one or more of the meta-analyses (<a id="xref-table-wrap-2-1" class="xref-table" href="#T2">Table 2</a>). For two of the studies, the outcome was all cancer <em>(<a id="xref-ref-26-3" class="xref-bibr" href="#ref-26">26</a>,<a id="xref-ref-35-1" class="xref-bibr" href="#ref-35">35</a>)</em>, for 11 (including one of the studies that reported on all cancer), it was breast cancer <em>(<a id="xref-ref-27-1" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-3" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-2" class="xref-bibr" href="#ref-34">34</a>–<a id="xref-ref-41-1" class="xref-bibr" href="#ref-41">41</a>,<a id="xref-ref-43-2" class="xref-bibr" href="#ref-43">43</a>)</em>, and for two, it was testicular cancer <em>(<a id="xref-ref-13-4" class="xref-bibr" href="#ref-13">13</a>,<a id="xref-ref-42-1" class="xref-bibr" href="#ref-42">42</a>)</em>. Of the 64 potentially relevant studies that were excluded from these meta-analyses, 46 related to childhood cancers [and
are the subject of a separate report <em>(<a id="xref-ref-25-2" class="xref-bibr" href="#ref-25">25</a>)</em>], 7 were reviews and 11 either did not report on breast-feeding-cancer outcomes or considered only breast cancers among breast-feeding
mothers.
</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="1446/T2.expansion.html">In this window</a></li>
<li><a class="in-nw" href="1446/T2.expansion.html">In a new window</a></li>
</ul>
</div>
</div>
<div class="table-caption"><span class="table-label"><strong>Table 2.</strong> </span>
<p id="p-33" class="first-child">Studies of breast-feeding and cancers in adulthood included in the meta-analyses by cancer type and in descending order of
year of publication*
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<h3>Breast-Feeding and All Cancers</h3>
<p id="p-44">In a meta-analysis of our Boyd Orr cohort findings together with the two published studies that examined the association of
breast-feeding with all cancers <em>(<a id="xref-ref-26-6" class="xref-bibr" href="#ref-26">26</a>,<a id="xref-ref-35-4" class="xref-bibr" href="#ref-35">35</a>)</em>, involving 802 cancer cases in total, there was no association between breast-feeding and all cancers (random-effects model:
RR = 0.95, 95% CI = 0.71 to 1.26; <em>P</em> = .7) (<a id="xref-fig-1-1" class="xref-fig" href="#F1">Fig. 1, A</a>). There was strong evidence of heterogeneity (I<sup>2</sup> statistic = 71%), and the possibility of a substantial increase or decrease in the risk of all cancers with breast-feeding
cannot be discounted. An analysis of factors explaining this heterogeneity is not possible because meta-regression analyses
based on only three studies could yield chance associations.
</p>
<div id="F1" class="fig pos-float odd">
<div class="fig-inline"><a href="1446/F1.expansion.html"><img alt="Fig. 1." src="1446/F1.small.gif" /></a><div class="callout"><span>View larger version:</span><ul class="callout-links">
<li><a href="1446/F1.expansion.html">In this window</a></li>
<li><a class="in-nw" href="1446/F1.expansion.html">In a new window</a></li>
</ul>
<ul class="fig-services">
<li class="ppt-link"><a href="/powerpoint/97/19/1446/F1">Download as PowerPoint Slide</a></li>
</ul>
</div>
</div>
<div class="fig-caption"><span class="fig-label"><strong>Fig. 1.</strong></span>
<p id="p-45" class="first-child">Relative risks and 95% confidence intervals (CIs) for cancer incidence, comparing individuals who were ever breast-fed as
infants with those who were never breast-fed, from a meta-analysis of published studies and Boyd Orr [shown as Martin (2004)].
<strong>A</strong>) Incidence of all cancer. <strong>B</strong>) Incidence of breast cancer. <strong>C</strong>) Incidence of premenopausal breast cancer. <strong>D</strong>) Incidence of postmenopausal breast cancer. The study author is indicated on the y-axis (ordered by year of publication).
The box for each study is proportional to the inverse of the variance; horizontal lines show 95% CIs on the relative risk.
The pooled estimates, based on a random-effects model, are shown by a dashed vertical line and diamond (95% CI).
</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<h3>Breast-Feeding and Breast Cancer</h3>
<p id="p-46">Ten published studies <em>(<a id="xref-ref-27-4" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-5" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-5" class="xref-bibr" href="#ref-34">34</a>–<a id="xref-ref-41-4" class="xref-bibr" href="#ref-41">41</a>)</em> plus the Boyd Orr cohort, involving 11 564 breast cancer cases in total, were included in the meta-analyses of the association
of breast-feeding with breast cancer. [An additional published study <em>(<a id="xref-ref-43-4" class="xref-bibr" href="#ref-43">43</a>)</em> was not included in the meta-analysis because this study investigated the association of breast-feeding with familial compared
with sporadic breast cancer rather than whether breast-feeding was associated with incident breast cancer per se.] Aspects
of the quality of these studies are summarized in <a id="xref-table-wrap-3-2" class="xref-table" href="#T3">Table 3</a>. Cohorts are considered to provide more robust estimates than case-control studies in the hierarchy of evidence, but only
three cohort studies examining the breast-feeding–breast cancer association were identified: the Boyd Orr cohort, Michels
et al. (the Nurses' Health Study) <em>(<a id="xref-ref-27-5" class="xref-bibr" href="#ref-27">27</a>)</em>, and Ekbom et al. (a record-linkage study based on the Swedish cancer registry) <em>(<a id="xref-ref-38-3" class="xref-bibr" href="#ref-38">38</a>)</em>. Nine studies relied on the long-term recall or reporting of having been breast-fed as a child among participants who were
questioned after the diagnosis of breast cancer, so the responses may have been influenced by recall bias. The exceptions
were the current Boyd Orr study and Ekbom's record-linkage study, in which infant feeding mode at discharge was recorded (on
average 10 days after delivery) among infants born between 1874 and 1954 <em>(<a id="xref-ref-38-4" class="xref-bibr" href="#ref-38">38</a>)</em>.
</p>
<div class="table pos-float" id="T3">
<div class="table-inline">
<div class="callout"><span>View this table:</span><ul class="callout-links">
<li><a href="1446/T3.expansion.html">In this window</a></li>
<li><a class="in-nw" href="1446/T3.expansion.html">In a new window</a></li>
</ul>
</div>
</div>
<div class="table-caption"><span class="table-label"><strong>Table 3.</strong> </span>
<p id="p-47" class="first-child">Characteristics and quality of published studies relating ever having been breast-fed with breast cancer risk*</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
<p id="p-50">All studies of the breast cancer association defined breast-feeding as ever having been breast-fed; none examined exclusive
breast-feeding beyond the first few days. Only some studies controlled for potentially important recognized confounders, including
reproductive factors (n = 6) <em>(<a id="xref-ref-27-6" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-6" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-6" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-36-2" class="xref-bibr" href="#ref-36">36</a>–<a id="xref-ref-38-5" class="xref-bibr" href="#ref-38">38</a>)</em>, family history of breast cancer (n = 5) <em>(<a id="xref-ref-27-7" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-7" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-7" class="xref-bibr" href="#ref-34">34</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>)</em> or a measure of socioeconomic/educational status (n = 5) <em>(<a id="xref-ref-34-8" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-37-3" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-38-6" class="xref-bibr" href="#ref-38">38</a>,<a id="xref-ref-40-2" class="xref-bibr" href="#ref-40">40</a>)</em> (and Boyd Orr). All six studies published since 1990, however, included multivariable adjustment for at least two of the
following: reproductive factors, family history of breast cancer, or socioeconomic/educational status <em>(<a id="xref-ref-27-8" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-8" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-9" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-36-4" class="xref-bibr" href="#ref-36">36</a>–<a id="xref-ref-38-7" class="xref-bibr" href="#ref-38">38</a>)</em>. Six case–control studies provided information on response rates <em>(<a id="xref-ref-28-9" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-10" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-37-4" class="xref-bibr" href="#ref-37">36</a>,<a id="xref-ref-37-5" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-39-2" class="xref-bibr" href="#ref-39">39</a>,<a id="xref-ref-40-3" class="xref-bibr" href="#ref-40">40</a>)</em>, and in all of these studies the response rates among both case patients and control subjects were less than 80%, raising
the possibility of selection bias.
</p>
<p id="p-51">Individually, five of the 11 studies included in the meta-analysis of breast cancer <em>(<a id="xref-ref-34-11" class="xref-bibr" href="#ref-34">34</a>–<a id="xref-ref-37-6" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-39-3" class="xref-bibr" href="#ref-39">39</a>)</em> suggested that women who had been breast-fed had a reduced risk of breast cancer (regardless of menopausal status), although
the associations were statistically significant in only two studies <em>(<a id="xref-ref-36-5" class="xref-bibr" href="#ref-36">36</a>,<a id="xref-ref-37-7" class="xref-bibr" href="#ref-37">37</a>)</em>. There was no evidence for an association between breast-feeding and breast cancer in the meta-analysis (RR = 0.94, 95% CI
= 0.85 to 1.04; <em>P</em> = .25) (<a id="xref-fig-1-2" class="xref-fig" href="#F1">Fig. 1, B</a>). There was moderate among-study heterogeneity (I<sup>2</sup> = 31%) but no evidence of small-study bias (Egger test <em>P</em> = .7).
</p>
<p id="p-52">Only three of the 11 studies <em>(<a id="xref-ref-27-9" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-10" class="xref-bibr" href="#ref-28">28</a>)</em> (and Boyd Orr) had information on duration of breast-feeding and breast cancer. In these studies, the association with breast
cancer was similar in those breast-fed <6 months compared with those never breast-fed (RR = 0.99, 95% CI = 0.86 to 1.15) and
in those breast-fed for 6 months or longer compared with those never breast-fed (RR = 1.21, 95% CI = 0.94 to 1.55), with little
evidence of heterogeneity in each duration specific analysis (I<sup>2</sup> = 0% and 29%, respectively).
</p>
<p id="p-53">Nine studies related having been breast-fed to premenopausal breast cancer <em>(<a id="xref-ref-27-10" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-11" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-12" 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-40-4" class="xref-bibr" href="#ref-40">40</a>)</em> (and Boyd Orr), with a total of 3347 cases of premenopausal breast cancer. Of these studies, six <em>(<a id="xref-ref-27-11" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-34-13" 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-39-4" class="xref-bibr" href="#ref-39">39</a>)</em> suggested a reduced risk of premenopausal breast cancer in women who had been breast-fed (<a id="xref-fig-1-3" class="xref-fig" href="#F1">Fig. 1, C</a>). In random-effects meta-analysis, the RR was 0.88 (95% CI = 0.79 to 0.98; <em>P</em> = .018); the heterogeneity was low (I<sup>2</sup> = 2%), suggesting that the study results were consistent with one another. There was no evidence of small-study bias (Egger
test <em>P</em> = .6). Five studies <em>(<a id="xref-ref-27-12" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-34-14" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-37-8" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-38-8" class="xref-bibr" href="#ref-38">38</a>)</em> (and Boyd Orr) examined associations of having been breast-fed with risk of postmenopausal breast cancer (5069 cases in total);
in a meta-analysis, there was no association (RR = 1.00, 95% CI = 0.86 to 1.16; <em>P</em> = 0.99) (<a id="xref-fig-1-4" class="xref-fig" href="#F1">Fig. 1, D</a>). These studies were reasonably consistent with one another (I<sup>2</sup> = 27%), and there was no evidence of small-study effects (Egger test <em>P</em> = .6). An analysis of the two Nurses' Health Study cohorts <em>(<a id="xref-ref-27-13" class="xref-bibr" href="#ref-27">27</a>)</em> found weak evidence of a higher risk of postmenopausal breast cancer in women who had been breast-fed for at least 9 months
versus those who had never been breast-fed, whether the analysis was based on self-reported infant feeding history (RR = 1.30,
95% CI = 0.98 to 1.72) or on information provided by the participants' mothers in a subsample of the cohorts (RR = 1.72, 95%
CI = 0.99 to 3.00).
</p>
<p id="p-54">Four studies examined associations between having been breast-fed and familial compared with sporadic breast cancer <em>(<a id="xref-ref-34-15" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-35-5" class="xref-bibr" href="#ref-35">35</a>,<a id="xref-ref-41-5" class="xref-bibr" href="#ref-41">41</a>,<a id="xref-ref-43-5" class="xref-bibr" href="#ref-43">43</a>)</em>. The data in three of these studies could be meta-analyzed <em>(<a id="xref-ref-34-16" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-41-6" class="xref-bibr" href="#ref-41">41</a>,<a id="xref-ref-43-6" class="xref-bibr" href="#ref-43">43</a>)</em> to assess whether a history of having been breast-fed is more common among women with familial breast cancer than women with
sporadic breast cancers [a test of the Bittner hypothesis <em>(<a id="xref-ref-10-2" class="xref-bibr" href="#ref-10">10</a>)</em> that a breast cancer–causing agent is transmitted in breast milk]. The pooled relative risk for the association of having
been breast-fed in relation to familial versus sporadic breast cancers was 1.05 (95% CI = 0.81 to 1.35; I<sup>2</sup> = 35%; <em>P</em> = .7), providing no evidence to support the suggestion that a transmissible agent in breast milk increases breast cancer
risk. The study excluded from the above meta-analysis reported that the risks of breast cancer among breast-fed women compared
with non-breast-fed women were the same regardless of maternal history of breast cancer <em>(<a id="xref-ref-35-6" class="xref-bibr" href="#ref-35">35</a>)</em>.
</p>
<h3>Breast-Feeding and Testicular Cancer</h3>
<p id="p-55">A meta-analysis of the two studies that provided testicular cancer outcomes (with 524 cases of testicular cancer in total)
provided only weak evidence of a lower risk of testicular cancer in breast-fed men <em>(<a id="xref-ref-13-6" class="xref-bibr" href="#ref-13">13</a>,<a id="xref-ref-42-2" class="xref-bibr" href="#ref-42">42</a>)</em> (pooled RR = 0.82, 95% CI = 0.62 to 1.10; I<sup>2</sup> = 0%; <em>P</em> = .18). One of the two studies, that of Coupland <em>(<a id="xref-ref-13-7" class="xref-bibr" href="#ref-13">13</a>)</em>, suggested a duration–response relationship between breast-feeding and a lower risk of testicular cancer (<em>P</em> for trend = .05), but recall bias was possible because mothers provided information about method of infant feeding after
the men were diagnosed at 15–49 years of age <em>(<a id="xref-ref-13-8" class="xref-bibr" href="#ref-13">13</a>)</em>. The second study was relatively small (78 cases) <em>(<a id="xref-ref-42-3" class="xref-bibr" href="#ref-42">42</a>)</em>. One small case–control study (n = 37), which was not included in the meta-analysis because the effect estimate was reported
as an odds ratio per month of breast-feeding <em>(<a id="xref-ref-44-1" class="xref-bibr" href="#ref-44">44</a>)</em>, suggested that breast-feeding was associated with an increased risk of testicular cancer: the odds ratio per month of breast-feeding
was 1.05 (95% CI = 0.99 to 1.11; <em>P</em> = .1).
</p>
<h3>Sensitivity Analysis</h3>
<p id="p-56">Sensitivity analysis was restricted to the breast cancer meta-analyses because so few reports investigated all cancers or
testicular cancers as outcomes. This analysis revealed that effect estimates differed according to study design, although
the difference was borderline statistically significant (<em>P</em> for difference in effect estimates = .08). In the eight case-control studies <em>(<a id="xref-ref-28-12" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-17" class="xref-bibr" href="#ref-34">34</a>–<a id="xref-ref-37-9" class="xref-bibr" href="#ref-37">37</a>,<a id="xref-ref-38-9" class="xref-bibr" href="#ref-38">39</a>–<a id="xref-ref-40-5" class="xref-bibr" href="#ref-40">41</a>)</em>, the pooled relative risk for the association of breast-feeding with breast cancer was 0.90 (95% CI = 0.82 to 0.99; I<sup>2</sup> = 4%). In the three cohort/nested case–control studies <em>(<a id="xref-ref-27-14" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-38-10" class="xref-bibr" href="#ref-38">38</a>)</em> (and Boyd Orr), the pooled relative risk was 1.07 (95% CI = 0.94 to 1.23; I<sup>2</sup> = 0%). There was no evidence that effects estimates differed by whether breast-feeding history was based on retrospective
recall; by whether the study controlled for one or all of socioeconomic, family history, or reproductive factors; by whether
the study was population or hospital based; by the region in which the study was conducted; by the year of birth of the participants;
or by whether or not the prevalence of any breast-feeding was at least 70% (all <em>P</em>>.3). Seven studies <em>(<a id="xref-ref-27-15" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-28-13" class="xref-bibr" href="#ref-28">28</a>,<a id="xref-ref-34-18" class="xref-bibr" href="#ref-34">34</a>,<a id="xref-ref-36-8" class="xref-bibr" href="#ref-36">36</a>–<a id="xref-ref-38-11" class="xref-bibr" href="#ref-38">38</a>)</em> (and Boyd Orr) presented both crude and adjusted estimates; the pooled relative risk for the association between breast-feeding
and breast cancer using crude estimates (RR = 0.94, 95% CI = 0.80 to 1.09) was the same as that for the association using
adjusted estimates (RR = 0.93, 95% CI = 0.82 to 1.06).
</p>
</div>
<div class="section" id="sec-13">
<div class="section-nav"><a href="#sec-6" title="RESULTS" class="prev-section-link"><span>Previous Section</span></a><a href="#sec-14" title="Next Section" class="next-section-link"><span>Next Section</span></a></div>
<h2>D<span class="sc">ISCUSSION</span></h2>
<p id="p-57">In this article, we describe two sets of analyses to examine associations between breast-feeding in infancy and subsequent
adult cancer. First, our analysis of the Boyd Orr cohort provided no evidence that ever having been breast-fed or duration
of breast-feeding is associated with incidence of all cancers or of prostate cancer, gastric cancer, or colorectal cancer.
There was only weak, and not statistically significant, evidence from Boyd Orr in support of the hypothesis of a positive
association between ever having been breast-fed in infancy and risk of breast cancer in adulthood.
</p>
<p id="p-58">Second, in a meta-analysis of published studies plus Boyd Orr (involving 11 564 breast cancer cases in total), ever having
been breast-fed in infancy and the duration of breast-feeding were not associated with breast cancer risk (regardless of menopausal
status). The meta-analyses showed that ever having been breast-fed was associated with a reduced risk of premenopausal breast
cancer (RR = 0.88, 95% CI = 0.79 to 0.98) but was not associated with an alteration in the risk of postmenopausal breast cancer
(RR = 1.00; 95% CI = 0.86 to 1.16). However, the confidence intervals for the two estimates overlap, and the differences in
the risks of pre- and postmenopausal breast cancer could have arisen by chance. There was no statistical evidence of differences
in associations of breast-feeding in infancy with breast cancer by menopausal status in the only study that formally tested
for this interaction (<em>P</em> = .23) <em>(<a id="xref-ref-34-19" class="xref-bibr" href="#ref-34">34</a>)</em>.
</p>
<p id="p-59">The analyses we conducted thus do not support the a priori hypothesis that ever having been breast-fed is linked with an increased
risk of breast cancer. Neither was breast-feeding associated with a greater increase in breast cancer risk among women with
a family history of breast cancer than among women with sporadic breast cancer, arguing against a transmissible agent in breast
milk that increases breast cancer risk. Although breast-feeding was associated with a reduced risk of testicular cancer (RR
= 0.82, 95% CI = 0.62 to 1.10), this reduction was not statistically significant. Moreover, the RR was derived from only two
studies with retrospective ascertainment of exposure and multiple hypothesis testing. Hence, recall bias and chance cannot
be excluded.
</p>
<p id="p-60">Both analyses in this article are subject to a number of limitations. In the Boyd Orr cohort, the confidence limits around
the estimates for site-specific cancers were wide, indicating limited precision and low power to detect associations or interactions.
The data did, however, add information to the breast cancer meta-analysis. We found no published studies relating having been
breast-fed with prostate, colorectal, or gastric cancer; given the imprecise estimates from Boyd Orr and the absence of published
data, definitive conclusions cannot be drawn for these cancers. Another limitation of the Boyd Orr study is the possibility
that associations were confounded by adult risk factors such as smoking, body weight, and reproductive influences. However,
the lack of an association between breast-feeding and smoking-related cancers suggests that breast-feeding in infancy was
not associated with smoking in adult life.
</p>
<p id="p-61">A limitation of the meta-analyses is that they were based on results from a group of studies of heterogeneous design and conduct.
However, there was generally little variation in effect estimates between studies. Moderate between-study heterogeneity was
found for the association of breast-feeding with all breast cancers (I<sup>2</sup> = 31%). Although only 11 studies were involved in the breast cancer analysis, a relatively large number of breast cancers
were included per study (total = 11 564; median = 508; interquartile range = 235–1192), allowing us to investigate this heterogeneity
using meta-regression (discussed below).
</p>
<p id="p-62">There are a number of possible alternative explanations for the associations observed in the meta-analyses. First, in meta-regression
analysis, we found some evidence (<em>P</em> = .08) that effect estimates from case-control studies (which suggested a 10% reduction in risk of breast cancer associated
with having been breast-fed) were qualitatively different than those from the cohort/nested case–control studies (which suggested
no association). Cohort studies are less likely than case–control studies (where breast-feeding history is obtained retrospectively
following the diagnosis of cancer) to be affected by recall or selection biases, and the findings from these studies may be
more robust. The possibility that retrospective versus prospective exposure ascertainment is a potential source of recall
bias is suggested by the finding that long-term recall of breast-feeding history differs by socioeconomic status <em>(<a id="xref-ref-45-1" class="xref-bibr" href="#ref-45">45</a>)</em>. Indeed, an apparent association between breast-feeding and type 1 diabetes was discounted once studies with prospective
measurement of exposure were conducted <em>(<a id="xref-ref-46-1" class="xref-bibr" href="#ref-46">46</a>)</em>. Moreover, in all case–control studies providing this information, response rates were less than 80% in both case and control
subjects, a potential source of selection bias.
</p>
<p id="p-63">Second, because the mother's choice to breast-feed may be related to other factors influencing the future health of the child,
the possibility of confounding in the studies included in the meta-analysis, particularly by socioeconomic status, needs to
be considered. Approximately half of the reports with breast cancer outcomes controlled for at least one of the following:
socioeconomic status, reproductive history, or a family history of breast cancer. There was no evidence that estimates adjusted
for selected confounding factors differed from crude effect estimates, either within or between studies. Most studies included
no or only limited information on childhood diet or factors during infancy (e.g., age at introduction of solids, birth weight,
and length), and residual or uncontrolled confounding remains a concern. It has recently been recognized that growth patterns
in early life may underlie associations relating early-life factors with breast cancer <em>(<a id="xref-ref-3-3" class="xref-bibr" href="#ref-3">3</a>,<a id="xref-ref-47-1" class="xref-bibr" href="#ref-47">47</a>,<a id="xref-ref-48-2" class="xref-bibr" href="#ref-48">48</a>)</em>. None of the published studies reviewed here investigated the role of childhood growth in the relationship between breast-feeding
and cancer risk, although two studies adjusted for final height <em>(<a id="xref-ref-27-16" class="xref-bibr" href="#ref-27">27</a>,<a id="xref-ref-37-10" class="xref-bibr" href="#ref-37">37</a>)</em>. In the Boyd Orr cohort, adjustment for height and BMI in childhood did not alter the point estimates of the associations
between breast-feeding and breast cancer (data not shown), arguing against childhood growth as an intermediate factor on the
causal pathway or for a confounding effect of factors associated with childhood growth, such as diet.
</p>
<p id="p-64">Third, many of the studies had collected data on a large number of exposures. At the individual study level, chance findings
in the context of multiple hypothesis testing and low prior probability that any one hypothesis is correct are a distinct
possibility <em>(<a id="xref-ref-49-1" class="xref-bibr" href="#ref-49">49</a>)</em>.
</p>
<p id="p-65">Fourth, publication bias may occur when small studies are differentially published only when they report large differences
between feeding groups. Other undiscovered or unpublished studies may have included breast-feeding among a large number of
variables tested for statistically significant associations. However, there was no evidence of small-study bias (i.e., the
tendency for the smaller studies in a meta-analysis to provide larger estimates of the effect size) on the Egger tests <em>(<a id="xref-ref-31-2" class="xref-bibr" href="#ref-31">31</a>)</em>.
</p>
<p id="p-66">Finally, there are several limitations to interpreting the infant feeding exposure data. The participants in these reports
were born between 1874 and 1972, during which time alternatives to breast-feeding were likely to be predominantly unmodified
cow's milk preparations. Formula feeds changed considerably from the late 1970s onward. Thus, the relevance of these analyses
to babies born since then is unclear. Also, historic breast-feeding rates were far higher than current rates, so babies who
were bottle-fed were likely to be a highly selected group, especially in the earlier part of the twentieth century. Nevertheless,
we found no evidence that associations altered in studies with different breast-feeding prevalence rates. Moreover, although
the studies were conducted on babies born during a period of considerable socioeconomic transition, the effect sizes were
consistent over time.
</p>
<p id="p-67">Another limitation to the infant feeding exposure data is that most studies defined breast-feeding as any breast-feeding and
did not specify the timing of breast-feeding initiation. Such information would be important to assess whether colostrum,
which is dense with immunologic factors, is important to cancer outcomes. Most studies also did not address whether breast-feeding
was exclusive, the duration of exclusive breast-feeding, and the quantities of supplementary feeds. Distinguishing between
exclusive and partial breast-feeding, and assessing its duration, would help to assess whether the amount of breast milk exposure
is associated with cancer outcomes.
</p>
<p id="p-68">Confounding in observational epidemiologic studies of the long-term effects of breast-feeding is a major limitation of existing
reports <em>(<a id="xref-ref-50-1" class="xref-bibr" href="#ref-50">50</a>)</em>. Two broad strategies could be used to overcome the problem of confounding. First, the relationship could be explored in
populations in which breast-feeding is not socially patterned <em>(<a id="xref-ref-51-1" class="xref-bibr" href="#ref-51">51</a>)</em>. The Boyd Orr cohort presented in this paper involved subjects born between 1918 and 1939, an era during which social and
educational factors played little part in a mother's decision to breast- or bottle-feed <em>(<a id="xref-ref-52-1" class="xref-bibr" href="#ref-52">52</a>)</em>. Therefore confounding by social, educational, and economic factors is likely to be less of an issue in this cohort than
in more recent cohorts <em>(<a id="xref-ref-53-1" class="xref-bibr" href="#ref-53">53</a>)</em>. Second, the relationship between breast-feeding and markers of later cancer risk [such as the IGF axis <em>(<a id="xref-ref-4-2" class="xref-bibr" href="#ref-4">4</a>,<a id="xref-ref-8-4" class="xref-bibr" href="#ref-8">8</a>,<a id="xref-ref-54-1" class="xref-bibr" href="#ref-54">54</a>)</em>] could be investigated in large, randomized, controlled trials of measures to promote breast-feeding. For example, long-term
follow-up of the 17 000 children in the PROBIT trial <em>(<a id="xref-ref-55-1" class="xref-bibr" href="#ref-55">55</a>)</em> would provide an experimental setting in which to rigorously test the association between breast-feeding and markers of cancer
risk. Given our largely null findings, the beneficial influence of breast-feeding on infant and child health <em>(<a id="xref-ref-56-1" class="xref-bibr" href="#ref-56">56</a>,<a id="xref-ref-57-1" class="xref-bibr" href="#ref-57">57</a>)</em> and cognitive development <em>(<a id="xref-ref-58-1" class="xref-bibr" href="#ref-58">58</a>)</em> support its promotion as the infant feeding method of choice.
</p>
<p id="p-69">In summary, we found that breast-feeding has little or no association with development of adult cancer. There are limitations
in the current epidemiologic evidence relating having been breast-fed with cancer, and the generalizability of the results
to modern cohorts is unclear. Better measurement of infant feeding is required if future studies are to improve understanding
of the association between breast-feeding and cancer.
</p>
</div>
<div class="section" id="sec-14">
<div class="section-nav"><a href="#sec-13" title="DISCUSSION" class="prev-section-link"><span>Previous Section</span></a><a href="#fn-group-1" title="Footnotes" class="next-section-link"><span>Next Section</span></a></div>
<p id="p-70">
</p>
<div class="table pos-float" id="T4">
<div class="table-inline">
<div class="callout"><span>View this table:</span><ul class="callout-links">
<li><a href="1446/T4.expansion.html">In this window</a></li>
<li><a class="in-nw" href="1446/T4.expansion.html">In a new window</a></li>
</ul>
</div>
</div>
<div class="table-caption"><span class="table-label"><strong>Appendix.</strong> </span>
<p id="p-71" class="first-child">Search terms used in systematic review</p>
<div class="sb-div caption-clear"></div>
</div>
</div>
</div>
<div class="section fn-group" id="fn-group-1">
<div class="section-nav"><a href="#sec-14" title="Previous Section" 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>Footnotes</h2>
<ul>
<li class="fn" id="fn-19">
<p id="p-72">G.D.S. and Stephen Frankel established the adult follow-up phase of the Carnegie Survey of Diet and Health in Pre-War Britain
using original records loaned by the Rowett Research Institute. R.M.M. and N.M. currently maintain the Boyd Orr cohort database.
R.M.M., G.D.S., D.G., and C.G.O. developed the hypothesis. N.M. analyzed the Boyd Orr dataset. R.M.M. undertook the literature
search, identified eligible papers, undertook data extraction, and did the analysis for the systematic review. R.M.M. wrote
the first draft of the paper and coordinated completion. All authors contributed to, and approved, the final version.
</p>
</li>
<li class="fn" id="fn-20">
<p id="p-73">C.G.O. is supported by the British Heart Foundation (grant no. PG/04/072). N.M. was supported by a World Cancer Research Fund
grant (grant no. 2001/31).
</p>
</li>
<li class="fn" id="fn-21">
<p id="p-74">We thank Peter Morgan, director of the Rowett Research Institute, for the use of the archived material on the original Boyd
Orr cohort, and in particular, Walter Duncan, honorary archivist to the Rowett. We also thank the staff at the National Health
Service Central Register at Southport and Edinburgh. We are grateful to Sara Bright for data entry, Mark Taylor for entering
breast-feeding data, and John Pemberton for information concerning the conduct of the original survey. We also acknowledge
all the research workers who participated in the original survey in 1937–1939. We thank Susie Potts and Jan Hill for secretarial
support.
</p>
</li>
</ul>
</div>
<ul>
<li class="fn" id="copyright-statement-1">© The Author 2005. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: journals.permissions@oupjournals.org.</li>
</ul>
<div class="section ref-list" id="ref-list-1">
<div class="section-nav"><a href="#fn-group-1" title="Footnotes" 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-other" id="cit-97.19.1446.1">
<div class="cit-metadata"><cite>Schwartzbaum JA, George SL, Pratt CB, Davis B. An exploratory study of environmental and medical factors potentially related
to childhood cancer. <span class="cit-source">Med Ped Oncol</span> <span class="cit-pub-date">1991</span>;<span class="cit-vol">19</span>:<span class="cit-fpage">115</span>–21.</cite></div>
<div class="cit-extra"><a href="/external-ref?access_num=1849220&link_type=MED"
class="cit-ref-sprinkles cit-ref-sprinkles-medline">Medline</a><a href="/external-ref?access_num=A1991FE88500008&link_type=ISI"
class="cit-ref-sprinkles cit-ref-sprinkles-webofscience">Web of Science</a></div>
</div>
</li>
<li><span class="ref-label">(2)</span><a class="rev-xref-ref" href="#xref-ref-2-1" title="View reference (2) in text"
id="ref-2">↵</a>
<div class="cit ref-cit ref-other" id="cit-97.19.1446.2">
<div class="cit-metadata"><cite>Potischman N, Troisi R, Vatten L. A lifecourse approach to cancer epidemiology. In: Kuh D, Ben Shlomo Y, editors. A life course
approach to chronic disease epidemiology. 2nd ed. Oxford (England): Oxford University Press; <span class="cit-pub-date">2004</span>. p. 260–80.</cite></div>
<div class="cit-extra"></div>
</div>
</li>
<li><span class="ref-label">(3)</span><a class="rev-xref-ref" href="#xref-ref-3-1" title="View reference (3) in text"
id="ref-3">↵</a>
<div class="cit ref-cit ref-other" id="cit-97.19.1446.3"
data-doi="10.1093/oxfordjournals.epirev.a000809">
<div class="cit-metadata"><cite>Gunnell D, Okasha M, Davey Smith G, Oliver SE, Sandhu J, Holly JMP. Height, leg length, and cancer risk: a systematic review.
<span class="cit-source">Epidemiol Rev</span> <span class="cit-pub-date">2001</span>;<span class="cit-vol">23</span>:<span class="cit-fpage">313</span>–41.</cite></div>
<div class="cit-extra"><a href="/cgi/ijlink?linkType=PDF&journalCode=epirev&resid=23/2/313"
class="cit-ref-sprinkles cit-ref-sprinkles-ijlinks"><span class="cit-reflinks-full-text"><span class="free-full-text">FREE </span>Full Text</span></a></div>
</div>
</li>
<li><span class="ref-label">(4)</span><a class="rev-xref-ref" href="#xref-ref-4-1" title="View reference (4) in text"
id="ref-4">↵</a>
<div class="cit ref-cit ref-other" id="cit-97.19.1446.4"
data-doi="10.1136/bmj.321.7265.847">
<div class="cit-metadata"><cite>Davey Smith G, Gunnell D, Holly J. Cancer and insulin-like growth factor-I. <span class="cit-source">BMJ</span> <span class="cit-pub-date">2000</span>;<span class="cit-vol">321</span>(7265):<span class="cit-fpage">847</span>–8.</cite></div>
<div class="cit-extra"><a href="/cgi/ijlink?linkType=FULL&journalCode=bmj&resid=321/7265/847"
class="cit-ref-sprinkles cit-ref-sprinkles-ijlinks"><span class="cit-reflinks-full-text"><span class="free-full-text">FREE </span>Full Text</span></a></div>
</div>
</li>
<li><span class="ref-label">(5)</span><a class="rev-xref-ref" href="#xref-ref-5-1" title="View reference (5) in text"
id="ref-5">↵</a>
<div class="cit ref-cit ref-other" id="cit-97.19.1446.5"
data-doi="10.1016/S0140-6736(04)16044-3">
<div class="cit-metadata"><cite>Renehan AG, Zwahlen M, Minder PC, O'Dwyer ST, Shalet PS, Egger PM. Insulin-like growth factor (IGF)-I, IGF binding protein-3,
and cancer risk: systematic review and meta-regression analysis. <span class="cit-source">Lancet</span> <span class="cit-pub-date">2004</span>;<span class="cit-vol">363</span>(9418):<span class="cit-fpage">1346</span>–53.</cite></div>
<div class="cit-extra"><a href="/external-ref?access_num=10.1016/S0140-6736(04)16044-3&link_type=DOI"
class="cit-ref-sprinkles cit-ref-sprinkles-webofscience">CrossRef</a><a href="/external-ref?access_num=15110491&link_type=MED"
class="cit-ref-sprinkles cit-ref-sprinkles-medline">Medline</a><a href="/external-ref?access_num=000220996300008&link_type=ISI"
class="cit-ref-sprinkles cit-ref-sprinkles-webofscience">Web of Science</a></div>
</div>
</li>
<li><span class="ref-label">(6)</span><a class="rev-xref-ref" href="#xref-ref-6-1" title="View reference (6) in text"
id="ref-6">↵</a>
<div class="cit ref-cit ref-other" id="cit-97.19.1446.6"
data-doi="10.1136/fn.87.3.F193">
<div class="cit-metadata"><cite>Martin RM, Davey Smith G, Mangtani P, Frankel S, Gunnell D. Association between breast feeding and growth: the Boyd-Orr cohort
study. <span class="cit-source">Arch Dis Child Fetal Neonatal Ed</span> <span class="cit-pub-date">2002</span>;<span class="cit-vol">87</span>:<span class="cit-fpage">F193</span>–F201.</cite></div>
<div class="cit-extra"><a href="/cgi/ijlink?linkType=ABST&journalCode=fetalneonatal&resid=87/3/F193"
class="cit-ref-sprinkles cit-ref-sprinkles-ijlinks"><span class="cit-reflinks-abstract">Abstract</span><span class="cit-sep cit-reflinks-variant-name-sep">/</span><span class="cit-reflinks-full-text"><span class="free-full-text">FREE </span>Full Text</span></a></div>