Appendix 1:The Search Strategy

Appendix 1:The search strategy

Electronic databases

PubMed:

#1 densitometry, X-ray [mh]

#2 densitometry [mh]

#3 bone density [mh]

#4 absorptiometry, photon [mh]

#5 dual x-ray [tw] OR dual energy x-ray [tw]

#6 bone density [tw] OR bone mineral density [tw] OR bmd [tw]

#7 osteoporosis, postmenopausal [mh]

#8 osteoporosis [mh]

#9 osteoporosis [tw]

#10 (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9)

#11 self-assessment [tw] or tool[tw]

#12 ost [tw] OR osta [tw] OR fosta [tw] OR osteorisk[tw]

#13 (#11 OR #12)

#14 (#10 AND #13)

Embase (Webspirs)

#1 explode "dual energy x-ray absorptiometry" / all subheadings

#2 dual NEAR absorptiometry

#3 explode "bone mineral" / all subheadings

#4 bone NEAR density

#5 explode "bone characteristics and functions" / all subheadings

#6 explode "photon absorptiometry" / all subheadings

#7 explode "osteopenia" / all subheadings

#8 osteopenia in ti OR osteopenia in ab

#9 explode "bone demineralization"/ all subheadings

#10 osteoporosis in ti OR osteoporosis in ab

#11 (#1 OR #2 OR #3 OR #4 OR #5 OR #6 OR #7 OR #8 OR #9 OR #10)

#12 ost in ti OR ost in ab

#13 osta in ti OR osta in ab

#14 fosta in ti OR fosta in ab

#15 osteorisk in ti OR osteorisk in ab

#16 self-assessment in ti OR self-assessment in ab

#17 tool in ti OR tool in ab

#18 (#12 OR #13 OR #14 OR #15 OR #16 OR #17)

#19 (#11 AND #18)

Synonyms for OST:

OST:The Osteoporosis Self-assessment Tool

OSTA: The Osteoporosis Self-assessment Tool for Asians

FOSTA:The Female Osteoporosis Self-assessment Tool for Asians

Osteorisk

Conference abstracts

We searched abstract CDs or abstract books from conferences hosted by the following societies since 2000:

- American Society of Bone mineral Metabolism

- International Osteoporosis Foundation

- National Osteoporosis Society

- European Calcified Tissue Society

- International Bone and Mineral Society

- International Society for Clinical Densitometry

as well as abstracts from the European Congresses on Clinical and Economic Aspects of Osteoporosis and Osteoarthritis.

Appendix 2:Calculation of OST and interpretation of the cut-off for a negative test

Illustration of how the calculation of OST influences the interpretation of the cut-off for a negative test. For cut-off values above or equal to zero the interpretation depends on whether OST is calculated with truncation or not. In the meta-analyses we grouped studies regardless of how OST was calculated as long as equivalent cut-offs were used so that the studies classified OST-positives and OST negatives concordantly. When we refer to cut-offs defining a negative OST outcome in the text we assume that OST was calculated with truncation.

Appendix 3:Assessment of methodological quality

The adapted QUADAS checklist used to assess aspects of study quality. All items were scored in the following categories: yes, no, unclear. Brief guidelines for scoring certain item are stated in parentheses. Item 3, 7, 12 are shaded grey as these items were excluded:

Item 3 was considered redundant because of the selection criteria

Item 7 was considered irrelevant, as the score on OST is never a part of the BMD measurement

Item 12 was considered inoperational for this review

We added one item to the checklist (item 15) because Lijmer et al. [52] have demonstrated that inadequate description of study participants can be associated with an overestimation of diagnostic performance. The remainder of the study characteristics found to be associated with a biased estimation of diagnostic performance by Lijmer et al. are included in the QUADAS checklist (item 7, 8, 9, 10, 11) or constituted an exclusion criterion for the present review (separate-samples design).

1: / Was the spectrum of patients clinically representative?
(Unselected patients recruited from the community or primary care) / [ ] Y [ ] N [ ] U
2: / Were selection criteria clearly described?
(Patient selection considered reproducible) / [ ] Y [ ] N
3: / Is DXA likely to correctly classify the target condition?
4: / Was the time period between DXA and OST reasonable?
(DXA less than 6 months after OST, OST less than 3 months after DXA) / [ ] Y [ ] N [ ] U
5: / Did the whole sample or a random selection hereof have DXA? / [ ] Y [ ] N [ ] U
6: / Did patients receive DXA regardless of the score on OST? / [ ] Y [ ] N [ ] U
7: / Did the score on OST form part of the DXA outcome?
8: / Was OST described in sufficient detail to permit its replication?
(Expression, truncation (y/n) and collection of weight described) / [ ] Y [ ] N
9: / Was DXA described in sufficient detail to permit its replication?
(Manufacturer, model, reference values and region described) / [ ] Y [ ] N
10: / Was OST determined without knowledge of the DXA results?
(OST before DXA or blinding explicitly stated) / [ ] Y [ ] N [ ] U
11: / Was DXA performed and analysed without knowledge of OST?
(DXA before OST or blinding explicitly stated) / [ ] Y [ ] N [ ] U
12: / Was the same data available when OST was calculated and when DXA was performed as would be in clinical practice?
13: / Were uninterpretable and intermediate test results reported? / [ ] Y [ ] N
14: / Were withdrawals from the study or missing data accounted for?
(A patient flow diagram could be established) / [ ] Y [ ] N
15: / Were study participants adequately described?
(Age, weight, BMD, risk factors for fracture – 3 out of 4) / [ ] Y [ ] N

Abreviations: DXA, dual energy x-ray absorptiometry; BMD, bone mineral density; Y, yes; N, No; U, unclear

Appendix 4:Structured summary of included studies

Study / Country
and
ethnicity / Setting
and
recruitment procedure / Age
and
weight
[mean(SD)] / DXA-device(s)
(source of reference
material for T-score calculation) / OST formula and
collection of
weight data / T-score
targets / Cut-off
for a negativeOST resulta / 2x2 table according to
T-score target / Compliance with
the QUADAS items
TP
no. / FP
no. / FN
no. / TN
no.
Cook 2005
[35] / Swindon, UK
Whites / Postmenopausal women
referred for DXA due to the presence of ≥ 1 risk factor for osteoporosis were recruited
from DXA scanning clinics / 59.7 y
65.6 kg / Hologic QDR 4500c
(manufacturer) / Tr[(w–a)/5]
by self-report / TTH,LS ≤ –2.5
TTH,LS ≤ –2.0
TTH,LS ≤ –1.0 / > 3
> 3
> 5 / 41
72
135 / 109
78
48 / 4
8
9 / 54
50
16 / Yes: 5,9,15
No: 1,2,8,13,14
Unclear: 4,10,11
Gaugris 2005
[53] / Canada
Whites / The performance of the OST
was assessed using data from the Canadian Multi-Centre Osteoporosis Study / 66.6 (9.4) y
68.6 (13.5) kg / NS
(NS) / (w–a)/5
NS / TFN ≤ –2.5 / > 1 / 828 / 1590 / 111 / 1557 / Yes: 1,2,5,15
No: 8,9,13,14
Unclear: 4,10,11
Siew-Leng 2005
[54] / Singapore
Asians / NS / 55.9 (6) y / Lunar DPX-L
(NS) / NS
NS / TFN ≤ –2.5 / > –1 / 78 / 496 / 21 / 881 / Yes: 5
No: 2,8,9,13,14,15
Unclear: 1,4,10,11
Lu 2005
[32] / ShengDu
Asians / Postmenopausal women
who were either attending
an outpatient clinic or
hospitalized / [Postmeno-pausal]
NS / NS
(NS) / NS
NS / TLS ≤ –2.5
TLS ≤ –2.0 / NS / 52
77 / 56
31 / 36
57 / 74
53 / Yes: 4,5
No: 1,2,8,9,13,14,15
Unclear: 10,11
Sen 2005 (a)
[Development cohort]
[15] / Argentina,
Venzuela, Brazil, Costa Rica, Mexico, Peru
Whites 44%
Mestizos 56% / Records of the most recent 300 postmenopausal women aged
> 49 y who had femoral neck DXA at osteoporosis clinics in
six Latin American countries were eligible / 64.6 (8.4) y
63.0 (11.4) kg / Hologic, Lunar
Norland
(manufacturer) / (w–a)/5
by weighing / TFN ≤ –2.5 / > 1 / 463 / 577 / 40 / 467 / Yes: 2,4,5,8,11,15
No: 1,9,13,14
Unclear: 10
Sen 2005 (b)
[Validation cohort]
[15] / Brazil
NS / Review of hospital records for women aged > 49 y who had
femoral neck DXA / 69.6 (9.2) y
60.3 (11.3) kg / NS
(manufacturer) / (w–a)/5
NS / TFN ≤ –2.5 / > 1 / 117 / 109 / 8 / 45 / Yes: 2,5,15
No: 8,9,13,14
Unclear: 1,4,10,11
Cadarette 2004
[55] / Toronto, Canada
Whites / Women >44 y:
1) attending two family practice clinics who had DXA according to the records
2) presenting consecutively for DXA at an ambulatory care center / 62,4 (11.2) y
65.6 (54) kg / Hologic, Lunar,
Norland
(NS) / Tr[(w–a)/5]
by weighing / TFN ≤ –2.5
TLS ≤ –2.5
TFN,LS ≤ –2.5
TFN ≤ –2.0
TLS ≤ –2.0
TFN,LS ≤ –2.0 / > 1 / 64
68
101
136
112
180 / 360
343
325
288
299
246 / 3
2
5
8
11
18 / 214
211
213
209
202
200 / Yes: 2,4,5,8,11,15
No: 1,9,13,14
Unclear: 10
Study / Country
and
ethnicity / Setting
and
recruitment procedure / Age
and
weight
[mean(SD)] / DXA-device(s)
(source of reference
material for T-score calculation) / OST formula and
collection of
weight data / T-score target / Cut-off
for a negativeOST resulta / 2x2 table according to
T-score target / Compliance with
the QUADAS items
TP
no. / FP
no. / FN
no. / TN
no.
Choi 2004
[56] / Seoul,
Korea
Asians / Women aged > 44 y who had undergone testing with DXA at both the femoral neck and the lumbar spine / > 45 y
NS / NS
(NS) / (w–a)/5
NS / TFN ≤ –2.5
TLS ≤ –2.5
TFN,LS ≤ –2.5 / > –1 / 75
104
134 / 246
217
187 / 1
29
30 / 679
651
650 / Yes: 5
No: 2,8,9,13,14,15
Unclear: 1,4,10,11
Gambacciani 2004
[57] / Italy
Whites / NS / 58 (7) y
NS / Lunar DPX
(manufacturer) / Tr(w–a)/5
NS / TTH ≤ –2.5
TLS ≤ –2.5
TTH,LS ≤ –2.5
TTH ≤ –2.0
TLS ≤ –2.0
TTH,LS ≤ –2.0 / > 1 / 10
46
47
22
76
77 / 180
146
143
168
116
113 / 1
9
7
3
16
15 / 82
75
75
80
68
67 / Yes: 5,9
No: 2,8,13,14,15
Unclear: 1,4,10,11
Geater 2004
[58] / Thailand
Asians / BMD measurements made at Songklanarind Hospital from Nov-2000 to Apr-2002 was reviewed and data for women whose indication for DXA was postmenopausal status was retrieved / 60.5 (9.7) y
57.2 (10.3) kg / Lunar
(manufacturer) / Tr(w/5) – Tr(a/5)
NS / TFN ≤ –2.5
TLS ≤ –2.5 / > –1 / 43
95 / 134
82 / 3
24 / 208
187 / Yes: 2,5,11,15
No: 1,8,9,13,14
Unclear 4,10
Mossman 2004
[59] / Oregon, USA
NS / Postmenopausal women referred to the Oregon Osteoporosis center for DXA testing / [Postmeno-pausal]
NS / NS
(NS) / Tr[(w–a)/5]
NS / TFN ≤ –2.5
TLS ≤ –2.5
TFN,LS ≤ –2.5 / > 1 / 459
734
928 / 4210
3986
3706 / 44
140
168 / 3267
3208
3124 / Yes: 2,5
No: 1,8,9,13,14,15
Unclear: 4,10,11
Nguyen 2004
[31] / Dubbo,
Australia
Whites / All women > 60 y living in Dubbo were invited to participate in the Dubbo Epidemiology Study / 70.5 (7.5) y
65.0 (12.9) kg / Lunar DPX-L
(NS) / Tr[(w–a)/5]
by weighing / TFN ≤ –2.5
TFN,LS ≤ –2.5 / ≥ –1 / 50
100 / 218
58 / 73
70 / 69
182 / Yes: 4,5,9,11,15
No: 2,8,13,14
Unclear: 1,10
Pongchaiyakul 2004
[36] / Khon Kaen
Province,
Thailand
Asians / Women ≥ 45 y in a specific geographic region were selected randomly for participation / 60.2 (9.6) y
56.3 (10.0) kg / Lunar DPX-IQ
(Local) / Tr[(w–a)/5]
by weighing / TFN ≤–2.5
TLS ≤ –2.5
TFN,LS ≤ –2.5 / > –1 / 34
82
86 / 128
80
76 / 1
19
19 / 154
136
136 / Yes: 1,2,4,5,9,13,14,15
No: 8
Unclear: 10,11
Poriau 2004
[60] / Belgium
NS / NS / 68 y
NS / Lunar Prodigy,
Lunar DPX-NT
(manufacturer) / (w–a)/5
NS / TLS ≤ –2.5 / > 1 / 292 / 445 / 40 / 310 / Yes: 5
No: 2,8,9,13,14,15
Unclear: 1,4,10,11
Study / Country
and
ethnicity / Setting
and
recruitment procedure / Age
and
weight
[mean(SD)] / DXA-device(s)
(source of reference
material for T-score calculation) / OST formula and
collection of
weight data / T-score target / Cut-off
for a negativeOST resulta / 2x2 table according to
T-score target / Compliance with
the QUADAS items
TP
no. / FP
no. / FN
no. / TN
no.
Richy 2004 (a)
[Clinic study]
[43] / Liège,
Belgium
Whites / Retrospective collection of medical data for women > 44 y who had DXA at an out-patient osteoporosis Center. / 61.5: (8.8) y
65.1 (11.9) kg / Hologic QDR1000,
QDR2000, QDR4500
(Local) / (w–a)/5
NS / TFN ≤ –2.5
TTH ≤ –2.5
TLS ≤ –2.5
TFN,TH,LS ≤ –2.5
TFN ≤ –2.0
TTH ≤ –2.0
TLS ≤ –2.0
TFN,TH,LS ≤ –2.0 / ≥ 2 / 698
372
833
1131
1045
644
1239
1565 / 2064
2410
1924
1632
1716
2149
1529
1191 / 61
12
147
184
129
34
291
344 / 1212
1242
1130
1088
1144
1209
977
936 / Yes: 2,5,9,11,15
No: 1,8,13,14
Unclear: 4,10
Richy 2004 (b)
[ORACLE study]
[29] / Begium
Whites / Healthy postmenopausal women >44 y attending a public screening for osteoporosis / 63.4 (6.6) y
68.8 (13) kg / Hologic QDR 4500
(manufacturer) / NS
NS / TFN ≤ –2.5
TFN ≤ –1.0 / NS
NS / NE
NE / NE
NE / NE
NE / NE
NE / Yes: 1,4,5,9,15
No: 8,13,14
Unclear: 1,10,11
Rud 2004
[44] / Denmark
Whites / Healthy, peri- and recent postmenopausal women were recruited randomly from public registers for the Danish Osteo-porosis Prevention Study / 50.5
(48.4-52.6)† y
65.9
(59.8-74.0)† kg / Hologic QDR1000
QDR2000
(FN, TH: NHANES III
LS: manufacturer) / Tr[(w–a)/5]
by weighing / TFN ≤ –2.5
TTH ≤ –2.5
TLS ≤ –2.5
TFN,TH,LS ≤ –2.5
TFN ≤ –2.0
TTH ≤ –2.0
TLS ≤ –2.0
TFN,TH,LS ≤ –2.0 / > 1 / 12
5
43
49
63
29
111
136 / 568
575
541
537
517
551
473
450 / 1
1
42
43
28
16
107
126 / 1416
1416
1381
1380
1389
1401
1316
1297 / Yes: 1,4,5,9,11,14,
15
No: 2,8,13
Unclear: 10
Taguchi 2004
[61] / Hiroshima, Japan
Asians / Women visiting the an osteoporosis clinic for BMD measurement
were recruited / 56.8 (7.7) y
51.5 (7.3) kg / Lunar DPX-alpha
(NS) / Tr(w/5) – Tr(a/5)
by weighing / TLS ≤ –2.5 / > –1 / 62 / 100 / 16 / 138 / Yes: 2,4,5,8,15
No: 9,13,14
Unclear: 1,10,11
Wallace 2004
[33] / Texas, USA
African-Americans / Subjects were recruited from African-American churches by a respected nurse / 59.4 (12.5) y
79.8 (20.7) kg / Hologic QDR 2000
(NHANES III) / (w–a)/5
by weighing / TFN ≤ –2.5
TFN ≤ –2.0 / NS
> –1 / 16
NE / 14
NE / 10
NE / 134
NE / Yes: 1,2,4,5,8,9,15
No: 13,14
Unclear: 10,11
Chan 2003
[62] / Malaysia
Asians / Postmenopausal women were recruited at the University of Malaya Medical Centre / > 50 y
NS / Lunar
(NS) / Tr[(w–a)/5]
NS / TFN ≤ –2.5 / > –1 / 56 / 219 / 10 / 220 / Yes: 5
No: 2,8,9,13,14,15
Unclear: 1,4,10,11
Study / Country
and
ethnicity / Setting
and
recruitment procedure / Age
and
weight
[mean(SD)] / DXA-device(s)
(source of reference
material for T-score calculation) / OST formula and
collection of
weight data / T-score target / Cut-off
for a negativeOST resulta / 2x2 table according to
T-score target / Compliance with
the QUADAS items
TP
no. / FP
no. / FN
no. / TN
no.
Kung 2003
[63] / Hong Kong,
China
Asians / Southern Chinese women recruited from the community when they passed public road shows, health fairs or health talks on osteoporosis / 62 (8) y
55 (8) kg / Hologic QDR2000+
(Local) / Tr(w×2) – Tr(a×2)
by weighing / TFN ≤ –2.5
TLS ≤ –2.5
TFN,LS ≤ –2.5 / > –1 / 136
175
216 / 261
222
181 / 19
46
56 / 306
279
269 / Yes: 2,4,5,8,9,11,15
No: 13,14
Unclear: 1,10
Lesnyak 2003 (a)
[Irkutsk study]
[64] / Irkutsk, Russia
Whites / Retrospective clinical data on postmenopausal women
aged ≥ 50 y / 62.3 (7.0) y
71.9 (12.7) kg / Lunar DPX-IQ
(NS) / Tr[(w–a)/5]
by weighing / TFN or TLS ≤ –2.5 / > 3 / 140 / 233 / 21 / 108 / Yes: 5,15
No: 2,8,9,13,14
Unclear: 1,4,10,11
Lesnyak 2003
(b)
[Yaroslavl study]
[64] / Yaroslavl, Russia
Whites / Retrospective clinical data on postmenopausal women
aged ≥ 50 y / 62.0: (8.2) y
73,2 (13.0) kg / Hologic QDR 4500C
Elite
(NS) / Tr[(w–a)/5]
by weighing / TFN,LS ≤ –2.5 / > 3 / 162 / 236 / 24 / 194 / Yes: 5,15
No: 2,8,9,13,14
Unclear: 1,4,10,11
Mellström 2003
[65] / Gothenberg, Sweden
Whites / Data was collected retro-spectively for women aged
> 54 who attended an academic osteoporosis clinic and had femoral neck DXA / 68,4 (8.3) y
65,2 (11,4) kg / NS
(NS) / (w–a)/5
NS / TFN ≤ –2.5 / > 1 / 1095 / 1859 / 178 / 1584 / Yes: 5
No: 1,2,8,9,13,14,15
Unclear: 4,10,11
Ongphiphadha-nakul 2003
[37] / Bangkok, Thailand
Asians / Women > 54 y recruited for BMD screening by advertising / 64.6 (5.7) y
56.9 (9.0) kg / Lunar Expert XL
(Local) / Tr[(w–a)/5]
NS / TFN ≤ –2.5 / > –1 / 23 / 125 / 10 / 200 / Yes: 5,9,15
No: 2,8,13,14
Unclear: 1,4,10,11
Park 2003
[66] / Korea
Asians / Menopause clinic. Existing data from medical records / 59.1 (7.7) y
58.0 (8.0) kg / Lunar DPQ-IQ
(Local) / Tr[(w–a)/5]
by self-report / TFN ≤ –2.5
TFN ≤ –2.0 / > –1 / 103
175 / 321
249 / 16
45 / 661
632 / Yes: 1,2,8,9,11,15
No: 13, 14
Unclear: 4, 5, 10
Cho 2002
[30] / Korea
Asians / For women ≥ 40 y information about risk factors and the
results of bone densitometry
was collected from 23 centers across Korea / [Postmeno-pausal]
NS / NS
(NS) / NS
NS / T-score ≤ –2.5
(Region NS) / NS / 54 / 298 / 12 / 473 / Yes: 2,5,11
No: 8,9,13,14,15
Unclear: 1,4,10
Geusens 2002 (a) [42]
[Fracture Intervention
Trial - Asians] / USA
Asians / Asian women screened with
DXA for participation in
the Fracture Intervention Trial / 67(7) y
NS / Hologic QDR 2000
(Local) / Tr[(w–a)/5]
by weighing / TFN ≤ –2.5 / > –1 / 21 / 74 / 1 / 44 / Yes: 2,4,5,8,9,11
No: 13,14,15
Unclear: 1,10
Geusens 2002 (b) [42]
[Fracture Intervention
Trial - Whites] / USA
Whites / Women screened with DXA
for participation in The
Fracture Intervention Trial / 67 (6.5) y
NS / Hologic QDR2000
(NHANES III) / Tr[(w–a)/5]
by weighing / TFN ≤ –2.5
TFN ≤ –2.0 / > 1 / 4782
9256 / 13251
8777 / 251
936 / 5549
4864 / Yes: 4,5,8,9,11
No: 2,13,14,15
Unclear: 1,10
Study / Country
and
ethnicity / Setting
and
recruitment procedure / Age
and
weight
[mean(SD)] / DXA-device(s)
(source of reference
material for T-score calculation) / OST formula and
collection of
weight data / T-score target / Cut-off
for a negativeOST resulta / 2x2 table according to
T-score target / Compliance with
the QUADAS items
TP
no. / FP
no. / FN
no. / TN
no.
Geusens 2002
(c) [42]
[Netherlands Clinic sample] / Holland
Whites / All healthy and 25% of the remaining postmenopausal women aged 50-80 y
registered with 23 GPs were invited to participate / 63 (7) y
71 (12) kg / Hologic QDR1000
(NS) / Tr[(w–a)/5]
by weighing / TLS ≤ –2.5
TLS ≤ –2.0 / > 2 / 599
929 / 1768
1438 / 134
280 / 1703
1557 / Yes: 1,4,5,8,9,1,
13, 15
No: 2,14
Unclear: 10
Geusens 2002
(d) [42]
[Rotterdam Study] / Holland
Whites / All inhabitants in the district of Ommoord, Rotterdam aged
>54y were invited to partici-
pate in the Rotterdam Study. / >54 y
NS / Lunar DPX-L
(NHANES III) / Tr[(w–a)/5]
by weighing / TFN ≤ –2.5
TFN ≤ –2.0 / > 1 / 592
1112 / 1651
1131 / 47
170 / 1050
927 / Yes: 5,8,9,11
No: 2,13,14,15
Unclear: 1,4,10
Geusens 2002
(e) [42]
[SCORE
developement cohort] / USA
Whites / Peri/ postmenopausal women attending 106 investigators specializing in family medicine, gynecology, or internal
medicine for routine check-up. / 61.3 (9.6) y
70.5 (15.7) kg / Hologic, Lunar
Norland
(FN: NHANESIII
LS: NS) / Tr[(w–a)/5]
by self-report / TFN ≤ –2.5
TLS ≤ –2.5
TFN ≤ –2.0
TLS ≤ –2.0 / > 1
> 2
> 1
> 2 / 133
151
269
215 / 459
513
323
449 / 19
25
53
44 / 491
358
457
339 / Yes: 1,5,8,9,11,15
No: 2,13,14
Unclear: 4,10
Koh 2002
[67] / Singapore
Asians / NS / 59 (7.4) y
58 (8.7) kg / Hologic QDR4500
(NS) / Tr(w×2) – Tr(a×2)
NS / TFN ≤ –2.5 / > –1 / 18 / 114 / 2 / 160 / Yes: 2,5,15
No: 8,9,13,14
Unclear: 1,4,10,11
Sen 2002
[34] / New Zealand
NS / Retrospective data collection / 73.5 (4.3) y
66.6 (11.4) kg / NS
(NS) / (w–a)/5
NS / TFN ≤ –2.5 / > –1 / 134 / 675 / 24 / 518 / Yes: 5,15
No: 2,8,9,13,14
Unclear: 1,4,10,11
Ben Sedrine 2001
[68] / China
Asians / NS / [Postmeno-pausal]
NS / NS
(NS) / NS
NS / TFN ≤ –2.5 / > –1 / 81 / 405 / 7 / 480 / Yes: 5
No: 2,8,9,13,14,15
Unclear: 1,4,10,11
Koh 2001,
Fujiwara 2001
[validation cohort]
[12;41] / Hiroshima,
Japan
Asians / Women follow-up during 1994-5 in the Adult Health Study on the effects of radiation exposure. / 65 (47-91) y
NS / Hologic QDR 2000
(Local) / Tr[(w–a)/5]
by weighing / TFN ≤–2.5
TLS ≤ –2.5 / > –1
≥ –1 / 179
255 / 664
475 / 4
38 / 276
359 / Yes: 4,5,8,9,11
No: 2,13,14,15
Unclear: 1,10

a: Cut-offs on OST are only listed once if the same cut-off is used for all T-score targets, otherwise cut-offs are listed according to T-score targets.

Abbreviations:DXA, dual energy x-ray absorptiometry; BMD, bone mineral density; NE, not extractable; NS, not stated; NHANES III, Third National Health and Nutrition Examination Survey

a, age; w, weight; Tr, truncation to an integer, FN, femoral neck; TH, total hip; LS, lumbar spine; TP, true positives; FP, False positives; FN, false negative; TN, true negatives

Appendix 5:Analysis of heterogeneity

Outcome of the meta-regression analysis of heterogeneity in primary study estimates of negative likelihood ratios with respect to femoral neck T-score ≤ –2.5 against predefined covariates. Estimates of negative likelihood ratios, coefficients and standard errors are on natural log-scale.

Covariate / Whites / Asians
Coefficient (SE) / p-value / Coefficient (SE) / p-value
Study characteristics
Publication type (article vs. abstract) / – 0.06 (0.19) / 0.86 / – 0.08 (0.46) / 0.84
Recruitment of women referred for DXA
(no vs. yes, unclear) / – 0.16 (0.11) / 0.42 / 0.03 (0.47) / 0.92
Manufacturer of densitometer
(Hologic vs. Lunar) / a / a / – 0.52 (0.40) / 0.22
Collection of body weight
(self-report vs. weighing, unclear) / 0.27 (0.24) / 0.43 / 0.11 (0.72) / 1.00
Calculation of OST
(Tr[(w-a)/5], (w-a)/5 vs. other) / b / b / – 0.15 (0.46) / 0.80
Involvement of pharmaceutical or
device industry (yes vs. no, unclear) / 0.02 (0.19) / 0.96 / 0.02 (0.46) / 0.95
Prevalence of femoral neck T-score ≤ –2.5
(for each 5% increase in prevalence) / – 0.04 (0.06) / 0.63 / – 0.07 (0.25) / 0.77
QUADAS Items (yes vs. no, unclear)c
1: Representative spectrum / 0.22 (0.23) / 0.58 / – 0.26 (0.63) / 0.71
2: Selection criteria described / 0.16 (0.11) / 0.47 / – 0.27 (0.45) / 0.49
4: Short time between DXA and OST / – 0.21 (0.10) / 0.03 / – 0.52 (0.48) / 0.28
8: OST described in detail / – 0.17 (0.11) / 0.38 / – 0.23 (0.47) / 0.54
9: DXA described in detail / – 0.01 (0.19) / 0.96 / 0.07 (0.46) / 0.88
11: DXA blind to OST / – 0.05 (0.19) / 0.87 / – 0.39 (0.43) / 0.34
13: Uninterpretable results reported / b / b / – 1.33 (1.23) / 0.32
14: Withdrawals accounted for / – 0.57 (0.97) / 0.72 / – 1.33 (1.23) / 0.35
15: Participants adequately described / 0.21 (0.12) / 0.25 / – 0.52 (0.42) / 0.24

a: meta-regression analysis was not feasible because less than seven studies reported on this item

b: meta-regression analysis was not feasible due to uniformity of classification

c: meta-regression analysis was not feasible for item 5, 6, 10 due to uniformity of classification.

Appendix 6:Sensitivity analyses.

The analyses explore the effect of excluding studies

in Whites that do not apply the ‘> 1’ cut-off to define

a negative OST result.

Two summary ROC-curves are estimated using the

Moses method [28] for each of the three targets:

femoral neck T-score ≤ –2.5,

lumbar spine T-score ≤ –2.5,

any region T-score ≤ –2.5.

One curve is estimated for all studies regardless of

cut-off and another is estimated for studies using the

‘> 1’ cut-off.

Bias in the summary estimate of LR– due to exclusion

of studies using cut-offs other than ‘> 1’ is unlikely if

the two summary ROC curves coincide.

The inclination of the straight line in the ROC

diagrams represents the summary estimate of LR–

derived from studies applying the ‘> 1’ cut-off to define

a negative OST result.