Self-monitoring of blood glucose in patients with diabetes who do not use insulin-are guidelines...
-
Upload
independent -
Category
Documents
-
view
3 -
download
0
Transcript of Self-monitoring of blood glucose in patients with diabetes who do not use insulin-are guidelines...
Do Guidelines for the Diagnosis andMonitoring of Diabetes Mellitus Fulfill
the Criteria of Evidence-BasedGuideline Development?
Eva Nagy,1 Joseph Watine,2 Peter S. Bunting,3 Rita Onody,1 Wytze P. Oosterhuis,4 Dunja Rogic,5
Sverre Sandberg,6 Krisztina Boda,7 and Andrea R. Horvath1*
BACKGROUND: Although the methodological quality oftherapeutic guidelines (GLs) has been criticized, little isknown regarding the quality of GLs that make diagnos-tic recommendations. Therefore, we assessed themethodological quality of GLs providing diagnosticrecommendations for managing diabetes mellitus(DM) and explored several reasons for differences inquality across these GLs.
METHODS: After systematic searches of published andelectronic resources dated between 1999 and 2007, 26DM GLs, published in English, were selected andscored for methodological quality using the AGREEInstrument. Subgroup analyses were performed basedon the source, scope, length, origin, and date and typeof publication of GLs. Using a checklist, we collectedlaboratory-specific items within GLs thought to be im-portant for interpretation of test results.
RESULTS: The 26 diagnostic GLs had significant short-comings in methodological quality according to theAGREE criteria. GLs from agencies that had clear pro-cedures for GL development, were longer than 50pages, or were published in electronic databases were ofhigher quality. Diagnostic GLs contained more preana-lytical or analytical information than combined (i.e.,diagnostic and therapeutic) recommendations, but theoverall quality was not significantly different. The qual-ity of GLs did not show much improvement over thetime period investigated.
CONCLUSIONS: The methodological shortcomings of di-agnostic GLs in DM raise questions regarding the va-lidity of recommendations in these documents that
may affect their implementation in practice. Our re-sults suggest the need for standardization of GL termi-nology and for higher-quality, systematically devel-oped recommendations based on explicit guidelinedevelopment and reporting standards in laboratorymedicine.© 2008 American Association for Clinical Chemistry
Clinical practice guidelines are systematically devel-oped statements to assist practitioner and patient deci-sions about appropriate health care for specific clinicalcircumstances (1 ). The methodological quality ofpractice guidelines (GLs)8 has been widely criticized(2 ). As effective treatment requires effective diagnosis,recommendations for the clinical use of tests shouldalso fulfill the criteria of evidence-based guideline de-velopment (3 ). Assuring methodological quality ofGLs requires that the potential biases of GL develop-ment have been addressed adequately and that the rec-ommendations are valid and feasible in practice.Therefore, the aim of our current study was to investi-gate whether GL development teams use appropriateand explicit methods for making diagnostic recom-mendations and whether diagnostic GLs meet basic re-porting standards. For these assessments, we chose lab-oratory diagnosis and monitoring of diabetes mellitus(DM), one of the global health problem areas in whichGLs are most widely used worldwide. In previous workusing the AGREE (Appraisal of Guidelines Researchand Evaluation) instrument (4 ), we showed that themethodological quality of 4 DM GLs, issued by presti-
1 Department of Clinical Chemistry, University of Szeged, Medical Faculty,Szeged, Hungary; 2 Laboratoire de Biologie Polyvalente, Hopital General, Rodez,France; 3 Department of Pathology and Laboratory Medicine, The OttawaHospital, Ottawa, Ontario, Canada; 4 Department of Clinical Chemistry, AtriumMedical Centre, Heerlen, The Netherlands; 5 Institute of Clinical LaboratoryDiagnosis, Zagreb University School of Medicine, Clinical Hospital Center,Zagreb, Croatia; 6 Laboratory of Clinical Biochemistry, Haukeland UniversityHospital, Bergen, Norway; 7 Department of Medical Informatics, University ofSzeged, Medical Faculty, Szeged, Hungary.
* Address correspondence to this author at: Department of Clinical Chemistry,University of Szeged, Medical Faculty, Somogyi Bela ter 1, Szeged, H-6725Hungary. E-mail [email protected].
Received August 6, 2008; accepted August 7, 2008.Previously published online at DOI: 10.1373/clinchem.2008.1090828 Nonstandard abbreviations: GL, guideline; DM, diabetes mellitus; AGREE, Ap-
praisal of Guidelines Research and Evaluation; D, AGREE domain; I, AGREEitem.
Clinical Chemistry 54:111872–1882 (2008)
Evidence-based Laboratory Medicine and Test Utilization
1872
gious authorities between 1999 and 2003, was ratherlow (5 ). The objectives of the current study were toexamine whether similar findings were present in alarger sample of recently published DM GLs, and if so,to identify some of the reasons for such findings. Wewere also interested in differences between GLs that areprimarily diagnostic compared to those that are com-bined with therapeutic recommendations. Owing tothe high number and heterogeneous nature of diagnos-tic questions and recommendations addressed in DMguidelines, and the fact that AGREE is designed to as-sess the methodological quality of GLs only, our studydid not investigate the accuracy of the content ofguidelines.
Materials and Methods
SEARCH STRATEGY
We carried out a systematic literature search to retrievediagnostic GLs in DM. The aim of the search was toobtain a representative sample of GLs, published in En-glish between 1 January 1999 and 31 December 2007,that can be easily accessed and are therefore likely to beread and used in many countries. In PubMed, 1 re-viewer (J. Watine) applied a broad search strategy usingthe Clinical Queries filter “systematic[sb],” which iscapable of retrieving systematic reviews and/or GLs(6 ). This term was combined with the laboratory-specific MeSH terms “Clinical Laboratory Techniques”[MeSH] AND systematic[sb] OR “Laboratory Tech-niques and Procedures”[MeSH] AND systematic[sb](7 ). Another independent reviewer (E. Nagy) searchedin electronic journals using the keywords “guideline”AND “diabetes,” and in dedicated GL databases andwebsites of professional organizations. The databasessearched are shown in Data Supplement 1, whichaccompanies the online version of this article athttp://www.clinchem.org/content/vol54/issue11.
SELECTION OF GUIDELINES ELIGIBLE FOR THE STUDY
Based on the titles and/or abstracts, references werescreened for relevance (E. Nagy, J. Watine). Using thissubset, 2 independent reviewers (E. Nagy, A.R. Hor-vath) applied the following inclusion criteria: the pub-lication fulfilled the definition of GLs (1 ) and dealtwith the use of laboratory tests for the diagnosis ormonitoring of DM, and the GL was publicly availablein a peer-reviewed journal and/or in nationally or in-ternationally endorsed GL databases. If several updatesof the GL were available during the studied time pe-riod, only the latest version was selected. All these cri-teria had to be met for enrollment into the study.
We excluded publications that contained thera-peutic recommendations only, were primarily focusedon technical/analytical/quality control/standardization/
quality management issues, referred to special patientgroups, or offered local protocols on best practice (i.e.,restricted to 1 particular health care setting).
EVALUATION OF THE METHODOLOGICAL QUALITY OF
GUIDELINES
It has been shown that compliance with the AGREEcriteria of most GL development programs is high (8 ),and, therefore, we used AGREE, a standardized, ge-neric, and validated checklist (4, 9, 10 ), along with itsaccompanying Training Manual, for the assessment ofGLs. AGREE arranges 23 criteria, thereafter referred toas AGREE items I1 through I23, into 6 key domains(D1 through D6). Selected GLs were randomly allo-cated to 2 assessment teams with 4 reviewers per team.Reviewers were trained how to use AGREE in a pilotstudy before conducting this larger survey (5 ). Tomake the appraisal process as objective as possible, re-viewers were provided with all supplementary files ref-erenced by each GL and found in the public domain (E.Nagy), including background supporting materials,technical papers, or general GL development manualsissued by the respective GL agency.
Reviewers independently assessed the fulfillmentof the AGREE criteria on a 4-point Likert scale. Dis-agreements in 2 or more scores between appraiserswere resolved by discussion and consensus. An inde-pendent reviewer’s opinion (A.R. Horvath) was re-quired in 1 case only for reaching consensus. Domainscores were expressed in percentages, and a final con-clusion was reached about the acceptability of the GLaccording to the instructions of AGREE. A GL was“strongly recommended” if the majority of itemsscored 3 or 4 and most domain scores (i.e., at least 4 of6) were �60%. A GL was “not recommended” if themajority of items rated 1 or 2 and most of the domainscores (i.e., 4 or more of 6) were �30%. Guidelineswere “recommended with provisos or alterations”when the GL rated high (3 or 4) or low (1 or 2) on asimilar number of items and most domain scores werebetween 30% and 60%. For investigating whether di-agnostic GLs meet additional reporting standards (11 )that are not covered in depth in the AGREE, we as-sessed the presence of the following items: (a) an evi-dence table, (b) a description of the grading system, (c)graded recommendations, and (d) an expiry or reviewdate. Additionally, we assessed whether the GL con-tained data thought to be important for test interpre-tation (3 ), such as (e) prevalence, (f ) diagnostic accu-racy of tests, (g) preanalytical, and (h) analyticalspecifications. All reviewers checked the availability ofthese items, and results were summarized by 1 inde-pendent assessor (E. Nagy).
We created 5 subgroups of GLs based on theirsource, scope, length, and origin and whether they were
Quality of Diagnostic Guidelines in Diabetes
Clinical Chemistry 54:11 (2008) 1873
supplemented with a guideline methods manual. Wealso investigated the quality of guidelines according tothe date and type of publication. In the statistical anal-yses (K. Boda), the mean item and standardized do-main scores of GL subgroups were compared by theKruskal–Wallis test. Pair-wise comparisons were car-ried out using the Mann–Whitney U-test with Bonfer-roni correction. The frequency of reporting laboratoryspecific information in different guideline subgroupswas compared with the Fisher exact test. The level ofsignificance was set at P � 0.01 because of multiplecomparisons. All analyses were performed using SPSSfor Windows, version 13.
Results
Of 2630 references retrieved in a broad search, wefound 497 GLs to be related to laboratory medicine(Fig. 1). After screening for relevance, we subjected 54GLs to the selection criteria described. Fig. 1 shows thereasons for excluding 28 GLs; 26 GLs became eligiblefor critical appraisal (12–37 ). The most importantcharacteristics of GLs are summarized in Table 1. NineGLs originated from the USA, 3 from Canada, 7 fromthe UK, 1 each from Australia, New Zealand, and SouthAfrica, and 4 were international. All but 2 GLs weredeveloped in the last 6 years.
CRITICAL APPRAISAL OF GUIDELINES
Based on the assessment of methodological quality, 22GLs were recommended by reviewers, of which only 11were strongly recommended and the rest “with provi-sos and alterations.” Four GLs had 4 or 5 domains withscores �30%, and reviewers did not recommend theiruse (Table 1).
The domain and item scores of individual GLs areshown in Table 1 and online Data Supplement 2, re-spectively. Table 2 summarizes the mean item scoresand the number and proportion of GLs scoring �3 onthe 4-point Likert scale. Overall, the best-performingdomains were D1, “scope and purpose” (77%; Table1), with a high proportion of GLs scoring above 3 for allitems (Table 2). Although D4, “clarity and presenta-tion,” scored highly (76%; Table 1), I18 within thisdomain performed poorly, as only 10 GLs (38%) weresupported with tools for application (Table 2).
Domains 2 and 3, which explored the process ofGL development, showed lower scores (Table 1). NineGLs (35%) scored �60% in “stakeholder involvement”and 14 (54%) in “rigor of development” domains. Ofindividual items in D2, only a small proportion ofGLs gave information about the composition andaffiliations of the guideline development group, pro-vided some information on patient involvement in thedevelopment process, defined their target users clearly,and pilot tested the GL by target users before publica-tion (Table 2). In D3, there are notable shortcomingsin using systematic methods for searching the evidenceor at least giving some information about literatureretrieval, describing clearly the criteria for selectingthe evidence, indicating the methods used for formu-lating recommendations, and giving information onthe peer review and updating process. The lowestscores were achieved with “applicability” (34%) and“editorial independence” (39%) domains, in whicheach items performed very poorly (Table 2 and onlineData Supplement 2).
QUALITATIVE ANALYSIS OF GUIDELINES
Date of publication. Quality of GLs was also investi-gated according to the date of publication to seewhether any improvement can be observed over time.Only the highest scoring D1 and D4 showed some mar-ginal development in quality over the time scale inves-tigated (Table 1). GLs seem to have become more spe-cific in stating their objectives (I1) and in creating morefocused clinical questions (I2), and the recommenda-tions in GLs have become more easily identifiable (I17)(online Data Supplement 2). However, the poor per-formance in D6 showed further deterioration from
Broad search and selection for laboratory-related guidelines in:–MEDLINE n = 222–electronic guideline databases n = 295 (overlap 20)
n = 497
Excluded for not being relevant tothe laboratory management of DMn = 443
Full articles read for detailed evaluationn = 54
Guidelines selected for appraisaln = 26 (12–37)
Excluded n = 28•Technical or analytical paper •Duplicate publicatio = 1•Inappropriate topi = 3•No recommendation = 4•Newer updates available n = 6•Special patient group n = 2•Local protocol = 6•Parts of 1 guideline n = 3(4-part guideline merged into 1)
Screening of titles/ abstracts for relevance
n = 3
Fig. 1. Searching and selecting guidelines.
1874 Clinical Chemistry 54:11 (2008)
Tabl
e1.
Crit
ical
appr
aisa
lof
diab
etes
mel
litus
guid
elin
esby
the
AG
REE
inst
rum
ent.
Gu
idel
ine
Dat
eo
fis
sue
Sou
rcea
Sco
peb
Len
gth
,p
ages
GL
man
ual
cO
rig
in
Do
mai
nsc
ore
,%
Ove
rall
asse
ssm
ent
D1d
D2
D3
D4
D5
D6
(12
)19
99Bo
thDi
agno
stic
�10
0N
oU
SA89
4069
3517
21Re
com
men
dw
ithal
tera
tion
(13
)20
01Da
taba
seCo
mbi
ned
51–1
00Ye
sU
K56
7574
718
71St
rong
lyre
com
men
d
(14
)20
02Jo
urna
lCo
mbi
ned
1–10
No
Nor
thAm
eric
a47
617
3314
4W
ould
not
reco
mm
end
(15
)20
02Bo
thDi
agno
stic
11–5
0N
oU
SA53
2331
6711
17Re
com
men
dw
ithal
tera
tion
(16
)20
02Da
taba
seCo
mbi
ned
�10
0Ye
sU
K92
8587
9833
42St
rong
lyre
com
men
d
(17
)20
02Da
taba
seCo
mbi
ned
�10
0Ye
sU
K92
8890
9833
42St
rong
lyre
com
men
d
(18
)20
02Da
taba
seCo
mbi
ned
1–10
No
Sout
hAf
rica
1423
656
00
Wou
ldno
tre
com
men
d
(19
)20
02Bo
thDi
agno
stic
1–10
No
Cana
da81
2140
7319
13Re
com
men
dw
ithal
tera
tion
(20
)20
03Da
taba
seCo
mbi
ned
�10
0N
oCa
nada
8633
6090
2542
Reco
mm
end
with
alte
ratio
n
(21
)20
03Da
taba
seCo
mbi
ned
�10
0Ye
sN
ewZe
alan
d86
8376
9656
100
Stro
ngly
reco
mm
end
(22
)20
03Da
taba
seDi
agno
stic
51–1
00Ye
sU
SA97
2177
9039
88St
rong
lyre
com
men
d
(23
)20
03Da
taba
seDi
agno
stic
�10
0Ye
sU
SA94
2374
8142
83St
rong
lyre
com
men
d
(24
)20
03Da
taba
seDi
agno
stic
51–1
00Ye
sIn
tern
atio
nal
100
3329
5258
42Re
com
men
dw
ithal
tera
tion
(25
)20
04Bo
thCo
mbi
ned
1–10
Yes
USA
9727
6479
675
Reco
mm
end
with
alte
ratio
n
(26
)20
04Da
taba
seCo
mbi
ned
1–10
No
USA
,Can
ada
4227
665
00
Wou
ldno
tre
com
men
d
(27
)20
04Da
taba
seCo
mbi
ned
�10
0Ye
sU
K97
8892
9872
92St
rong
lyre
com
men
d
(28
)20
05Da
taba
seDi
agno
stic
11–5
0Ye
sCa
nada
7235
1385
6929
Reco
mm
end
with
alte
ratio
n
(29
)20
05Da
taba
seCo
mbi
ned
51–1
00Ye
sIn
tern
atio
nal
5846
5579
4496
Reco
mm
end
with
alte
ratio
n
(30
)20
05Da
taba
seDi
agno
stic
�10
0Ye
sAu
stra
lia97
7390
8139
21St
rong
lyre
com
men
d
(31
)20
06Da
taba
seDi
agno
stic
51–1
00Ye
sIn
tern
atio
nal
7815
2669
2521
Wou
ldn’
tre
com
men
d
(32
)20
07Bo
thCo
mbi
ned
�10
0Ye
sU
SA64
4239
6917
46Re
com
men
dw
ithal
tera
tion
(33
)20
07Bo
thCo
mbi
ned
11–5
0Ye
sU
SA61
3139
9239
0Re
com
men
dw
ithal
tera
tion
(34
)20
07Da
taba
seDi
agno
stic
0–50
Yes
Inte
rnat
iona
l86
2755
6028
9Re
com
men
dw
ithal
tera
tion
(35
)20
07Da
taba
seCo
mbi
ned
51–1
00Ye
sU
K97
7164
9056
21St
rong
lyre
com
men
d
(36
)20
07Da
taba
seCo
mbi
ned
51–1
00Ye
sU
K97
6967
8856
21St
rong
lyre
com
men
d
(37
)20
07Da
taba
seCo
mbi
ned
51–1
00Ye
sU
K75
7167
8172
29St
rong
lyre
com
men
d
Mea
n77
4554
7634
39
Rang
e14
–100
6–88
6–92
33–9
80–
720–
100
aBo
th,j
ourn
alan
del
ectr
onic
guid
elin
eda
taba
se.
bCo
mbi
ned,
diag
nost
ican
dth
erap
eutic
reco
mm
enda
tions
.c
GL
deve
lopm
ent
man
ualo
rte
chni
cald
ocum
ent
was
avai
labl
ebe
fore
GL
publ
icat
ion.
dD1
,sco
pean
dpu
rpos
e;D2
,sta
keho
lder
invo
lvem
ent;
D3,r
igor
ofde
velo
pmen
t;D4
,cla
rity
and
pres
enta
tion;
D5,a
pplic
abili
ty;D
6,ed
itoria
lind
epen
denc
e.
Quality of Diagnostic Guidelines in Diabetes
Clinical Chemistry 54:11 (2008) 1875
2005 onward, with failures to report editorial indepen-dence and conflict of interest in the majority of GLs(Table 1).
Type of publication. We investigated how GL develop-ers defined the type of their publications and whetherthese reflected the methods used for their develop-
Table 2. Performance of AGREE item scores in diabetes mellitus guidelines.
Domain and item
All GLs (n � 26)
Mean score Score >3, nRate with
score >3, %
D1a
1 The overall objective of the guideline is specifically described. 3.34 19 73
2 The clinical questions covered by the guideline are specificallydescribed.
3.16 17 65
3 The patients to whom the guideline is meant to apply arespecifically described.
3.45 21 81
D2
4 The guideline development team involves all relevantprofessional groups.
2.53 9 35
5 The patients’ views and preferences have been sought. 2.12 9 35
6 The target users of the guideline are clearly defined. 3.03 15 57
7 The guideline has been piloted among target users. 1.75 6 23
D3
8 Systematic methods were used to search for evidence. 2.50 10 38
9 The criteria for selecting the evidence are clearly described. 2.42 11 42
10 The methods used for formulating the recommendations areclearly defined.
2.30 7 27
11 The health benefits, side effects, and risks have beenconsidered.
3.14 17 65
12 There is an explicit link between the recommendations and thesupporting evidence.
3.12 19 73
13 The guideline has been externally reviewed by experts beforeits publication.
2.58 11 42
14 A procedure for updating the guideline is provided. 2.31 10 38
D4
15 The recommendations are specific and unambiguous. 3.57 22 85
16 The different options for management of the condition areclearly presented.
3.37 21 81
17 The recommendations are easily identifiable. 3.75 24 93
18 The guideline is supported with tools for application. 2.43 10 38
D5
19 Potential barriers in applying the recommendations have beendiscussed.
1.88 4 15
20 Potential cost implications of applying the recommendationshave been considered.
2.21 6 23
21 The guideline presents key review criteria for monitoring and/or audit purposes.
1.95 8 31
D6
22 The guideline is editorially independent from the funding body. 2.60 11 42
23 Conflicts of interest of guideline development members havebeen recorded.
1.95 8 31
a D1, scope and purpose; D2, stakeholder involvement; D3, rigor of development; D4, clarity and presentation; D5, applicability; D6, editorial independence.
1876 Clinical Chemistry 54:11 (2008)
ment. There was diversity in definitions: 19 publica-tions were labeled as GLs or recommendations, ofwhich 7 stated that they were evidence-based, 4 posi-tion statements or reports, and 3 guidance documents(Table 3). Among the 7 evidence-based GLs, 5 had ev-idence summaries and 6 graded recommendations.Three GLs that had evidence tables did not define theirpublications as evidence-based GLs (22, 23, 25 ). Morethan two-thirds of GLs (n � 18, 69%) defined theirgrading system, but only 16 (62%) graded their finalrecommendations (Table 3).
Procedure for updating guidelines. Item 14 investigateswhether GL developers describe the procedures for up-dating recommendations, including the timescale, re-sponsibilities, and methods used. Fifteen GLs (58%)gave a timescale or expiration date, of which 1 GL pro-vided this information in a separate GL developmentmanual of the issuing authority (Table 3). The mostfrequent review date was 3 and 4 years. Only 10 GLs(38%) provided adequate information on the updatingprocess (Table 2).
Reporting of laboratory-specific information in diagnos-tic guidelines. We investigated whether GLs covered es-sential laboratory-specific information, such as preva-lence/pretest probability and diagnostic accuracy dataor preanalytical and analytical factors critical for thecorrect interpretation and application of laboratory re-sults in clinical practice (Table 3). About 60% of theGLs mentioned these factors. Reporting these pieces ofinformation was more frequent in diagnostic com-pared to combined GLs, but the difference was not sta-tistically significant in the various GL subgroups, asdiscussed below (online Data Supplement 3).
SUBGROUP ANALYSIS
GLs were grouped according to source, scope, length,origin, and availability of a guideline methods manual,to investigate whether there were statistically signifi-cant differences in GL quality in these subsets. Resultsare shown in Table 4 and online Data Supplements 3and 4.
Subgrouping by source. Grouping GLs by source of pub-lication revealed that 1 GL was published in a peer-reviewed journal, 19 were available in electronic GLdatabases, and 6 in both sources. The GL that was pub-lished exclusively in a peer-reviewed journal (14 ) wasnot recommended for use by the assessors. None of the6 GLs published in both peer-reviewed journals and GLdatabases were strongly recommended. GLs publishedin electronic guideline databases only received a morefavorable overall assessment. A notable difference, at alevel of significance of P � 0.05, could be observed inthe D5 only for the electronic GLs (Table 4).
Subgrouping by scope. The rate of occurrence ofstrongly recommended GLs was higher for the com-bined (50%) than for the diagnostic (30%) GLs, but therate of GLs not recommended was also higher in thecombined group. The difference was moderate (P �0.05) in D2 only, with combined GLs scoring higher(Table 4). Moderate differences were also found in 4individual items (online Data Supplement 4). Diagnos-tic GLs defined their objectives better (I1) and consid-ered the cost implications of the recommendationsmore frequently (I20), whereas combined GLs definedtheir target users (I6) and their updating processesmore precisely (I14) than diagnostic ones.
Subgrouping by length. A clear relationship could bedemonstrated between GL length and methodologicalquality (Table 4). Most GLs that were not recom-mended were shorter, and all strongly recommendedguidelines were longer than 50 pages. Significant differ-ences between these subgroups could be found formost domains, with higher quality of the longer GLs.Moderate differences (P � 0.05) could be observedwith the “applicability” and “clarity and presentation”domains. However, the best-performing GLs, scoring�50% in the “applicability” domain (21, 24, 27, 28,35–37 ), were generally longer than 50 pages, and allwere published in electronic databases (Table 1).
Subgrouping by origin. The majority of the strongly rec-ommended GLs (7 of 11) originated from the UK; theother 4 were from New Zealand, Australia, and theUSA (Table 1). Significant differences (P � 0.01) couldbe observed in fulfilling the criteria of D2, with higherscores for the British GLs. In the “rigor of develop-ment” and “clarity and presentation” domains, the dif-ference was moderate (P � 0.05) (Table 4).
Subgrouping by availability of guideline methods man-ual. Two-thirds of GLs had some accompanying man-uals describing the methods of their development insome form. All strongly recommended GLs had such amanual (Table 4). All domain scores were better in thesubset where these manuals were available, and the dif-ferences were highly statistically significant (P � 0.01)in D4, D5, and D6. In D1, D2, and D3, the P values werealso significant, but at values somewhat �0.01.
Discussion
In the current study, we made special efforts to retrieveand use all available background technical materialswhen appraising GLs to avoid a biased assessment ofmethodological quality by the AGREE Instrument.Our evaluation revealed that diagnostic recommenda-tions in the field of DM suffer from the same method-ological weaknesses as GLs developed by prestigious
Quality of Diagnostic Guidelines in Diabetes
Clinical Chemistry 54:11 (2008) 1877
Tabl
e3.
Qua
litat
ive
anal
ysis
ofre
port
ing
diab
etes
mel
litus
guid
elin
es.
Gu
idel
ine
Typ
eo
fp
ub
licat
ion
asd
escr
ibed
by
auth
ors
Evid
ence
tab
leD
escr
ipti
on
of
gra
din
gsy
stem
Gra
ded
reco
mm
end
atio
ns
Rev
iew
dat
e,ye
arPr
eval
ence
/Pre
test
pro
bab
ility
Dia
gn
ost
icac
cura
cyPr
ean
alyt
ical
info
rmat
ion
An
alyt
ical
info
rmat
ion
(12
)Re
view
ofth
eev
iden
cean
dre
com
men
datio
ns�
––
––
�–
–
(13
)N
atio
nalc
linic
algu
idel
ines
–�
�3
��
–�
(14
)Co
nsen
sus
repo
rt–
––
1–
––
–(1
5)
Gui
delin
esan
dre
com
men
datio
ns–
��
–�
��
�
(16
)Cl
inic
algu
idel
ines
and
evid
ence
revi
ew�
��
4–
�–
�
(17
)Cl
inic
algu
idel
ines
and
evid
ence
revi
ew�
��
4–
––
–
(18
)G
uide
line
––
––
––
––
(19
)Cl
inic
alpr
actic
egu
idel
ines
–�
�–
��
��
(20
)Cl
inic
alpr
actic
egu
idel
ines
–�
�–
��
��
(21
)Ev
iden
ce-b
ased
best
prac
tice
guid
elin
es–
��
3�
–�
�
(22
)Re
com
men
datio
nan
dra
tiona
lest
atem
ent
��
�–
��
��
(23
)Re
com
men
datio
nan
dra
tiona
lest
atem
ent
��
�–
��
��
(24
)Re
port
––
––
��
––
(25
)Cl
inic
alpr
actic
egu
idel
ines
�–
–5
––
––
(26
)G
uide
lines
––
––
––
––
(27
)Cl
inic
algu
idel
ines
and
evid
ence
revi
ew�
��
4–
��
�
(28
)G
uide
lines
and
prot
ocol
s–
––
3–
–�
–(2
9)
Glo
balg
uide
line
–�
–3–
5�
��
�(3
0)
Evid
ence
-bas
edgu
idel
ines
��
�3
��
��
(31
)Re
port
––
––
��
�–
(32
)M
edic
algu
idel
ines
(evi
denc
eba
sed)
–�
�–
�–
�–
(33
)Po
sitio
nst
atem
ent
–�
�1a
��
��
(34
)G
uide
line
–�
�3
––
––
(35
)G
uida
nce
–�
�co
ntin
uous
––
��
(36
)G
uida
nce
–�
–co
ntin
uous
��
��
(37
)G
uida
nce
–�
�co
ntin
uous
�–
��
Perc
enta
geof
GLs
fulfi
lling
crite
ria31
6962
5858
5862
58
aIn
form
atio
non
upda
ting
ispr
ovid
edin
ase
para
tegu
idel
ine
deve
lopm
ent
man
ual.
1878 Clinical Chemistry 54:11 (2008)
Tabl
e4.
Subg
roup
anal
ysis
D1a
D2
D3
D4
D5
D6
Stro
ng
lyre
com
men
ded
,n
(%)
Rec
om
men
ded
wit
hal
tera
tio
n,
n(%
)
Wo
uld
n’t
reco
mm
end
,n
(%)
Mea
nd
om
ain
sco
re,
%SE
Ran
ge
Mea
nd
om
ain
sco
re,
%SE
Ran
ge
Mea
nd
om
ain
sco
re,
%SE
Ran
ge
Mea
nd
om
ain
sco
re,
%SE
Ran
ge
Mea
nd
om
ain
sco
re,
%SE
Ran
ge
Mea
nd
om
ain
sco
re,
%SE
Ran
ge
Sour
ce
Gui
delin
eda
taba
se(n
�19
)80
5.2
14–1
0052
6.1
15–8
858
6.5
6–92
803.
352
–98
405.
10–
7245
7.6
0–10
011
(58)
5(2
6)3
(16)
Jour
nala
ndG
Lda
taba
se(n
�7)
b70
7.1
47–9
727
4.6
6–42
436.
817
–69
648.
333
–92
183.
96–
3925
10.0
0–75
0(0
)6
(86)
1(1
4)
P0.
209
0.05
50.
169
0.08
30.
018c
0.15
2
Scop
e
Diag
nost
ic(n
�10
)85
4.5
53–1
0031
5.2
15–7
350
8.2
13–9
069
5.3
35–9
035
5.8
11–6
934
8.9
9–88
3(3
0)6
(60)
1(1
0)
Com
bine
d(n
�16
)73
6.2
14–9
754
6.8
6–88
566.
96–
9280
4.5
33–9
833
6.1
0–72
438.
80–
100
8(5
0)5
(31)
3(1
9)
P0.
286
0.02
3c0.
660
0.09
70.
776
0.55
1
Leng
th
1–50
page
s(n
�9)
618.
514
–97
242.
76–
3530
7.0
6–64
685.
833
–92
217.
40–
6916
8.0
0–75
0(0
)6
(67)
3(3
3)
�50
page
s(n
�17
)86
3.5
56–1
0056
6.2
15–8
867
4.8
26–9
280
4.1
35–9
841
4.6
8–72
527.
221
–100
11(6
5)5
(29)
1(6
)
P0.
009d
0.00
3d0.
001d
0.05
10.
018c
0.00
1d
Orig
in
Nor
thAm
eric
a(n
�12
)74
5.7
42–9
727
2.8
6–42
447.
16–
7772
5.7
33–9
225
5.5
0–69
359.
30–
882
(17)
8(6
6)2
(17)
Briti
sh(n
�7)
875.
956
–97
623.
269
–88
674.
564
–92
823.
871
–98
428.
90–
7244
10.1
0–92
7(1
00)
0(0
)0
(0)
Oth
er(n
�7)
7411
.314
–100
439.
815
–83
4811
.26–
9070
5.9
52–9
636
7.6
0–58
4115
.40–
100
2(2
8.5)
3(4
3)2
(28.
5)
P0.
355
0.00
1d0.
028c
0.03
7c0.
112
0.60
6
Man
ual
yes
(n�
19)
843.
556
–100
535.
915
–88
625.
352
–98
823.
050
–98
424.
56–
7249
7.4
0–10
011
(58)
7(3
7)1
(5)
No
(n�
7)59
10.5
14–8
925
4.0
6–40
339.
56–
6960
7.7
33–9
012
3.6
0–25
145.
60–
420
(0)
4(5
7)3
(43)
P0.
013c
0.01
5c0.
022c
0.01
0d0.
001d
0.00
4d
aD1
,sco
pean
dpu
rpos
e;D2
,sta
keho
lder
invo
lvem
ent;
D3,r
igor
ofde
velo
pmen
t;D4
,cla
rity
and
pres
enta
tion;
D5,a
pplic
abili
ty;D
6,ed
itoria
lind
epen
denc
e.b
One
guid
elin
e(1
4)
was
publ
ishe
din
jour
nalo
nly.
cP
�0.
05.
dP
�0.
01.
Quality of Diagnostic Guidelines in Diabetes
Clinical Chemistry 54:11 (2008) 1879
authorities in many other disciplines (2, 5, 38 ). Sub-group analyses of our study demonstrated that longerand electronically published GLs and the availability ofGL development manuals yielded higher methodolog-ical scores in most AGREE domains (Table 4). Onesimple explanation is the lack of space available in jour-nals for detailed and accurate reporting (39 ). Poormethodological scores could just as well reflect faultymethods that could lead to biased and/or inaccurateconclusions. Paradoxically, lengthy GLs are thought tobe less practical for daily use (39 ), so one may arguethat length of GLs adversely affects implementation. Inour case, GLs that achieved high scores for “applicabil-ity” were indeed longer documents, but they alsocovered additional information on organization, costimplications, and monitoring of the use of recommen-dations in practice. All these tools help GL implemen-tation, and thus we cannot confirm that lengthy GLsare not applicable in practice. The Conference onGuideline Standardization defined a standard for GLreporting to promote quality and facilitate implemen-tation (11 ). Such GL reporting standards have not yetbeen adopted by most journals, and peer reviewers alsorarely use the AGREE or other criteria for systematicassessment of recommendations before publication(40 – 42 ). These shortcomings suggest the need for GLreporting standards, similar to the STARD documentfor reporting diagnostic accuracy studies (43 ), andclear publication and peer review policies for GLs bymajor medical journals.
In our study, the quality of purely diagnostic GLswas not significantly different from that of combinedGLs (Table 4). Our additional evaluation in Table 3showed that nearly half of all GLs do not report preana-lytical, analytical, and diagnostic accuracy data (3 ),which may lead to inappropriate interpretation of testresults in clinical practice (44 ). Fulfilling these criteriawould be desirable in any GLs that deal with laboratorytesting–related issues, since it is expected that practicerecommendations are developed in a multidisciplinaryprocess (45 ). Unfortunately, this could not be con-firmed by our study, as only 41% of the criteria werefulfilled in D2, which explored the involvement of allrelevant stakeholders in the GL development process.
All GLs that scored better in the comparison byorigin were from agencies with detailed GL manualsthat provided a clear description and standards for thedevelopment process (Table 1). The availability of a GLmanual, however, does not always guarantee that GLteams follow those processes consistently, and it hasbeen shown that it is often not clear how decisions aremade by the GL team when arriving at final recommen-dations (8 ). The substantial heterogeneity, in both howthe type of publication is defined and the adherence tothis definition in the final presentation of the GL, sug-
gests that there is likely a disparity between the meth-odology GL developers describe and what is actuallyfollowed in practice (Table 3). We found several GLsthat described a grading system but did not grade theirfinal recommendations. The lack of evidence tables inGLs that claim to be evidence-based may also point topotential deviations from the processes set in GL man-uals. Therefore it is advisable that diagnostic GL devel-opment teams adhere to preset methodology and doc-ument the procedures followed explicitly.
We could not demonstrate major improvementsin GL quality for most domains, and in the “editorialindependence” domain, deterioration in scores wasobserved over time. We further evaluated the quality ofGLs over time in some cases where the authorities is-sued several GLs [e.g., National Institute for ClinicalExcellence (NICE), WHO, International Diabetes Fed-eration (IDF)] within the time scale investigated (datanot shown). The NICE GL in 2004 is of higher qualitythan the NICE 2002 version due to improvements in“applicability” and “editorial independence” domains.It is noteworthy that many international organizationshave improved the rigor of their guideline develop-ment process and are moving toward internationalstandardization (11, 46 – 48 ). Surprisingly, the inter-national WHO and IDF GLs in 2006 and 2007 hadlower scores in most domains than the 2003 and 2005versions, despite the fact that both agencies releasedguideline development manuals in 2003 (http://whqlibdoc.who.int/hq/2003/EIP_GPE_EQC_2003_1.pdf, http://www.idf.org). Therefore, we assume that the lowerAGREE scores are due to the lack of reporting somemethodological details rather than the lack of followingthe methodology described in the manuals. Explicit re-porting of methodology and adherence to that meth-odology is particularly important for influential agen-cies (e.g., American Diabetes Association and WHO)whose recommendations are followed or adaptedworldwide.
There are several limitations in our study. By eval-uating English publications only, our results may sufferfrom language bias. However, several publications, in-cluding our own review of the topic, confirm no signif-icant differences in the quality of English vs non-En-glish publications of guidelines or trials (2, 49, 50 ).Because most national DM GLs are based on orstrongly influenced by international recommendationsprimarily published in English, we believe our resultsare likely to be generalizable.
Our study evaluated different publications thatwere defined in various ways by their authors. Suchheterogeneity of definitions (such as guideline, guid-ance, protocols, position statement, recommendationand rationale statement, consensus report) may high-light different approaches in formulating recommen-
1880 Clinical Chemistry 54:11 (2008)
dations for practice. We also found several GLs that,while having proof of using evidence-based methods,failed to define their publication as such (22, 23, 25 ).This suggests that the definitions used in the interna-tional guideline community may be confusing for bothGL developers and users, and that simplification andstandardization of terminology is needed. One may ar-gue that AGREE can be used for assessing evidence-based GLs only. However, AGREE is a generic andwidely accepted toolbox (8 ) that can investigate the GLdevelopment process irrespective of whether it appliesevidence- or consensus-based methodology (4 ). Infact, most evidence-based GLs have a substantial ele-ment of consensus-based judgment, especially whenevidence is conflicting or lacking. In the latter case, GLdevelopers should still search for and appraise the “bestavailable” evidence before they conclude that the bestthey can do is to reach consensus.
Our study does not determine whether there arerelationships between the methodological quality ofGLs and the validity of their content. The AGREE In-strument or other GL appraisal tools can investigateneither the accuracy of the content of recommenda-tions nor their impact on patient outcomes (51, 52 ).Another shortcoming of all critical appraisal tools isthat they do not differentiate between whether the pub-lication fails certain criteria due to lack of reporting orto poor methodology and design. Therefore, our re-sults should not be interpreted as criticisms of the truthof scientific statements or the validity of recommenda-tions made in a given publication. However, the dem-onstrated shortcomings in reporting and/or the meth-odology applied by different GL developers could leadto distrust in and/or misuse of recommendations (53 ).With such shortcomings, the energy put into develop-
ing scientifically accurate but otherwise poorly pre-sented GLs could end up being wasted, whereas inac-curate but otherwise nicely presented GLs might bepromoted and used widely. This is why we advise thatGLs be critically evaluated for both methodology andcontent before recommendations are used in clinicalpractice (38 ).
In conclusion, our results suggest the need for sys-tematically developed, explicit recommendationsbased on evidence-based guideline development andreporting standards in laboratory medicine. Our studyalso highlights the need for simplification and stan-dardization of GL terminology. Further studies areneeded to explore in depth the relationship betweenthe scientific validity and the methodological quality ofdiagnostic recommendations in DM.
Author Contributions: All authors confirmed they have contributed tothe intellectual content of this paper and has met the following 3 require-ments: (a) significant contributions to the conception and design, acqui-sition of data, or analysis and interpretation of data; (b) drafting orrevising the article for intellectual content; and (c) final approval of thepublished article.
Authors’ Disclosures of Potential Conflicts of Interest: No authorsdeclared any potential conflicts of interest.
Role of Sponsor: The funding organizations played no role in thedesign of study, choice of enrolled patients, review and interpretationof data, or preparation or approval of manuscript.
Acknowledgments: Several authors of this article (J. Watine, W.Oosterhuis, D. Rogic, S. Sandberg, P. S. Bunting, A.R. Horvath) aremembers of the Committee on Evidence-Based Laboratory Medicineof IFCC, and this work was carried out in collaboration with theIFCC Task Force on the Global Campaign for Diabetes Mellitus.
References
1. Field MJ, Lohr KN, eds. Guidelines for clinicalpractice: from development to use. Washington,DC: National Academy Press; 1992. 426 p.
2. Horvath AR, Nagy E, Watine J. Critical appraisalof guidelines. In: Evidence-based LaboratoryMedicine: Principles, Practice, Outcomes. PriceCP, Christenson RH (eds) AACC Press, Washing-ton. 2nd edition, 2007;295–319.
3. Oosterhuis WP, Bruns DE, Watine J, Sandberg S,Horvath AR. Evidence-based guidelines in labora-tory medicine: principles and methods. Clin Chem2004;50:806–18.
4. The AGREE Collaboration. Appraisal of Guidelinesfor Research and Evaluation (AGREE) Instrument.http://www.agreecollaboration.org (Accessed De-cember 2007).
5. Horvath AR, Nagy E, Watine J. Quality of guide-lines for the laboratory management of diabetesmellitus. Scand J Clin Lab Invest Suppl 2005;240:41–50.
6. Hunt DL, McKibbon KA. Locating and appraisingsystematic reviews. Ann Intern Med 1997;126:
532–8.7. Horvath AR, Pewsner D. Systematic reviews in
laboratory medicine: principles, processes andpractical considerations. Clin Chim Acta 2004;342:23–39.
8. Van der Wees PJV, Hendriks EJM, Custers JWH,Burgers JS, Dekker J, de Bie RA. Comparison ofinternational guideline programs to evaluate andupdate the Dutch program for clinical guidelinedevelopment in physical therapy. BMC HealthServices Research 2007;7:191. http://www.biomedcentral.com/1472–6963/7/191 (AccessedDecember 2007).
9. The AGREE Collaboration. Development and val-idation of an international appraisal instrumentfor assessing the quality of clinical guidelines: theAGREE project. Qual Saf Health Care 2003;12:18–23.
10. MacDermid JC, Brooks D, Solway S, Switzer-McIntyre S, Brosseau L, Graham ID. Reliabilityand validity of the AGREE instrument used byphysical therapists in assessment of clinical prac-
tice guidelines. BMC Health Services Research2005;5:18. (http://www.biomedcentral.com/1472–6963/5/18) (Accessed December 2007).
11. Shiffman RN, Shekelle P, Overhage JM, Slutsky J,Grimshaw J, Deshpande AM. Standardized re-porting of clinical practice guidelines: a proposalfrom the Conference on Guideline Standardiza-tion. Ann Intern Med 2003;139:493–8.
12. Woolf SH, Davidson MB, Greenfield S, Bell HS,Ganiats TG, Hagen MD et al. American Academyof Family Physicians and American Diabetes As-sociation. The benefits and risk of controllingblood glucose levels in patients with type 2 dia-betes mellitus: a review of the evidence andrecommendations. AAFP Policy Action April 1999.http://www.aafp.org/online/etc/medialib/aafp_org/documents/clinical/clin_recs/diabetespolicy.Par.0001.File.tmp/clinicalrecs_diabetespolicy05.pdf(Accessed December 2007).
13. Scottish Intercollegiate Guidelines Network.Management of diabetes CPG55. 2001. http://www.sign.ac.uk/guidelines/fulltext/55/index.html
Quality of Diagnostic Guidelines in Diabetes
Clinical Chemistry 54:11 (2008) 1881
(Accessed December 2007).14. Reece EA, Homko C, Miodovnik M, Langer O. A
consensus report of the Diabetes in PregnancyStudy Group of North America Conference, LittleRock, Arkansas, May 2002. J Matern Fetal Neo-natal Med 2002;12:362–4.
15. Sacks DB, Bruns DE, Goldstein DE, Maclaren NK,McDonald JM, Parrott M. Guidelines and recom-mendations for laboratory analysis in the diagno-sis and management of diabetes mellitus. ClinChem 2002;48:436–72.
16. McIntosh A, Hutchinson A, Home PD, Brown F,Bruce A, Damerell A, Davis R et al. (2001) Clinicalguidelines and evidence review for type 2 dia-betes: management of blood glucose. Sheffield:ScHARR, University of Sheffield. http://www.nice.org.uk/cat.asp?c�36733 (Accessed December2007).
17. McIntosh A, Hutchinson A, Feder G, Durring-ton P, Elkeles R, Hitman GA, et al. (2002) Clinicalguidelines and evidence review for type 2 dia-betes: lipids management. Sheffield: ScHARR,University of Sheffield. http://www.nice.org.uk/cat.asp?c�38551 (Accessed December 2007).
18. Society for Endocrinology, Metabolism and Dia-betes of South Africa. Revised SEMDSA Guide-lines for diagnosis and management of type 2diabetes mellitus for primary health care in 2002.http://www.semdsa.org.za/guidelines.htm (Ac-cessed December 2007).
19. Berger H, Crane J, Farine D, Armson A, De LaRonde S, Keenan-Lindsay L, et al. Screening forgestational diabetes mellitus. J Obstet GynaecolCan 2002;24:894–912.
20. Canadian Diabetes Association Clinical PracticeGuidelines Expert Committee. Canadian Dia-betes Association 2003 Clinical Practice Guide-lines for the Prevention and Management ofDiabetes in Canada. http://www.diabetes.ca/cpg2003/downloads/cpgcomplete.pdf (AccessedDecember 2007).
21. New-Zealand Guidelines Group. Management ofType 2 Diabetes. http://www.nzgg.org.nz/guidelines/dsp_guideline_popup.cfm?guidelineCatID�32&guidelineID�36 (Accessed December 2007).
22. U.S. Preventive Services Task Force. Screening fortype 2 diabetes mellitus in adults: recommenda-tions and rationale. February 2003. Agency forHealthcare Research and Quality, Rockville, MD.http://www.ahrq.gov/clinic/uspstf/uspsdiab.htm(Accessed December 2007).
23. U.S. Preventive Services Task Force. Screening forgestational diabetes mellitus: recommendationsand rationale. http://www.ahrq.gov/clinic/uspstf/uspsgdm.htm (Accessed December 2007).
24. World Health Organization. Screening for type 2diabetes: report of a World Health Organizationand International Diabetes Federation meeting.Geneva, 2003. http://whqlibdoc.who.int/hq/2003/WHO_NMH_MNC_03.1.pdf (Accessed December2007).
25. Snow V, Aronson MD, Hornbake ER, Mottur-Pilson C, Weiss KB. Clinical Efficacy AssessmentSubcommittee of the American College of Physi-cians. Lipid control in the management of type 2diabetes mellitus: a clinical practice guideline
from the American College of Physicians. AnnIntern Med 2004;140:644–9.
26. Kaiser Permanente, Care Management Institute’sDiabetes Guidelines workgroup. Guidelines forthe Management of Adult Diabetes in PrimaryCare. http://members.kaiserpermanente.org/kpweb/pdf/feature/247clinicalpracguide/CMI_AdultDiabetesGuideline_public_web_033104.pdf (Accessed De-cember 2005).
27. National Collaborating Centre for Women’s andChildren’s Health (NICE guidance). Type 1diabetes: diagnosis and management of type 1diabetes in children and young people. http://www.nice.org.uk/page.aspx?o�213575 (Ac-cessed December 2007).
28. Guidelines and Protocols Advisory Committee (BrColumbia Ministry of Health). Diabetes Care, re-vised 2005. http://www.health.gov.bc.ca/gpac/guideline_diabetes.html (Accessed December2007).
29. IDF Clinical Guidelines Task Force. Global guide-line for type 2 diabetes. Brussels: InternationalDiabetes Federation, 2005. http://www.idf.org/home/index.cfm?node�1457 (Accessed Decem-ber 2007).
30. National Health and Medical Research Council.National evidence based guidelines for the man-agement of type 2 diabetes mellitus. http://www.health.gov.au/nhmrc/publications/pdf/cp86.pdf(Accessed November 2007).
31. World Health Organization: Definition and diag-nosis of diabetes mellitus and intermediatehyperglycemia: report of a WHO/IDF Consulta-tion. Geneva, 2006. http://www.who.int/diabetes/publications/Definition%20and%20diagnosis%20of%20diabetes_new.pdf (Accessed December2007).
32. American Association of Clinical Endocrinolo-gists. American Association of Clinical Endocri-nologists medical guidelines for clinical practicefor the management of diabetes mellitus. EndocrPract 2007;13(Suppl 1):1–68.
33. American Diabetes Association. Standards ofmedical care in diabetes. Diabetes Care 2007;30:S4–41.
34. IDF Clinical Guidelines Task Force. Guideline formanagement of postmeal glucose. Brussels: In-ternational Diabetes Federation, 2007. http://www.idf.org/home/index.cfm? node�1620 (AccessedDecember 2007).
35. Sowerby Centre for Health Informatics at New-castle. PRODIGY Guidance: Diabetes type 2: lipidmanagement. http://www.prodigy.nhs.uk/guidance.asp?gt�Diabetes%20-%20lipid%20management(Accessed December 2007).
36. Sowerby Centre for Health Informatics atNewcastle: PRODIGY Guidance: Diabetes type 2:renal disease. http://www.prodigy.nhs.uk/guidance.asp?gt�Diabetes%20-%20renal%20disease (Ac-cessed December 2007).
37. Sowerby Centre for Health Informatics atNewcastle: PRODIGY Guidance: Blood glucose.http://www.prodigy.nhs.uk/guidance.asp?gt�Diabetes%20-%20renal%20disease (AccessedDecember 2007).
38. Qaseem A, Vijan S, Snow V, Cross T, Weiss KB,
Owens DK, for the Clinical Efficacy AssessmentSubcommittee of the American College of Physi-cians. Glycaemic control and type 2 diabetesmellitus: the optimal hemoglobin A1c targets. Aguidance statement from the American College ofPhysicians. Ann Intern Med 2007;147:417–22.
39. Deeks JJ. Word limits best explain failings ofindustry supported meta-analyses [Letter]. BMJ2006;333:1021.
40. Fervers B, Burgers JS, Haugh MC, Brouwers M,Browman G, Cluzeau F, Philip T. Predictors ofhigh quality clinical practice guidelines: examplesin oncology. Int J Qual Health Care 2005;17:123–32.
41. Miller J, Petrie J. Development of practice guide-lines [Commentary]. Lancet 2000;355:82–3.
42. vanTulder MW, Tuut M, Pennick V, Bombardier C,Assendelft WJJ. Quality of primary care guidelinesfor acute low back pain. Spine 2004;29:E357–62.
43. Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA,Glasziou PP, Irwig LM, et al. The STARD state-ment for reporting studies of diagnostic accuracy:explanation and elaboration. Clin Chem 2003;49:7–18.
44. Skeie S, Nordin G, Oosterhuis WP, Araczki A,Horvath AR, Perich C et al. Post-analytical exter-nal quality assurance of blood glucose andHbA1c: an international survey. Clin Chem 2005;51:1145–53.
45. Grimshaw GM, Khunti K, Baker R. Diagnosis ofheart failure in primary care: an assessment ofinternational guidelines. Br J Gen Pract 2001;51:384–6.
46. Raine R, Sanderson C, Black N. Developing clin-ical guidelines: a challenge to current methods.BMJ 2005;331:631–3.
47. Oxman AD, Fretheim A, Schunemann HJ, SURE.Improving the use of research evidence in guide-line development: introduction. Health Res PolicySyst 2006;4:12.
48. Schunemann HJ, Fretheim A, Oxman AD. Improv-ing the use of research evidence in guidelinedevelopment: I. Guideline for guidelines. HealthRes Policy Syst 2006;4:13.
49. Moher D, Fortin P, Jadad AR, Juni P, Klassen T, LeLorier J, et al. Completeness of reporting of trialspublished in languages other than English: impli-cations for conduct and reporting of systematicreviews. Lancet 1996;347:363–6.
50. Burgers JS, Grol R, Klazinga NS, Makela M, ZaatJ. Towards evidence-based clinical practice: aninternational survey of 18 clinical guideline pro-grams. Int J Qual Health Care 2003;15:31–45.
51. Vlayen J, Aertgeerts B, Hannes K, Sermeus W,Ramaekers Dirk. A systematic review of appraisaltools for clinical practice guidelines: multiple sim-ilarities and one common deficit. Int J QualHealth Care 2005;17:235–42.
52. Burgers JS. Guidelines quality and guidelinescontent: are they related [Editorial]. Clin Chem2006;52:3–4.
53. Woolf SH, Grol R, Hutchinson A, Eccles M, Grim-shaw J. Potential benefits, limitations, and harmsof clinical guidelines. BMJ 1999;318:527–30.
1882 Clinical Chemistry 54:11 (2008)