EDITORIAL

How can clinical epidemiology better support evidence-based

guidelines and policies in low-income countries?
GRADE (Grading of Recommendations, Assessment,
Development, and Evaluation) has been adopted by over
50 organizations that develop clinical and policy guidelines
(http://www.gradeworkinggroup.org); the Journal of Clini-
cal Epidemiology is proud to be publishing their latest
20-part series. In this issue, English and Opiyo present an in-
teresting viewpoint on GRADE from a low-income country
perspective. They are concerned that low income settings
could increasingly be found standing still, while best prac-
tices accelerate away from them. They believe that this
can be avoided by a strategic approach in low-income coun-
tries to explain, popularize, and build the capacity in the
preparation of evidence using systematic reviews and the
GRADE approach. To move from evidence to recommenda-
tions and local evidence-informed policies, they call for the
adaptation and adoption of local data. This call is being
heeded. The Cochrane Collaboration leadership have priori-
tized this issue e they held a special workshop in March
2011 on how the Cochrane network can increase low- and
middle-income country (LMIC) involvement [1]. In addition
to the resources mentioned in this commentary, the
SUPPORT (http://www.support-collaboration.org) and
EVIPNET (http://www.who.int/rpc/evipnet/) resources are
useful in providing LMIC summaries and other knowledge
translation materials based on systematic reviews, many of
which are from the Cochrane Library. The Campbell Collab-
oration have just established a newReviewGroup in Interna-
tional Development (http://www.campbellcollaboration.org/
news_/International_Development_invitation_to_register.ph)
and are calling for reviews, especially those focusing on social
and economic development interventions. There is a major
focus on developing low- and middle-income country capac-
ity, partnering with the Cochrane Collaboration (especially
the Effective Practice and Organisation and Care group, the
Equity methods group, and the Public Health Groups) and
other international organizations in both LMIC and high-
income countries to establish a network of researchers, policy
makers, and practitioners who will produce and disseminate
systematic reviews of high policy relevance on the effects of
international development interventions. The International
Clinical Epidemiology Network (INCLEN; http://www.
inclentrust.org/) are also contributing with their 89 clinical ep-
idemiology groups in 36 low- and middle-income countries.
JCE readers can watch for the coming series from the
IndiaCLEN and the West Indies Clinical Epidemiology
groups. Most recently, the International Society for Evidence
Based Health Care has been established [2]. Their mission is
to develop and encourage research in evidence-based
healthcare and to promote and provide professional and pub-
lic education in the eld. They already have many members
who live in and are committed to capacity development in
lower- and middle-income countries for evidence-based-
medicine (EBM) and evidence-based healthcare (EBHC)
and the related professional and public education. There is
good collaboration between the Society and JCE via mutual
Board membership. We would welcome comments and ideas
on EBM and EBHC capacity development in lower- and
middle-income countries from our readers.
Another issue acutely affecting low- and middle-income
countries is the issue of cost effectiveness. Value for money
for the community as a whole, based on cost-effectiveness,
is being increasingly demanded by policy makers in all
countries, and rightly so. A commentary by Calvert, Wood,
Freemantle et al. challenges us in the clinical epidemiology
community to pay more attention to ways in which our
methods can optimally contribute to cost-effectiveness.
Clinical epidemiologists are frequently opinion leaders
who can inuence the design of studies to evaluate inter-
ventions. Establishing consensus on pragmatic trials was
an important rst step [3], but we need to be more open
to better designing and tailoring real-world trials that
go beyond traditional assessment of efcacy and safety to
examine preferences, adherence, and value for money ques-
tions, to provide the evidence base for evidence-informed
practice and regulatory decisions.
As mentioned above, we need to improve and utilize
measures for assessing patient preferences. Murad et al
sought to examine patients preferences regarding the design
of diabetes trials. They sought to determine, rstly, whether
patients prefer trials to focus on patient-important outcomes,
as opposed to surrogate outcomes, and secondly, whether
patients prefer that trials provide practical/pragmatic an-
swers as opposed to mechanistic/explanatory answers. They
found that individuals with diabetes report a strong prefer-
ence for practical trials measuring the effect of treatments
on patient-important outcomes.
Reporting-guidelines feature prominently in this issue.
Many journals endorse reporting guidelines and expect
0895-4356/$ - see front matter 2011 Elsevier Inc. All rights reserved.
doi: 10.1016/j.jclinepi.2011.04.008
Journal of Clinical Epidemiology 64 (2011) 705e707
authors to follow them. Little or no attention has been paid
to their quality. In the rst systematic review of reporting
guidelines, Moher et al reviewed 81 reporting guidelines
to see how many of these met the guidance criteria recently
proposed by these authors. As they argue, transparency and
documentation should be required for the following: the
preparation process, any face-to-face meeting, structured
discussion, rationale for inclusion of items, authorship, pro-
vision of an explanatory document, feedback and criticism,
and endorsement plans. No criterion was met in more than
30%, except for plans for endorsement (49%). These nd-
ings represent an important message when reviewing re-
porting guidelines for possible publication.
In another article on reporting guidelines, Gagnier et al ex-
amine the quality of reporting and predictors of reporting in
randomized clinical trials (RCTs) of herbal medicine inter-
ventions using the newly developed set of reporting guide-
lines building on the Consolidated Standards of Reporting
Trials (CONSORT) statement. Their results indicate that
RCTs of herbal medicine interventions frequently do not re-
port important characteristics of the intervention. Trialists
are recommended to refer to the CONSORT for herbal med-
icines when reporting their trials.
Literature searching is still an art as well as a science.
A well-designed database-specic search strategy is an
important topic that forms the basis of any research. Katch-
amart et al compared the performance of Ovid-MEDLINE
with PubMed as a means of identifying randomized con-
trolled trials of methotrexate in patients with rheumatoid
arthritis. They found that PubMed had slightly higher sen-
sitivity than Ovid-MEDLINE in terms of comparable preci-
sion and number needed to read. They also point to the
importance of seeking the assistance of experienced infor-
mation specialists or medical librarians for developing
comprehensive search strategies for systematic reviews.
Comorbidity measures are rightly attracting ongoing at-
tention. Gagne et al argue that by summarizing various med-
ical conditions into single numerical indices, comorbidity
scores can provide a standardized summary of the burden of
comorbidity in a study group, increase analytic efciency,
and allow for adjustment of more potentially confounding
baseline conditions than otherwise possible. These authors
sought todevelopandvalidate a single numerical comorbidity
score for predicting short- and long-term mortality, by com-
bining conditions in the Charlson and Elixhauser measures.
They conclude that the combined score may offer improve-
ments in comorbidity summarization over existing scores.
How many JCE readers know what spie charts are?
Better graphical displays are urgently needed for displaying
multivariate healthcare data in the context of comparative
evaluation of clinical outcomes among populations or
healthcare providers. Stafoggia et al analyzed the strengths
and weaknesses of different graphical techniques including;
target plots, radar plots, and spie charts, for comparing the
performances of different healthcare providers. They found
that the spie chart was able to distinguish data subtleties
missed by other graphical techniques, in terms of clarity, in-
tuitiveness, and completeness of the information displayed.
Two groups report on specic outcome instruments to
exemplify methods challenges. De Witt et al present self-
report physical function measures for children using mod-
ern psychometric methods for item analysis as part of
patient-reported outcomes measurement information sys-
tem (PROMIS). PROMIS pediatric physical function scales
show excellent test properties in preliminary large-
scale validation testing in children aged 8 to 17 years.
Sherrington and coworkers report a prospective cohort
study that aims to develop and internally validate a falls
prediction tool for people being discharged from inpatient
aged care rehabilitation. Once this simple tool has been
externally validated, it could be used to quantify the prob-
ability with which an individual will fall in the 3 months
after an aged care rehabilitation stay. It may assist in the
discharge process by identifying high-risk individuals
who may benet from ongoing assistance or intervention.
Instrumental variables are making a come-back with in-
creasing numbers of manuscripts submitted to JCE. In this
issue, the article by Chen and Briesacher propose the in-
strumental variable approach as a method for addressing
measured and unmeasured confounding in observational
studies. In this review of instrumental variable methods in
prescription drug research, they show that instrumental
variables may be useful if they are strongly and unbiasedly
associated to treatment assignment and if they are uncorre-
lated with factors predicting the outcome.
Burns et al tackle the important issues related to missing
data. In their study using the Mini-Mental State Examination
(MMSE), which is used to estimate current cognitive status
and is a screen for possible dementia, they indicate that mul-
tiple imputation (MI) is an effective method for imputing
missing item-level data for the Mini-Mental State Examina-
tion (MMSE) and, in this case, was preferred over other
methods of dealing with cases with missing data. Bayesian
approaches are conceptually appealing but are devilishly dif-
cult to implement. Hsiao et al successfully used Bayesian
hierarchical correlated random-effects models for the infer-
ence of intra-rater and test-retest reliability simultaneously.
The assessment of inter- and intra-rater reliability usually
involves more than one level of nesting structures in the col-
lected data, where repeated observations are made by multi-
ple raters. Most approaches, however, are not designed to
accommodate both inter- and intra-rater reliability jointly.
They argue that this Bayesian hierarchical correlation model
offers a wide applicability, exibility, and feasibility in mod-
eling inter- and intra-rater reliability together.
Peter Tugwell
Andre Knottnerus
Editors
Leanne Idzerda
Assistant Editor
E-mail address: lidzerda@uottawa.ca (L. Idzerda)
706 Editorial / Journal of Clinical Epidemiology 64 (2011) 705e707
References
[1] McDonald S. Cochrane Strategic Session on Ensuring The Cochrane
Collaboration enables better global participation. Split, Croatia, March
30, 2011. The Cochrane Collaboration 2011. Available at http://www.
cochrane.org/news/tags/authors/cochrane-strategic-session-ensuring-
cochrane-collaboration-enables-better-global-p.
[2] Prasad K. International Society for Evidence-Based Health Care: Mis-
sion, Vision, and Structure, in International Society for evidence-based
healthcare. Newsletter 1, November 2010. Available at http://ebm.
mcmaster.ca/documents/ebhc_newsletter_2010.pdf.
[3] Zwarenstein M, Treweek S, Gagnier JJ, Altman DG, Tunis S, Haynes B,
Oxman AD, Moher D. Improving the reporting of pragmatic trials: An
extension of the CONSORT statement. BMJ 2008;337(7680):1223e6.
707 Editorial / Journal of Clinical Epidemiology 64 (2011) 705e707