Journal of Clinical Epidemiology 67 (2014) 359e360

EDITORIAL

Basic science, evidence, and clinical judgement

Knowledge resulting from basic research may lead to out in systematic reviews for most public health evalua-
clinical research and subsequent practical application. But tions. And when it is done, often the simplest methods
it also happens that clinical observation and experience are used, while more sophisticated methods are availabe
evoke research that confirms previous judgement. In addi- to make the evaluations more useful for decision makers.
tion, evidence of effectivenes has been produced without The authors suggest that researchers working on public
any plausible pathophysiological explanation yet available. health evaluations should expand their toolbox.
In other words, it seems that there are no standard recipes in According to Kim et al., a serious limitation in develop-
how effective clinical practice is being scientifically pre- ing the evidence base to make the best clinical decisions is
pared and realized. But, whatever route is followed, patho- that comparisons between interventions for a certain health
physiological insights, evidence for effectiveness, and problem are often made within a certain class of interven-
clinical judgement are coherenty contributing to the best tions (e.g., pharmaceutical or surgical) or settings. How-
care for individual patients [1e3]. ever, for optimizing clinical practice, it is often necessary
The complex and sometimes cumbersome relation be- to make comparisons across different types of interven-
tween basic science, observational studies, and RCT- tions. In a review of treatments for basal cell carcinoma,
based evidence is addressed in a commentary by Prasad these authors for the first time applied a network geometry
and Ho. They critically discuss the translation of preclinical to address this issue and found that many important com-
findings to medical practice, and also make suggestions to parisons have still not been made. This finding has impor-
promote all available previous research findings being ap- tant implications for future trial comparisons.
propriately used and reviewed when evaluating the added In the triad of basic science, evidence, and judgement, the
value of new studies. By the same token, systematic re- latter e while being relevant in each consultation - has been
views representing the available clinical evidence must be much less investigated than the other two. This concerns both
as useful as possible, and topics nominated for reviews the process of individual judgement and consensus proce-
should therefore be well-refined and focused on relevant dures. Therefore, the study by Diamond and colleagues is
questions. For this purpose, based on the experience from welcomed. In a systematic review, they investigated how
the Agency for Healthcare Research and Quality (AHRQ) consensus is defined and operationalized in Delphi studies,
effective Health Care Program, Buckley et al. developed and highlight the role of consensus in those studies. The au-
a comprehensive set of guiding principles and methodolog- thors conclude that definitions of consensus vary consider-
ical recommendations that may help investigators refine ably, and propose methodologic criteria for the reporting of
topics for reviews. A broad range of elements, from the Delphi studies.
state of science to responsiveness to stakeholder inputs, Optimal translation from research to practice also im-
are among the criteria and principles to be used. The au- plies requirements for data analysis and presentation. In an-
thors expect that the methods in this field will continue to alyzing a trial, it is important that demonstrated effects are
evolve. summarized and reported appropriately. In a commentary,
Optimally and efficiently contributing to an appropriate Furuya- Kanamori and Doi evaluate the pros and cons of
picture of available evidence requires that overlap in pri- using relative risk and the odds ratio, and the role of the
mary studies in overviews (reviews of reviews) is well ad- baseline risk. The importance of the baseline situation is
dressed. However, according to Pieper and his group, the also addressed by Rouquette and her team: using data from
degree of such overlap has not been examined systemati- a large cohort of hospitalized patients, their study demon-
cally. In a systematic review of overviews, these authors strates that the minimally clinically important difference
show that overlaps are often not mentioned. They propose (MCID) defined as a function of a range of baseline scores
the CCA (corrected covered area) method to comprehen- leads to a better classification than the MCID without con-
sively report overlaps. Another relevant observation as to sidering baseline severity. Furthermore, a classic challenge
the quality of synthesis of evidence is reported by Achana in the analysis of trials is how to deal with subgroup anal-
and coworkers. In a systematic review of National Institute ysis. As results of statistical testing of subgroup analyses
for Health Care Excellence (NICE) public health apprais- can be different when relative or absolute effect measures
als, they found that quantitative synthesis is not carried are used, Venekamp and coworkers systematically reviewed
0895-4356 Ó 2014 Elsevier Inc. Open access under CC BY-NC-ND license.
http://dx.doi.org/10.1016/j.jclinepi.2014.01.005
360 Editorial / Journal of Clinical Epidemiology 67 (2014) 359e360

randomized trials published in 5 major journals. It was among elderly people. It was found that, whereas the GALI
found that, almost always, relative effect measures were mainly measured functional disability, SRH mainly meas-
used. The authors advise that the CONSORT statement ured physical morbidity.
should recommend the reporting of both relative and abso- In striving for more efficient collection of evidence, it is
lute reduction for subgroup analyses. Another field where often assumed that reducing questionnaire length would im-
more methodological guidance is suggested is the predic- prove response rate in surveys among physicians. This was
tion of chronic disease evolution from a prognostic marker. not confirmed in a randomized trial of Bolt et al., comparing
Dantan et al. propose the use of simple equations to assess the responses to a long and short version of a questionnaire
time-dependent sensitivity, specificity, predictive values, on end-of-life decisions. But sending a a drastically short-
likelihood ratios, and posttest probability ratios, which ened questionnaire version to non-responders did improve
can help readers to better evaluate research articles report- the response rate. Another method to increase physician sur-
ing on prognostic markers. Their approach is illustrated us- vey response, a charitable donation incentive, was tested by
ing data from reported prognostic studies on kidney Nesrallah c.s. in a randomized trial. In their letter the au-
transplantation and breast cancer. thors report no effect, and given the growing demand on
Although for diagnostic accuracy research the develop- physicians time they make a plea for better strategies.
ment of methodological criteria is much younger than for
randomized trials, standards for reporting primary studies
and reviews have been available now for a number of years.
This enabled Henschke c.s. to provide a literature review of Many thanks to Martin Prins, welcome to John
all systematic reviews of diagnostic test accuracy studies in Ioannidis
the musculoskelal field. They found much room for im- After many years of service to the Journal, professor
provement of primary studies, and positively evaluated Martin Prins, professor of Clinical Epidemiology at Maas-
the validity and consistency of the QUADAS checklist for tricht University, has retired from his post of associate edi-
the accuracy of diagnostic accuracy studies. tor. We have highly appreciated his contributions to further
Medical decisions in the field of clinical diagnosis are developing the journal and his excellent inputs at our
often related to optimal thresholds. In a simulation study, annual editorial retreats. At the same time, we are delighted
Hirschfeld and colleagues evaluated the performance of to welcome professor John Ioannidis as new associate edi-
empirically defined thresholds in various samples. They tor. He is professor of Medicine and director of the Stanford
found that optimal thresholds for tests that are known to re- Prevention Research Center at Stanford University School
sult in medium to large differences, at the population level of Medicine (USA), holding also various other positions
still result in many misclassifications. Researchers must at that university, and professor and chairman at the Depart-
therefore be careful in defining such cut points, and optimal ment of Hygiene and Epidemiology, University of Ioannina
thresholds should be validated in prospective studies before School of Medicine, Greece. He is one of the world’s most
recommending them for clinical use. creative and productive medical and epidemiological re-
There has been substantial methodological debate on the searchers. We are happy that John joins the core team of
stepped-wedge design, and opinions and experiences differ the Journal of Clinical Epidemiology.
[4]. Zhan et al. report on the advantages and disadvantages
of this design type, based on their experience with a trial J. Andre Knottnerus
on detecting curable recurrences during follow-up after col- Peter Tugwell
orectal cancer. They conclude that the design is a strong Editors
alternative for pragmatic cluster randomized trials, but diffi- E-mail address: anneke.germeraad@maastrichtuniversity.nl
culties are the timing of the informed consent procedure and (J.A. Knottnerus)
the complexity of the data analysis.
Self-reported health is an important source of informa- References
tion, but opportunities to compare this with objective meas-
[1] Sackett DL, Rosenberg WM, Gray JA, Haynes RB, Richardson WS.
urements are scarce. In an excercise program for sedentary
Evidence based medicine: what it is and what it isn’t. BMJ
women, Gademan and her group were able to compare self- 1996;312:71e2.
reported physical activity with VO2max. No significant as- [2] Feinstein AR. Clinical judgment. Baltimore, MD: Williams & Wil-
sociation was found, and therefore, according to the authors, kins; 1967.
physical activity and VO2max represent different aspects of [3] Knottnerus JA, Buntinx F, editors. The Evidence Base of Clinical Di-
health in this group and cannot be used interchangeably. agnosis: Theory and Methods of Diagnostic Research. 2nd ed. Oxford:
Blackwell Publishing; 2009.
Cabrero-Garcıa and co-authors used self-rated health [4] Kotz D, Spigt M, Arts IC, Crutzen R, Viechtbauer W. The stepped
(SRH) as a reference to evaluate the Global Activity Limi- wedge design does not inherently have more power than a cluster
tation Index (GALI) in the context of a national survey randomized controlled trial. J Clin Epidemiol 2013;66:1059e60.