Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Frontiers_Rinchuse.

qxd 5/7/08 2:18 PM Page 167

Frontiers IN CLINICAL RESEARCH

Donald J. Rinchuse, DMD,


MS, MDS, PhD1 DECONSTRUCTING EVIDENCE IN
Daniel J. Rinchuse, DMD, ORTHODONTICS: MAKING SENSE OF
MS, MDS, PhD1

Sanjivan Kandasamy,
SYSTEMATIC REVIEWS, RANDOMIZED
BDSc, BScDent, DocClin-
Dent, MOrthRCS2
CLINICAL TRIALS, AND META-ANALYSES
Marc B. Ackerman, DMD3 Much has been written in support of systematic reviews and the ran-
domized clinical trials and meta-analyses upon which they are based.
Clearly, the medical profession (as opposed to the dental profession)
has been the leader in publishing the benefits of systematic reviews
over the traditional, qualitative narrative reviews. At the same time,
the medical profession also appears to be ahead of the dental profes-
sion in recognizing the limitations of such reviews. That said, there are
a number of inherent problems with systematic reviews, as well as the
randomized clinical trials and meta-analyses that back them up. To
better facilitate evidence-based decision making, this article discusses
the shortcomings of systematic reviews so that practitioners are fully
aware of their drawbacks, as well as their benefits. World J Orthod
2008;9:167–176.

1Clinical Professor, Department of


Orthodontics and Dentofacial Ortho-
pedics, University of Pittsburgh
t has been said that the 20th century published each year. Therefore, it takes
School of Dental Medicine, Pitts-
burgh, Pennsylvania, USA.
2Senior Lecturer, Department of
I was the age of science, while the 21st
century ushered in the age of evidence.1
a systematic approach to summarize
the large volume of literature. Further,
Orthodontics, School of Dentistry, Evidence-based dentistry is defined as the evidence-based approach argues
University of Western Australia,
Perth, Australia; private orthodontic
“an approach to oral health care that that lower-level evidence, such as expert
practice, Perth, Australia. requires the judicious integration of sys- opinions and even case studies, is “not
3Associate Professor and Director, temic assessments of clinically relevant sufficiently strong or credible evidence
Fellowship in Orthodontic Clinical scientific evidence relating to the in the decision-making process.”2 The
Research, Jacksonville University patient’s oral and medical condition and best evidence is considered that which
School of Orthodontics, Jacksonville,
Florida, USA.
history, with the dentist’s clinical exper- is the “least biased in terms of design,
tise and patient’s treatment needs and analyses, or interpretation.”2 The least
Correspondence preference.”2 The ultimate goal of evi- biased evidence is said to come from
Dr Sanjivan Kandasamy dence-based dentistry is for clinicians to prospective, longitudinal, double-blind
University of Western Australia
provide the best possible care to studies utilizing randomized clinical
17 Monash Avenue
Nedlands, Western Australia 6009 patients. It is difficult for clinicians to trials; this is considered the gold stan-
Australia read and assimilate into their practice dard by which new treatments are best
E-mail: sanj@kandasamy.com.au the information from the many articles evaluated.

COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO 167
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 168

FRONTIERS IN CLINICAL RESEARCH WORLD JOURNAL OF ORTHODONTICS

Fig 1 Hierarchy of evidence with the high-


HIERARCHY OF EVIDENCE est and strongest evidence at the “top.”
(Based on the quality of research) RCTs (randomized clinical trials).

Systematic reviews based on RCTs (highest level) Evidence-based


Meta-analyses
RCTs
Non-RCTs
Cohort studies
Case-control studies
Cross-over studies
Cross-sectional studies
Case studies
Consensus opinion of experts
Anecdotal reports and testimonies Experience-based

According to the evidence-based para- using various scientific databases such


digm, there exists a hierarchy of evi- as PubMed. Nonetheless, in the litera-
dence with systematic reviews utilizing ture review in model level No. 2, evi-
randomized clinical trials at the top (ie, dence is in a narrative format and the
the best); the experience-based view that interpretations and evaluation of the arti-
includes testimonies and anecdotal cles are qualitative. Model level No. 3 is
reports is at the lowest level 2 (Fig 1). the highest level of evidence and
There are 3 hierarchical models for evi- involves a systematic review of the litera-
dence-based decision-making as sug- ture that incorporates an exhaustive
gested by Ismail and Bader.2 At the low- search for the best scientific evidence
est level is model level No. 1, which (eg, PubMed, Medline, Cochrane Library),
consists of evidence derived solely from followed by the application of a quantita-
the practitioner’s experience. Although tive, statistical analysis: the meta-analy-
model level No. 1 is considered within sis. Again, the best scientific evidence is
the framework of the evidence-based considered to be derived from random-
dentistry paradigm, it is really not an evi- ized clinical trials.2
dentiary approach. Model level No. 2 is According to Shekelle et al,3 there is a
the next highest level and is evidence 6-level hierarchy of evidence:
gleaned from both the clinical experience
of the practitioner, but more importantly, Ia: Evidence from systematic reviews of
knowledge and information acquired randomized clinical trials
from a review of selected published arti- Ib: Evidence from at least 1 randomized
cles and research. The literature search clinical trial
in the model No. 2 approach is not IIa: Evidence from at least 1 controlled
exhaustive and may be as limited as a study with randomization
cursory reading of several publications IIb: Evidence from at least 1 other type of
readily available to the clinician. A major quasiexperimental study
limitation of this approach is that it III: Evidence from nonexperimental
requires a constant search for evidence. descriptive studies, such as compara-
Of note, there can be quite a range and tive studies, correlation studies, and
disparity in the quality of the evidence in case-control studies
model level No. 2 clinical decision-mak- IV: Evidence from expert committee
ing. For instance, a clinician can review reports or opinions or clinical experi-
only several articles while a more diligent ence of respected authorities.
practitioner may do an online search

168 COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 169

VOLUME 9, NUMBER 2, 2008 RINCHUSE ET AL

In an attempt to facilitate evidence- clinical trials, meta-analyses, and system-


based decision making, guidelines such atic reviews. For example, Dietrich 7
as CONSORT (Consolidated Standards of argued that “the conduct of high-quality
Reporting Trials), QUOROM (Quality of RCTs may not be a realistic goal for many
Reporting of Meta-analyses), and MOOSE dental interventions and procedures.
(Meta-analysis of observational studies in Well-designed observational studies may
epidemiology) were established by a be a viable, cost-efficient alternative.”
team of dedicated journal editors, epi- Likewise, Keim,8 editor of the Journal of
demiologists, and statisticians. 4,5 For Clinical Orthodontics, as well as Williams
sure, evidence-based dentistry does not and Garner,9 have challenged the obsti-
suggest that orthodontists consult only nate adherence to evidence-based sys-
systematic reviews and papers that tematic reviews because of the very
report randomized clinical trials. There restrictive inclusion criteria used while
still exists in the evidence-based para- ignoring other valid sources of knowl-
digm a useful place for traditional, narra- edge and clinical experience. As noted by
tive, qualitative literature reviews. Johnston,10 “ . . . I doubt that we need to
The value of systematic reviews and wait 20 years for an ostensibly perfect
randomized clinical trials have been well- answer from a randomized clinical trial
documented.2,3,5 And the present authors [RCT] that in the end will be diminished
clearly support the modern-day evidence- by sample attrition, ‘Hawthorne effect,’
based paradigm including systematic lack of blinding, ‘data peeking,’ and other
reviews. Nonetheless, there are some shortcomings that are intrinsic to ortho-
inherent problems and limitations of dontic versions of the medical trials we
some systematic reviews and randomized seek to imitate.”
clinical trials that are not as well-publi- There are other limitations of systematic
cized and recognized in dentistry as they reviews. First, Flores-Mir et al11 found that
are in medicine. The purpose of this arti- the search and selection methods of cur-
cle is to point out and discuss some of the rent systematic reviews in orthodontics (ie,
shortcomings of systematic reviews, ran- from 2000 to 2004) are limited in that key
domized clinical trials, and meta-analyses methodological components are frequently
(fully aware and appreciative of the bene- absent or not appropriately described. For
fits of such reviews). A recurrent theme of the 16 orthodontic systematic reviews for
this paper is that orthodontic practitioners this time period, many failed to search
and journal editors alike should not erro- more than Medline (56%), failed to docu-
neously accept as holy writ the conclu- ment the database names and search
sions of all systematic reviews. dates (37%), failed to document the search
strategy (62%), did not use several experts
to select studies (75%), and did not include
SYSTEMATIC REVIEWS AND all languages (81%).11
RANDOMIZED CLINICAL TRIALS This same research group also
assessed dental systematic reviews pub-
A systematic review is a critical review of lished for time periods between January
primary studies that involves the selec- 1, 2000, and July 14, 2005, and found
tion of only those studies that have used that many failed to search more than
explicit and reproducible methods and Medline (20%), document the search
then applies an appropriate statistical strategy (20%), use multiple reviewers for
technique that quantitatively appraises selecting studies (25%), and include all
the chosen studies.6 As previously men- languages (39%).12 On the plus side, in
tioned, we ardently support science and 2005, most systematic reviews docu-
the current evidence-based decision- mented database names and search
making paradigm, including randomized dates (90%), used electronic search
clinical trials and systematic reviews. terms (95%) and appropriate inclusion-
With this said, we are also very much exclusion criteria (95%), and employed
aware of the limitations of randomized secondary searches (100%). Parentheti-

COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO 169
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 170

FRONTIERS IN CLINICAL RESEARCH WORLD JOURNAL OF ORTHODONTICS

cally, search and selection methods were to a systematic review20 dealing with the
best for the most current (at the time of mandibular effects produced by func-
this writing) year of 2005.12 tional appliances in the treatment of
Next, by asking general and broad Class II malocclusion. Johnston, a critic
questions, systematic reviews often pro- of the proposition that functional appli-
duce results and conclusions with ques- ances can “grow mandibles,” 1 0 , 2 0 , 2 1
tionable validity: “Poorly focused questions offered this critique: inappropriate inclu-
in systematic reviews lead to unclear deci- sion and exclusion criteria; bias in the
sions about what research to include and weighting of articles (a tendency of the
how to summarize it.”13 Part of the prob- authors to select their own studies, even
lem is that some systematic reviews, or when the finding may not be statistically
the research studies they are based upon, significant); duplication of a previous sys-
have not accounted for confounders that tematic review that already had
may preclude appropriate interpretations. addressed the question; and so forth.
For example, a recent systematic review of Poor systematic reviews will invariably
the effect of rapid maxillary expansion by lead to inaccurate conclusions that will
Lagravere et al14 did not identify the types then negatively impact upon clinical prac-
of malocclusions (phenotypes) the rapid tice. Further, when systematic reviews
maxillary expansion was used to treat; all are based on randomized clinical trials
malocclusions were “lumped” together. that are also poorly defined and directed,
Was rapid maxillary expansion used for the error and impact on clinical practice
crowding or Class II correction in the multiplies. There are perhaps more nega-
absence of a posterior crossbite as advo- tive ramifications from poor systematic
cated by McNamara?15–17 Or, was rapid reviews in medicine than in dentistry,
maxillary expansion used only when a pos- given that many of their clinical ques-
terior crossbite was present as suggested tions are life-threatening. Moreover, the
by Gianelly?18 Parenthetically, Gianelly is results and conclusions of medical sys-
baffled by transverse maxillary expansion tematic reviews vis-a-vis dental reviews
in the absence of a posterior crossbite.19 are more likely to be thoroughly scruti-
Thus, how much clinical validity can this nized and the research they are based
review have? Asking the right questions is upon often replicated using larger trials.
of utmost importance in systematic With regard to the problem in medicine,
reviews. Even the authors14 of this exhaus- Rees and Ebrahim22 offered this caveat
tive systematic review were aware of the for basing medical decisions (ie, cardiol-
limitations of their review and gave this ogy) on initial and preliminary systematic
caveat and disclaimer as to the value of reviews utilizing rather small trials: “In
what they concluded: “All of the studies general, smaller trials are completed
had methodological deficiencies. Lack of before larger trials, and if meta-analyzed,
intra- and interexaminer agreement report- may show a beneficial effect, making
ing was common; only 1 study clearly subsequent large clinical trials unneces-
reported these values. The report of con- sary and even unethical. The results of
founders and dropouts in the sample meta-analyses of small trials may also be
analysis, as well as descriptions of the more applicable at the population level
inclusion criteria, was absent in the major- than a single large trial, in terms of allow-
ity of studies. Therefore, the scientific evi- ing stratified analyses by patient type,
dence we found in this meta-analysis severity of disease, age, and so on. Even
should be interpreted carefully. Clinicians in cardiology, where very large random-
will have to consider their experience, the ized controlled trials have been per-
opinion of experts, and the limited evi- formed, it is unlikely that single large tri-
dence on RME [rapid maxillary expansion] als of all potentially useful treatments
to decide whether to use this treatment on will be undertaken.”
patients.” Furthermore, Richards13 believed that
Another example of the ambiguity and many systematic reviews in dentistry are
controversy regarding systematic reviews compromised because of poorly defined
was pointed out by Johnston10 in regard outcomes measures with surrogate,

170 COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 171

VOLUME 9, NUMBER 2, 2008 RINCHUSE ET AL

rather than true, endpoints. True end- attrition. Not all subjects comply with the
points are “tangible to the patients and regimen to which they are assigned, and
directly measure how a patient feels, for studies that require long follow-up peri-
functions or survives (eg, tooth loss or ods, there is a natural tendency for a high
pain or oral health-related quality-of-life dropout rate. Moreover, because it takes
measurements).”13 Surrogate endpoints such a long time to complete prospective
are “intangible to the patient, such as trials, by the time the studies are com-
changes in probing attachment level or plete, the appliance and/or procedure
gingival crevicular fluid level.”13 Richards that was investigated may not even be
wrote that dentistry tends to focus on considered and/or utilized in practice (ie,
surrogate endpoints or outcomes rather the question may be obsolete).27
than on true endpoints dealing with oral Although the goal of conducting high-
health-related quality of life measures.13 quality randomized clinical trials is noble,
the reality is that many orthodontic clini-
cal research questions are amenable to
MORE ON RANDOMIZED well-designed and cost-effective, obser-
CLINICAL TRIALS vational (cross-sectional) studies such as
cohort or case-control studies. 28 Fein-
Randomized clinical trials24–26 are some- stein stated: “Although I have great rever-
times also referred to as randomized ence, admiration, and devotion for ran-
controlled trials.7 There is certainly a dif- domized trials, they have a limited
ference between a randomized clinical spectrum of application, and they cannot
trial and a randomized controlled trial. be used to answer all the clinical ques-
For instance, the Class II early treatment tions that need to be answered. For rea-
study at the University of Pennsylvania by sons of ethics, feasibility, or costs, we
Gharfari et al26 was said to be a random- simply will not be able to use randomized
ized clinical trial (and appropriately so) trials to answer all the cause-effect ques-
and could only be described as a clinical tions that arise about both therapy and
trial, not a controlled trial. Although the etiology in clinical medicine.”28
trial was prospective and the subjects As mentioned, there are ethical con-
were randomly assigned to 1 of 2 treat- cerns involving randomized clinical trials
ment groups, there was no control group. studies using human subjects. In general,
That is, subjects were not assigned to a there is the moral foundation that health
no-treatment (control) group. care providers should not disadvantage
Randomized controlled trials are con- subjects on account of their research par-
sidered the best type of studies because ticipation. 7 There must be a genuine
they are, theoretically, not susceptible to uncertainty on behalf of the expert ortho-
various biases; the distribution of charac- dontic community concerning the merits
teristics of all known and unknown vari- of each trial arm (clinical equipoise); oth-
ables are identical among the study erwise, obtaining proper informed consent
groups.9 Conversely, retrospective stud- becomes an issue. 7 More importantly,
ies, the traditional method of assessing when there is no clinical equipoise, there
the efficacy of orthodontic treatments, may be an additional ethical concern with
have been criticized because of selection randomized clinical trials due to random-
bias, inadequate sample size, lack ization into experimental and control
of contemporaneous controls, poor groups whereby subjects in the control
research design, and so on.24 group may be disadvantaged significantly
by not receiving the more appropriate
treatment in the long term (eg, extraction
Limitations of Randomized versus nonextraction, orthodontics versus
Clinical Trials surgery, and long versus short treat-
ments). In addition, a researcher cannot
As in other types of prospective studies, ethically create a disease or disorder in 1
randomized clinical trials are susceptible group of subjects, study the effects of the
to biases of compliance and long-term disease (and several treatment modali-
COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO 171
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 172

FRONTIERS IN CLINICAL RESEARCH WORLD JOURNAL OF ORTHODONTICS

ties), and compare it with a sample that grown any more than the given morpho-
does not have the disease. genetic pattern of the individual subjects
Importantly, there are many research and that 2-phase growth-modification
projects (and questions) that are not treatments have little effect above and
favorable to a true experiment design, ie, beyond what would be expected of more
longitudinal, prospective, double-blind (or conventional 1-phase treatments.21,26,30–35
at least single blind), randomized clinical A strength of early treatment random-
trial research protocol. For instance, it is ized clinical trials is their assignment of
not possible or logical to consider a treatments at random in order to mini-
prospective, randomized clinical trial mize bias. Proponents of the early treat-
design for descriptive (epidemiological) ment randomized clinical trials also argue
studies that do not involve issues of the view that whether 1 particular person
cause/effect. Descriptive studies in medi- or someone else could have assigned the
cine investigate the incidence or preva- treatments more appropriately in isolated
lence of a certain disease state, patho- instances or whether a Twin Block,
logic condition, diagnostic factor, health Herbst, or a Bionator was used to posture
variable in a population, and so forth. In the mandible in a forward position for it
orthodontics, an example would be a sur- to grow more is beside the point. The ran-
vey of the frequency of individuals in a domized clinical trials were designed to
certain location who possess the various test for an effect. Had 1 treatment or
Angle’s static occlusion types (ie, normal, another been able, on average, to do
Class I, II, III)29 to address perhaps the ulti- something that the others could not, the
mate question as to which of the various randomized clinical trials certainly would
Angle types does this population possess. have had the power to detect it. Had an
effect been detected, everyone could
then apply the proper treatment in the
Class II Early Treatment best interest of our patients.
Randomized Clinical Trials The individuals, or groups, within the
orthodontic profession who still dismiss
Arguably, the best studies conducted in the results of these current Class II early
orthodontics are the Class II early treat- treatment randomized clinical trials base
ment, multicenter randomized clinical tri- their criticisms on the following: ambigu-
als. The orthodontic profession has ous definition of Class II; objectives of
clearly recognized and embraced the phase I treatments not clearly defined as
value and strengths of these trials. How- would be in private practice (university
ever, these trials have some inherent setting); did not use headgear long
weaknesses. Certainly, we believe the enough; evaluated Bionator and Twin
strengths of these studies far outweigh Block (as opposed Frankel or others) as
their shortcomings. The strengths of the the choice of functional appliances;
early Class II randomized clinical trials could have used a more contemporary
will be discussed first. The Class II early appliance or method including fixed
treatment randomized clinical trials were appliances; no retention between
aimed essentially at answering 2 contro- phases; lack of blinding or control for
versies once and for all; the ability of Hawthorne effect; too many uncontrolled
functional appliances to modify dentofa- variables; large costs; and so forth.36,37
cial growth and the effectiveness of Meikle27 pointed out some additional
1- versus 2-phase treatments of Class II limitations of randomized clinical trials
malocclusions (overjets greater than 7 that have evaluated the efficacy of func-
mm). They generally found that 2-phase tional appliances: “Given the variability in
treatments are less beneficial than 1 the timing, magnitude, and duration of
comprehensive phase. 23,25,26,30–35 The pubertal dentofacial growth, differing lev-
early treatment randomized clinical trials, els of motivation and patient compliance,
as well as the compelling evidence from the inherent inaccuracy of cephalometry
the literature, seem to be converging on and the questionable validity of the mea-
the conclusion that mandibles cannot be surements themselves used to quantita-

172 COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 173

VOLUME 9, NUMBER 2, 2008 RINCHUSE ET AL

tive change, it is not surprising that the skewed and uninterpretable results. Care-
conclusions have not been as clear-cut fully designed, planned, and statistically
as anticipated.” analyzed retrospective studies aimed at
Furthermore, regarding the issue of controlling or minimizing biases (selection,
what exactly constitutes a Class II, division detection/exclusion, proficiency, and
1 malocclusion as used in the early treat- transfer) have been shown to yield highly
ment Class II randomized clinical trials accurate results.40,42 Methods of control-
and whether all Class II malocclusions are ling for these biases have been clearly
alike, Darendeliler37 made several com- outlined by Johnston.42
ments. The early treatment RCTs report
the average effect of treatment on a Class
II (1) malocclusion that result from a com- META-ANALYSES
bination of different deformities or mor-
phologies (phenotypes).37 The problem is: A meta-analysis is a mathematical and
“Not all patients with an increased overjet quantitative (statistical) synthesis of the
or increased ANB angle have the same results of 2 or more primary studies that
malocclusion.” For example, a Class II (1) address the same hypothesis or topic in
malocclusion with an overjet of greater the same way.6 Part of the responsibility
than 7 mm may be due to maxillary dental of the meta-analysis is to tabulate rele-
protrusion, mandibular retrognathism, vant information on “the inclusion crite-
maxillary prognathism, retroclined ria, sample size, baseline patient charac-
mandibular anterior teeth, a combination teristics, withdrawal rate, and results of
thereof, and so on. It is therefore argued primary and secondary end points of all
that a stratified, refined sample selection the studies included.” 6 It is important
of Class II (1) malocclusions would have that the methods used for the review are
provided more accurate data.37 reliable, valid, and well-characterized.
All considered, we believe the early Meta-analyses are by all accounts supe-
treatment Class II randomized clinical tri- rior to qualitatively based evaluations of
als did answer the questions they initially numerous studies. The preliminary
proposed. That is, we cannot grow aspects of the meta-analysis (prior to
mandibles and that in general, 2-phase applying the actual statistical test), how-
treatments are less efficient that 1 com- ever, are subjective (even though there
prehensive phase (at least under the are certain rules and guidelines); there is
conditions of the random clinical trial the subjective judgment in deciding
research design).23,25,26,30–35 which studies to include.43
The issue does not really appear to be Eysenck44 lists a number of problems
whether randomized clinical trials are that are inherent to meta-analyses:
capable of addressing various controver- “regressions are often nonlinear; effects
sies in dentistry and orthodontics: They are often multivariate rather than univari-
are. The issue is whether we can justify ate; coverage can be restricted; bad stud-
the large costs and time associated with ies may be included; the data summa-
such trials when simple, cost-effective ret- rized may not be homogeneous; grouping
rospective or observational cohort studies different causal factors may lead to mean-
may arguably reveal the same results.38–43 ingless estimates of effects; and the the-
In a specialty for which only limited ory-directed approach may obscure dis-
research funding is available, we must crepancies.” Ultimately, Eysenck
seek not only evidence but also frugality. concluded, “Meta-analysis may not be the
Retrospective studies are quick, cost- best method for studying the diversity of
effective, and ethically unambiguous.40 A fields for which it has been used.”44
major problem with retrospective studies Let us now discuss several additional
is achieving an approximation of a bias- limitations of meta-analysis as described
free sample, which is a prominent feature by Eysenck.44 First, there is the possibility
in prospective trials. Failure to control for that the data used in meta-analyses are
these biases when comparing samples not homogeneous. Homogeneity is an
retrospectively will invariably lead to important requisite of meta-analyses.
COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO 173
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 174

FRONTIERS IN CLINICAL RESEARCH WORLD JOURNAL OF ORTHODONTICS

Homogeneity means: “The results of each statistical test can overcome and rectify
individual trial are mathematically com- the methodological shortcomings of
patible with the results of any of the oth- poorly designed primary studies. In sum-
ers.”6 It implies that treatments, patients, mary, there is no doubt that the meta-
and end points must be similar or at least analysis has its place in the evidence-
comparable. Clinical heterogeneity is a based dentistry paradigm and is an
problem for meta-analyses. Greenhalgh6 integral part of a systematic review; how-
cited a poignant example of a fallacious ever, the validity of its findings is greatly
meta-analysis based partly on heterogene- dependent on the quality of the individual
ity: “ . . . Reservations about meta-analysis studies incorporated into the analysis.
are borne out in the infamously discred-
ited meta-analysis which showed (wrongly)
that giving intravenous magnesium to CONCLUSIONS
people who had had heart attacks was
beneficial. A subsequent megatrial involv- Ackerman 46 appropriately stated, “The
ing 58,000 patients (ISIS-4) failed to find challenge facing orthodontists in the 21st
any benefit, and the meta-analysts’ mis- century is the need to integrate the
leading conclusions were subsequently accrued scientific evidence into clinical
explained in terms of publication bias, orthodontic practice.” However, exactly
methodological weaknesses in the what constitutes the evidence is some-
smaller trials, and clinical heterogeneity.” what debatable. In orthodontic practice,
Altman45 believed that the meta-analy- clinical judgment entails an integration of
ses (and systematic reviews) of prognos- clinical experience and systematic
tic studies are difficult. Prognostic stud- assessment of relevant scientific evi-
ies include clinical studies of variables dence in the context of the patient’s
predictive of future events as well as epi- orthodontic condition, treatment need
demiological studies of etiological risk and preference.47 Currently, systematic
factors. Authors often have concluded reviews of the literature favoring studies
that a meaningful meta-analysis for prog- utilizing randomized clinical trials are con-
nostic studies is not possible due to a set sidered the gold standard and the top of
of studies being too diverse and/or too the hierarchical pyramid of evidence.
weak. That is, a high proportion of prog- Although systematic reviews are the best
nostic studies are methodologically poor. available source of evidence, a critical
Altman45 stated, “The poor quality of the appraisal and a cautious interpretation of
published literature is a strong argument the results are essential. The notion that
in favor of systematic reviews but also an systematic reviews and randomized clini-
argument against formal meta-analysis . . cal trials are without limitations is a
. Meta-analysis of prognostic studies rather naïve view and detrimental to
using individual data from patients may appropriate clinical decision-making.
overcome many of these difficulties.” Extrapolation of findings from systematic
When there are serious methodologi- reviews into clinical practice must con-
cal difficulties in the primary studies, it is sider the methodological quality and
generally impossible to conduct a sensi- external validity of the studies used in
ble meta-analysis without first having such reviews. Importantly, well-designed,
access to the individual patients’ raw planned, and statistically analyzed retro-
data. 45 The problem here may be that spective studies (observational) aimed at
many of the prognostic studies are obser- controlling or minimizing biases (selec-
vational rather than experimental (ran- tion, detection/exclusion, proficiency, and
domized clinical trials) and the individual transfer) may also yield valuable informa-
raw data are not readily available. Fur- tion for the clinician. Finally, it is inappro-
ther, meta-analyses are sometimes used priate in the age of evidence for health
incorrectly to recover something from care practitioners to totally abrogate their
poorly designed studies; studies with responsibility to critically evaluate the evi-
insufficient statistical power and studies dence, including the quality of systematic
resulting in apparent contradictions.45 No reviews.

174 COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 175

VOLUME 9, NUMBER 2, 2008 RINCHUSE ET AL

REFERENCES 18. Gianelly A. Rapid palatal expansion in the


absence of crossbites: Added value? Am J
Orthod Dentofacial Orthop 2003;124:362–365.
1. Turpin DL. Consensus builds for evidence-
19. Gianelly A. Dr. Lawrence White, JCO editor,
based methods. Am J Orthod Dentofacial
interviews Dr. Anthony Gianelly on current
Orthop 2004;125:1–2.
issues in orthodontics. J Clin Orthod 1996;30:
2. Ismail AI, Bader JD. Evidence-based dentistry in
439–445.
clinical practice. J Am Dent Assoc 2004;135:
20. Cozza P, Baccetti T, Franchi L, De Toffol L,
78–83.
McNamara Jr JA. Mandibular changes pro-
3. Shekelle PG, Woolf SH, Eccles M, Grimshaw J.
duced by functional appliances in Class II mal-
Clinical guidelines: Developing guidelines. Br
occlusion: A systematic review. Am J Orthod
Med J 1999;318:593–596.
Dentofacial Orthop 2006:129:599.e1–599.e12.
4. Turpin DL. CONSORT and QUOROM guidelines
21. Johnston Jr LE. If wishes were horses: Func-
for reporting randomized clinical trials and sys-
tional appliance and growth modification. Prog
tematic reviews. Am J Orthod Dentofacial
Orthod 2005;6:36–47.
Orthop 2005;128:681–686.
22. Rees K, Ebrahim S. Promises and problems of
5. Stroup D, Berlin J, Morton S, et al. Meta-analy-
systematic reviews. Heart Drug
sis of observational studies in epidemiology:
2001;1:247–248.
A proposal for reporting. JAMA 2000;283:
23. Keeling SD, Wheeler TT, King GJ, et al. Antero-
2008–2012.
posterior skeletal and dental changes after
6. Greenhalgh T. How to read a paper: Papers
early Class II treatment with bionator and head-
that summarize other papers (systematic
gear. Am J Orthod Dentofacial Orthop 1998;
reviews and meta-analyses) Brit Med J
113:40–50.
1997;315: 672–675.
24. Sadowsky PL. Craniofacial growth and the tim-
7. Dietrich T. Commentary: Evidence-based den-
ing of treatment. Am J Orthod Dentofacial
tistry and the absence of evidence from ran-
Orthop 1998;113:19–23.
domized controlled trials for common dental
25. Tulloch JFC, Phillips C, Proffit WR. Benefit of
procedures. In: Pena RA (ed). Dental Abstracts
early Class II treatment: Progress report of a
2006;51:192–193.
two-phase randomized clinical trial. Am J
8. Keim RG. The weight of evidence. J Clin Orthod
Orthod Dentofacial Orthop 1998;113:62–72.
2004;38:121–122.
26. Ghafari J, Shaffer S, Jacobsson-Hunt U,
9. Williams DDR, Garner J. The case against “the
Markowitz DL, Laster LL. Headgear versus
evidence:” A different perspective on evidence-
function regulator in the early treatment of
based medicine. Br J Psychol 2002;180:8–12.
Class II, division 1 malocclusion: Randomized
10. Johnston Jr LE. Commentary. Am J Orthod
clinical trial. Am J Orthod Dentofacial Orthop
Dentofacial Orthop (online only) 2006:
1998; 113:51–61.
129(5):e1–e4. [Cozza P, Baccetti T, Franchi L,
27. Meikle MC. What do prospective randomized
De Toffol L, McNamara Jr JA. Mandibular
clinical trials tell us about the treatment of
changes produced by functional appliances in
class II malocclusions? A personal viewpoint.
Class II malocclusion: A systematic review. Am
Eur J Orthod 2005;27:105–114.
J Orthod Dentofacial Orthop (online only)
28. Feinstein AR. The clinician as scientist. In: PS
2006:129:599.e1–599.e12.]
Vig, KA Ribbens (eds). Science and Clinical
11. Flores-Mir C, Major MP, Major PW. Search and
Judgment in Orthodontics. Monograph 19-
selection methodology of systematic reviews in
Craniofacial Growth Series: Center for Human
orthodontics (2000-2004). Am J Orthod Dento-
Growth and Development. Ann Arbor: The Uni-
facial Orthop 2006;130:214–217.
versity of Michigan, 1986.
12. Major MP, Major PW, Flores-Mir C. An evalua-
29. Angle EH. Classification of malocclusion. Den-
tion of search and selection methods used in
tal Cosmos 1899;41:248–264.
dental systematic reviews published in English.
30. O’Brien K, Wright J, Conboy F, et al. Effective-
J Am Dent Assoc 2006;137:1252–1257.
ness of early orthodontic treatment with the
13. Richards D. Outcomes, what outcomes? Evid
twin-block appliance: A multicenter, random-
Based Dent 2005;6:1.
ized, controlled trial. Part 1: Dental and skele-
14. Lagravere MO, Heo G, Major PW, Flores-Mir C.
tal effects. Am J Orthod Dentofacial Orthop
Meta-analysis of immediate changes with rapid
2003;124:234–243.
maxillary expansion treatment. J Am Dent
31. O’Brien K, Wright J, Conboy F, et al. Effective-
Assoc 2006;137:44–53.
ness of early orthodontic treatment with the
15. McNamara JA. Maxillary transverse deficiency.
twin-block appliance: A multicenter, random-
Am J Orthod Dentofacial Orthop 2000;117:
ized, controlled trial. Part 2: Psychosocial
567–570.
effects. Am J Orthod Dentofacial Orthop 2003;
16. McNamara JA Jr. Early intervention in the trans-
124:488–494.
verse dimension: Is it worth the effort? Am J
32. O’Brien K, Wright J, Conboy F, et al. Effective-
Orthod Dentofacial Orthop 2002;121:572–574.
ness of treatment for class II malocclusion with
17. McNamara JA Jr, Brudon WL. Orthodontics and
the herbst or twin-block appliances: A random-
dentofacial orthopedics. Ann Arbor: Needham
ized controlled trial. Am J Orthod Dentofacial
Press, 2001.
Orthop 2003;124:128–137.

COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO 175
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER
Frontiers_Rinchuse.qxd 5/7/08 2:18 PM Page 176

FRONTIERS IN CLINICAL RESEARCH

33. Tulloch JFC, Phillips C, Koch G, Proffit WR. The


effect of early treatment on skeletal pattern in
class II malocclusion: A randomized clinical
trial. Am J Orthod Dentofacial Orthop 1997;
111:391–400.
34. Tulloch JFC, Proffit WR, Phillips C. Outcomes in
a 2-phase randomized clinical trial of early
Class II treatment. Am J Orthod Dentofacial
Orthop 2004;125:657–667.
35. Koroluk LD, Tulloch JFC, Phillips C. Incisor Jarabak Endowed Professorship
trauma and early treatment for Class II division
1 malocclusion. Am J Orthod Dentofacial The Joseph R. and Louise Ada Jarabak Professorship is a full-time faculty
Orthop 2003;123:117–125. position endowed at the full Professor level. The specified purpose of the
36. Dugoni S, Lee JS, Varela J, Dugoni A. Early Professorship is to enhance and further orthodontic education; stimulate
mixed dentition treatment: Postretention evalu- research in anatomy, physiology, and the function of the stomatognathic
ation of stability and relapse. Angle Orthod system; continue research in anatomy, physiology, and the function of the
1995;65:311–320. stomatognathic system; continue research in TMJ disorders; and stimulate
37. Darendeliler MA. Validity of randomized clinical a continued development of orthodontic teachers. All successful candi-
trials in evaluating the outcomes of class II dates must either have a DDS/DMD degree, clinical specialty training in
treatment. Semin Orthod 2006;12:67–79. orthodontics and eligibility for Indiana dental licensure. In addition the can-
38. Johnston LE Jr. A comparative analysis of Class didates must either have: (1) completed an NIH Fellowship in biochemistry,
II treatments. In: Vig PS, Ribbens KA (eds). Sci- molecular biology, or immunology with a PhD in anatomy, physiology, or
ence and Clinical Judgment in Orthodontics. similar qualifications; (2) a MS or PhD in biophysics or a basic science
Ann Arbor: Center for Human Growth and equivalently related to orthodontics. Development of an independent
Development, 1986. research program is required. Opportunity for intramural or extramural pri-
39. Luppannapornlarp S, Johnston LE Jr. The vate practice is available. The position will remain open until filled.
effects of premolar-extraction treatment:
A long-term comparison of outcomes in “clear-
cut” extraction and nonextraction Class II Interested individuals should email an electronic application, an electronic
patients. Angle Orthod 1993;63:257–272. curriculum vitae and three electronic reference letters to:
40. Livieratos FA, Johnston LE Jr. A comparison of
Dr Jeffrey A. Dean
one- and two-stage non-extraction alternatives
Executive Associate Dean
in matched Class II samples. Am J Orthod
Dentofacial Orthop 1995;108:118–131. Indiana University School of Dentistry
41. Paquette DE, Beattie JR, Johnston LE Jr. A long- 1121 W. Michigan Street
term comparison of nonextraction and premo- Indianapolis, IN 46202-5186
lar-extraction edgewise therapy in “borderline” jadean1@iupui.edu
Class II patients. Am J Orthod Dentofacial
Indiana University is an Equal Opportunity/Affirmative Action Employer.
Orthop 1992;102:1–14.
42. Johnston LE Jr. Moving forward by looking
back: “Retrospective” clinical studies. J Orthod
2002;29:221–226.
43. Concato J, Shah N, Horwitz RI. Randomized
controlled trials, observational studies, and the
hierarchy of research designs. N Engl J Med
2000;342:1887–1892.
44. Eysenck HJ. Systematic reviews: Meta-analysis
and its problems. Br Med J 1994;309:789–792.
45. Altman DG. Systematic reviews in health care:
Systematic reviews of evaluations of prognos-
tic variables. Br Med J 2001;323:224–228.
46. Ackerman M. Evidence-based orthodontics for
the 21st century. J Am Dent Assoc 2004;135:
162-167.
47. Ackerman JL, Kean MR, Ackerman MB. Evi-
dence-bolstered orthodontics. Aust Orthod J
2006;22:69–70.

176 COPYRIGHT © 2008 BY QUINTESSENCE PUBLISHING CO, INC. PRINTING OF THIS DOCUMENT IS RESTRICTED TO PERSONAL USE ONLY. NO
PART OF THIS ARTICLE MAY BE REPRODUCED OR TRANSMITTED IN ANY FORM WITHOUT WRITTEN PERMISSION FROM THE PUBLISHER

You might also like