Why randomized controlled trials of calcium and vitamin D sometimes fail.

2012·0.17 MB·English

by  Lappe, Joan M.

REVIEW
Dermato-Endocrinology4:2,95–100;April/May/June2012;G2012LandesBioscience
Why randomized controlled trials of calcium
and vitamin D sometimes fail
Joan M. Lappe and Robert P. Heaney*
CreightonUniversity;OsteoporosisResearchCenter;Omaha,NEUSA
Keywords: chronic disease, nutrient dose response, systematic reviews, meta-analyses, design flaws, evidence-based medicine
.
e
c
n
the risk ofthediseases concerned.There are,however, reasons to
The importance of nutrients for promotion of health and e
reject that conclusion.
preventionofdiseasehaslongbeenrecognized.Nonetheless,
i
scientists are still trying to delineate the optimal intakes of There are wellcover 100 RCTs of vitamin D with respect to
various nutrients and their potential benefits for different various healthsendpoints, and several times that number involv-
populations.Tothatend,evidence-basedmedicine(EBM)has ing calciuom as the principal intervention. For the most part, the
beenappliedtothestudyofnutrition.EBMmethodsbasically resuilts for both nutrients fall into just two categories: many of
B
call for the use of randomized controlled trials (RCTs) to the trials are positive, many are null, but almost none is actually
.
establishcausalconnectionbetweentheinterventionandansy negative. And most of the eeffects, when positive, are small. If
particular endpoint.
e a particular interventiontwere, in fact, unrelated to a particular
ThispaperfocusesonproblemsthatariseintheuseofRCTS u
d disease risk, one would expect a more symmetrical distribution
toestablishacausallinkbetweennutrientsandvariousclinical b
endpoints. While many RCTS of calcium andn vitamin D have of results,6 with the majority of the trials being null and a
minority siplit roughly evenly between positive and negative.
been positive, many others have beenanull. In this paper, we r
discuss the reasons why effectivLe nutrient agents may be Howtever, as noted, the preponderance of the evidence tilts
s
found to be ineffective in pa rticular studies, giving examples stronglytowardapositiveresult,andthepurposeofthisreviewis
2 i
of such null results, and focusing on the nearly universadl to examine why, if the agent is in fact efficacious, randomized
failure to consider bi1ological criteria in designing RCTs. Our controlled trials sometimes fail to find the underlying causal

purpose is (1) to0 inform future study design so as tto ensure connection. When RCTs (or observational studies) produce this
o
that releva2nt biological facts are considered and (2) to aid in kind of mixed result, systematic reviews and meta-analyses can
n
the interpretation of the abundant, but often inconsistent oftenhelptodiscernanunderlyingpattern.Byaggregatingseveral

©literature onthis topic. o trials they effectively increase sample size and narrow the range
D of uncertainty around estimates of effect. Accordingly we will
also examine several of the larger reviews concerning these
relationships.
Introduction EBM, in its grading of the evidence in particular papers,
focuses on certain methodological issues which can confound
Enthusiasm for evidence-based medicine (EBM) has resulted in the results of even the best RCTs. These include such features
the extension of its methods to the evaluation of nutrient effects. as inadequate randomization and blinding, excessive losses of
However, as has become increasingly clear, EBM, as applied in participants during the course of a trial, and other such issues of
theevaluationofdrugs,ispoorlysuitedtothestudyofnutrients. unquestioned importance. Typically, the analyses and reviews
The reasons have been discussed in depth elsewhere1-5 and will of EBM admit studies into review, and/or grade studies, mainly
notberevisitedindetailhere.Basicallythosemethodscallforthe on these methodological grounds. They usually presume that all
use of randomized controlled trials (RCTs) to establish nutrient the trials so aggregated measured the same thing, under uniform
effects. The reason is that this design is the only certain way to exposure conditions, in participants of similar nutritional status.
establish a causal connection between an intervention and the Unfortunately that is often not the case. If these reviews are
productionofaparticularendpoint.Theexperiencehasbeenthat conducted by individuals or teams with limited understanding of
RCTs of calcium and vitamin D, while often positive, have the biology concerned, what emerges in the process may be a set
sometimes failed to find the sought for causal link. The most of studies that, while methodologically “pure” are nevertheless
obvious explanation for such failure is that the intervention biologically mixed, if not actually invalid.
concerned is not actually efficacious with respect to the endpoint In addition to biology-related issues, trials may fail for pro-
being studied, i.e., calcium and vitamin D have little to do with babilistic reasons common to any clinical trial. The latter are
generally well known and will be discussed only briefly. The
biologicalissues thatarethebasis ofthefollowingcriteriaare less
*Correspondenceto:RobertP.Heaney;Email:[email protected]
familiar but probably more important, at least for nutrients, and
Submitted:02/15/12;Revised:02/20/12;Accepted:02/22/12
http://dx.doi.org/10.4161/derm.19833 will be our principal focus.
www.landesbioscience.com Dermato-Endocrinology 95
Nutrient-Specific Issues
Before evaluating the major clinical trials and systematic reviews
of calcium and vitamin D, we describe and illustrate certain
features which arelargelyuniquetonutrients andwhichmustbe
factored into the design of nutrient efficacy studies.
Examples of critical biological criteria needed for a RCT to be
informative (and for inclusion into a systematic review) would
include such features as:
N
Use of a single form of the nutrient .
N e
Use of a low exposure control group
N c
Adequacy of dose in the treatment group
N n
Demonstration/documentation of the altered intake/
exposure, i.e., was a “therapeutic” blood level achieved e
N i
Use of a uniform response measure c
N
Optimization of co-nutrient status s
This is not an exhaustive list of relevant biological criteria, but o
it serves to focus attention on some of the reasons why, in a i
B
particularstudy,aneffectiveagentmayseemtobeineffective,and
.
specifically why meta-analyses and systematic reviews of calcsium e
Figure1.Typicalsigmoidcurveshowingphysiologicalresponseas
and vitamin D, when they ignore these criteria, have soemetimes t
afunctionofnutrientiuntakeorstatus.Depictedaretheexpected
been null. We then show, in actual RCTs, why thdese biological responsesfromequalincrementsinintake/status,startingfromalow
b
criteria are important and point out how ignonring them leads to basalintake,andmovingtoprogressivelyhigherstartinglevels.Intake
erroneous conclusions. incrementsi(A–C)produceresponses,(a–c),respectively.Onlyintakesin
a r
Sigmoid response vis-à-vis starting level. A unique feature of thet(B)regionproduceresponseslargeenoughadequatelytotestthe
L
shypothesisthatthenutrientconcernedelicitstheresponseinquestion.
the physiological response to nutrients is the sigmoid character
2 i (CopyrightRobertP.Heaney,MD,2010.Allrightsreserved.Usedwith
of the response. This relationship is depicted in Figure1, whdich permission.).
illustrates a phenomen1on common to virtually all nu trients.4
t
At low intakes (o0r low nutrient status) there is relatively little
o
response; th2e effect increases fairly rapidly over a particular providingatruezeroexposuretotheagentbeingtested.Itshould
n
intake or exposure range; and then at higher intakes the res- be immediately apparent that this approach is not possible with
po©nse plateaus. (This latter feature isoin sharp contrast to drug nutrients. All nutrients are essential (or they wouldn’t be
responses.)Well recognizedexamDplesoftheplateau characteristic nutrients), and a true zero intake is neither feasible nor ethical.
include such familiar phenomena as the treatment of iron The contrasting exposures have to be located somewhere along
deficiency anemia (in which hemoglobin rises to a normal value, the plausible intake range for the nutrient concerned (i.e., the
but then plateaus despite continuing or even increasing iron horizontalaxisinFig.1).Thatlocation,asitturnsout,iscrucially
doses), rehydration in water and electrolyte depleted patients, important.
refillingofthemusclecompartmentbyproteininfaminevictims, Low-exposure control group. Figure1, in addition to dis-
on and on. In all these cases response plateaus once a particular playing the typical response curve, also plots responses for three
physiological norm is reached. Continued dosing may produce identical intake/status increments (doses) differing only in the
harm or toxicity, but usually by mechanisms different from the starting, or control group value. Scenario “A,” starting at a very
one relating to the primary response (e.g., ironoverload does not low status, barely gets the group up to the start of the ascending
produce and is not the same as polycythemia). limboftheresponse curve,andScenario“C,”atexactlythesame
The intake that gets an individual (or a population) up onto dose, but with the starting value (the control group) high up on
the response plateau is, virtually by definition, an intake that is theresponsecurve,pushesthenutrientstatusinthetreatedgroup
“adequate”foraparticular healthoutcome.Wherethatthreshold well up onto the response plateau. Neither scenario produces a
may be located along the range of plausible intakes is the locus very large change in the response variable—in the first instance
of current disagreement in the field of calcium and vitamin D becausethedosedidnotproducetheneedednutrientstatus,and
nutrition. It is not our purpose to debate that issue here, but in the second instance because the group as a whole already had
to explore (and document) why, given the constraints of the nearly enough of the nutrient concerned. Only Scenario “B”
sigmoid response characteristic, RCTs of calcium and vitamin D produces a large enough response to be readily detectable in a
may fail even when the sought-for causal connection exists. typical trial. In this example, all three interventions (doses) were
RCTs usually consist of two or more contrast groups with identical in magnitude, but differed radically in apparent effect.
differing levels of exposure to the test agent. With drugs that In other words, starting value is critically important, first in
arrangement is relatively straightforward, because one of the designofRCTsfornutrients,and second intheinterpretation of
contrast groups will usually be a placebo, i.e., an inert agent, their results. In brief the control group must have an intake low
96 Dermato-Endocrinology Volume4Issue2
enough to ensure that its members are near the left hand end of Once again, this is a precise duplication of the scenario depicted
the response curve and the intervention must be large enough to with intervention “C” in Figure1. In brief: neither CPEP nor
produce a meaningful change in nutrient status. Reasonable calcium-WHI had a low-exposure contrast group. Neither,
as this may seem, it is difficult to implement in practice as there therefore, could test the actual underlying hypothesis, i.e., that
are almost always serious ethical and feasibility barriers to such a low exposure to calcium causes disease.
design feature in humans. Dose/exposure. A second and related consideration is the size
Two large trials, directly managed by the National Institutes oftheintervention itself. Ithardly needs mention thatadosetoo
of Health, serve to illustrate these nutrient-specific issues. They small to change the exposure appreciably is not likely to produce
are the calcium and preeclampsia prevention trial (CPEP),7 and muchofaneffect,irrespectiveofstartingvalue.Whilethiswould
thecalciumandvitaminDarmoftheWomen’sHealthInitiative seem obvious, and perhaps even trivial, f.ailure to observe this
e
(WHI).8 Each contains defects of design with regard to one or constraint has been the reason for several of the failed trials of
c
more of the foregoing biological criteria which preclude their calcium and vitamin D (see below). Bischoff-Ferrari and her
n
answering the research question they set out to address. colleagues have repeatedly shown that trials that fail to use more
e
AtthetimeoftheCPEPtrial,severalrelativelysmallRCTshad than 400 IU/d and/or fail to elevate serum 25(OH)D above
i
shown that calcium supplementation during pregnancy signifi- certain levels alsocfail to reduce falls or fractures.12,13
cantly reduced the risk of preeclampsia, and in fact a Cochrane WHI exemsplifies precisely this exposure problem for vitamin
meta-analysis of these trials had concluded definitively that D. In thoe early to mid-1990s, when WHI was designed, the
calcium supplementation was efficacious for this endpoint9 and RDiA for vitamin D was 400 IU/d, and there was a general
B
asserted that no further trials were warranted. In those prior belief in the medical community that if people got that much,
.
trials, the control group intakes had been relatively low (si.e ., they would have all the vitamein D they needed for bone health.
toward the bottom of the response curve in Fig.1). Heowever, So, accordingly, the catlcium and vitamin D treatment arm
u
in designing and conducting the CPEP trial, ethicdal constraints of WHI included, in addition to the 1,000 mg of additional
b
madeiteffectivelyimpossibletouseinpregnanntwomenacontrol calcium, a daily supplemental intake of 400 IU of vitamin D.
intake appreciably below official US government recommenda- Once agaiin, after participants were enrolled, and their vitamin
a r
tionsfor calcium intake during pregnancy. Hence, when the trial D sttatus ascertained, it became clear that they had pre-study
L
s
was mounted, both treatment groups were assigned a calcium values for serum 25(OH)D well down toward the bottom end

intake considered nutrition2ally adequate for pregnancy. Tdheiof the response range (median: 17 ng/mL).14 Furthermore,
“treatment” group simp1ly got extra calcium, i.e., it was precisely when compliance was taken into consideration, it emerged that
t
the analog of the0situation depicted in Figure1 as intervention the actual mean vitamin D intake, rather than 400 IU/d, was
o
“C.” It was,2in hindsight, a test of the hypothesis that “more is closer to 200 IU/d, an intervention, which, in today’s under-
n
better,” ortheanalogofaproposalthatmoreironproducesmore standing, would have to be considered homeopathic. There was
he©moglobin once normal levels had boeen reached. no follow-up measurement of 25(OH)D in WHI to document a
Not surprisingly, the result oDf this trial was null. There was change in vitamin D status, so the level actually achieved is
no significant difference between the treated and the control unknown. It can be estimated that the average induced rise in
groups in terms of preeclampsia incidence, but the investigators 25(OH)D would have been no greater than ~2 ng/mL. Thus,
noted that the preeclampsia rate in their study was substantially for vitamin D, WHI illustrated something close to scenario “A”
below what had been expected (which is hardly surprising given in Figure1 (with the additional feature that the dose was itself
the hypothesis concerned and the fact that virtually every actuallysmallandhenceunlikelytochangetheeffectiveexposure
participant had an adequate or nearly adequate calcium intake). appreciably wherever it might have fallen along the response
This situation is clearly frustrating for investigators who feel, on curve).
the one hand, that they must use an RCT design, and on the Co-nutrient optimization. Another reason why RCTs of
other hand are constrained from using a control group with a nutrients might fail is lack of attention to co-nutrient status in
clearly deficient intake. the participants enrolled in a trial. Unlike drugs, for which co-
A similar problem occurred in the calcium supplementation therapy is either minimized or serves as an exclusion criterion,
arm of WHI. At the time WHI was designed,10 the median co-therapyinstudiesofnutrientefficacyisessential.Forexample,
calcium intake for women in the target age range was estimated for their skeletal effects calcium and vitamin D each need the
to be somewhere under 600 mg/d, and the design intervention other, and trials that fail to ensure an adequate intake of the
(an additional 1,000 mg Ca/d) would have elevated that intake nutrient not being tested will often show a null effect for the
to levels just slightly above the level recommended in the 1984 one actually being evaluated. Two Cochrane reviews, one of
NIH Consensus Development Conference on Osteoporosis.11 calciumandoneofvitaminD,15,16explicitlyexcludedstudiesthat
This seemed to be the right choice at the time, but when the used both nutrients, rejecting in the calcium review any study
participants were enrolled and randomized to treatment or using vitamin D, and in the vitamin D review, any study using
placebo, it was discovered that the control group had an actual calcium. They both thus failed on the issue of optimizing
intake of calcium above 1,100 mg/d. Not surprising, therefore, co-nutrient status, and in hindsight would have been predicted,
the effect of the calcium intervention on bone mineral density if not actually to fail, to produce at most only a small effect.
and fracture risk was small and, to some extent, inconclusive. Similarly, for calcium to exert a positive effect on bone, protein
www.landesbioscience.com Dermato-Endocrinology 97
intake needs to be adequate (actually somewhat above the but either 1-a-hydroxyvitamin D or calcitriol. Both of these
current RDA for protein).17 Virtually none of the published agents bypass normal physiological controls and produce phar-
calcium trials assessed or attempted to optimize protein intake. macologic responsesvery different fromnative vitaminD.Hence
Some may have had a protein intake adequate to enable a the studies concerned differed critically from the others included
skeletal response to calcium; others may not. The result would in the reviews, and pooling their results in a meta-analysis was
be a mixed group of outcomes—some positive, some null, but inappropriate. The reviews that admitted such studies into
none negative—exactly as the aggregate evidence shows. Other analysis failed on the criterion of using a single agent.
examples abound. The often ignored reality is that nutrients are Useofasingleoutcomemeasure.Cappuccioetal.,inameta-
not soloists; they are ensemble players. analysis of calcium and blood pressure, pooled studies reporting
We use these examples not as hindsight criticism of the absolute changes in blood pressure (in m.m Hg) with a single
e
studies concerned but to explain why their results were null (or study using the Z-transform of the blood pressure changes,20
c
nearly so), and to stress why their evidence does not contribute therebysubstantiallyunderstatingthesizeoftheaggregateeffect.21
n
to an understanding of the underlying research questions. When Even so, the authors found the estimated calcium effect to be
WHI was designed, we didn’t know the dose response relation- statisticallysignificant,buettoosmalltobe“clinicallyinteresting.”
i
ship for vitamin D, nor how much was needed to test whether it This is an instancecof pooling incommensurable endpoints.
had an effect. Nor was there any way to anticipate the healthy s
volunteer effect which contributed to the high calcium intake o Systematic Reviews
of the women who chose to be a part of WHI. Nor was the i
B
constructive interaction of protein and calcium recognized at In what follows we examine application of the criteria employed
.
the time when most of the calcium intervention trials wse re for inclusion of studies in esystematic reviews of calcium and
mounted. What we do criticize is the continued use todaey of the vitamin D in an attemtpt to address the question of whether
u
results of such trials as evidence that calcium andd vitamin D they were capable of answering the research questions posed. We
b
may not have certain of the effects attributednthereto. WHI was note that the nearly universal absence of biological criteria for
a very large trial and hence its seemingly null results heavily admittingistudies into review does not, in itself, prove that such
a r
weight any kind of meta-analysis or systematic review in which revietws are flawed, but it does raise that possibility. For example,
L
this study is admitted into the analysis. ifswe don’t know the baseline vitamin D status in the studies

2 i
Population heterogeneity. It is hardly necessary to remdind whose outcomes are pooled, we have no way to judge whether
ourselves that not ever1yone is the same (as if we wer e inbred that pooling was valid. Moreover, as we have already shown,
t
mice); neverthele0ss it may be useful to illustrate how much CPEP and WHI should not have been included in any such
o
differenceth2atheterogeneitycanmakeintheoutcomeofnutrient review. Unfortunately, systematic reviewers have not usually
n
trials. A perfect example, from the field of nutrition, is the fact realized that.
th©at a substantial fraction of the popoul ation has a mutation in Literally dozens of systematic reviews of calcium and vitamin
the 5,10-methylene-tetrahydrofolDate reductase gene which causes D have been commissioned and/or undertaken. Two, however,
them to have a greatly increased requirement for choline.18 A haveplayedaparticularlyprominentroleinhelpingtodetermine
typical RCT testing choline response in a general population health policy in both the United States and Canada. One was
sample would find an average response that was either not performed by the University of Ottawa Evidence-Based Practice
statistically significant or too small to be “interesting.” However, Center,22labeledinwhatfollows“Cranney,”andtheotherbythe
inpointoffact,thatresponsewouldbeacompositeofindividuals Tufts Evidence-Based Practice Center,23 labeled “Chung.”
who, on the one hand, were already at or near the plateau of the Cranneyaddressedstudiesoftheefficacy and safety ofvitamin
choline dose response curve (and hence would be expected to D in relation to bone health, while Chung included studies that
experiencenoperceptibleresponse)andaminorityofindividuals, evaluated multi-system health outcomes. Chung actually used
ontheother,whogotalargeresponsebecause,giventheirunique Cranney for its bone-related effects, supplementing it with six
genetic composition, the same basal intake was at the bottom more recent studies reporting bone health outcomes.
of their response curves; hence only they were in a position to Both Cranney and Chung admitted only studies that used
respond. Understanding this, as we do today, would prevent native vitamin D (either D or D ). None of the studies that
2 3
an investigator from designing a trial of choline supplementa- had been inappropriately included by Papadimitropoulos et al.15
tion without taking genetic composition of the participants into andWangetal.19(usingcalcitrioloritscongeners)wereincluded
consideration. However, similar allelic differences in vitamin D in either review. Both thereby met the criterion of testing a
metabolism appear to be present in the general population, but single agent. And, while D is now generally considered to have
2
are today not adequately understood or quantified, and could lower molar potency than D ,24-27 most of the included studies
3
not readily be ascertained in advance of a trial. Hence such trials using D employed a dose large enough to overcome the
2
may be indeterminate because only some of the enrollees would potency difference. However, neither Cranney nor Chung used a
be in a position to respond. minimum dose criterion; nor did they require documentation of
Utilizing a single form of the nutrient. Reviews by Papadi- a therapeutic blood level in the treated group. Doses too small
mitropoulos et al.15 and Wang et al.19 both included studies in to change serum 25(OH)D by at least 5–8 ng/mL would, for
which the treatment agent was not actually vitamin D itself, practical purposes, be null-effect doses. (Once again, this is not
98 Dermato-Endocrinology Volume4Issue2
to criticize the original studies using small doses, since effect it provides a useful way of understanding the results from a
size may have been unknown when they were designed. We group of studies, as well. Briefly, if the sought for effect is
stress, however, that an inclusion criterion that is blind to dose relatively small, and if, in a group of studies, the average power
effectively treats vitamin D as a binary variable, rather than the is about 0.6, then one would expect about two out of five of
continuous variable that it is). thestudiestobenull,preciselythesortofpatternwhichhasbeen
Neither Cranney nor Chung required that a study have theexperienceofinvestigatorsevaluatingmultisystemresponsesto
reported low basal vitamin D status. While some of the included nutrient intake changes. Even with a power of ~0.85, generally
studies did report basal status, many did not. Nevertheless the consideredadequate,oneoutofsixstudiesofanactuallyeffective
outcomes of all included studies were pooled in deriving an agent can be expected to be null. Because nutrient effects, in
estimate of the aggregate effect. As already noted, pooling studies general, tend to be small, this issue of po.wer is larger than is
e
spanning different regions of the x-axis of Figure1 would commonlyrecognized.However, itneeds noparticulartreatment
c
inevitably result in diminution of apparent effect size. here beyond acknowledging that it undoubtedly is a part of the
n
Neither Cranney nor Chung made any apparent attempt to explanation for failed RCTs. Nevertheless, to characterize the
use co-nutrient optimization as a criterion for inclusion of a resultsofsuchstudiesas“ienconsistent,”whiletechnicallyaccurate,
i
studyintoanalysis.Indeed,iftheyhad,itislikelythattheywould is to fail to undersctand the role of statistical power.
have come up with empty or nearly empty sets. Some of the Another reasson forfailure ofan RCT isloss of subjects during
vitamin D studies used supplemental calcium as well, and vice the courose of the trial. Such losses broaden the confidence
versa, but others did not. In any event, for neither Cranney nor inteirvals around outcome measures in the contrast groups and
B
Chung was there an attempt at analysis to determine whether thereby greatly obscure differences (if any) between them. The
.
the effect with both nutrients differed from the effects of onse o r RECORD24trialisagoodexaemple,withdocumentedcompliance
the other alone. e at less than 40%. Intenttion-to-treat analysis (ITT), used in this
u
Both Cranney and Chung used the WHI andd RECORD28 instance, found no benefit from vitamin D. ITT is designed to
b
studies in their analysis. Chung used CPEP ans well. For reasons protect investigators from drawing inappropriate conclusions of
discussed in more detail under General Clinical Trial Issues, efficacywhiensuchlossesoccur.ButITTinevitablybiasestoward
a r
below, all three trials heavily weight the estimate of pooled effect the ntull. This is not so much to argue for per-protocol analysis,
L
s
toward a null value. as to stress that studies in which there have been more than

2 i
In brief, Cranney and Chung, between them costing closedto minimal subject dropouts, may be null in part because ITT
$2 million, by failing to1 use appropriate and necessary b iological designedly minimizes effects. But null is not negative. It must
t
criteria, did not, 0and could not provide the solid evidence base also be mentioned that dropouts themselves imperil (or destroy)
o
needed to in2form nutritional policy deliberations. the randomization and thereby convert a RCT to a concurrent
n
By w ay of contrast, we call attention to yet another systematic cohort study.
rev©iew, in this case one published tooo late to be used in the
formulation of current policy. PaDrker et al.29 screened over 6,000 Conclusion
prospective cohort studies, identifying 28 (with a total of 99,745
participants) that met criteria for inclusion. They based their We have focused mainly on certain biology-based, dose response
evaluation of effect on the association between achieved serum issues, primarily because they are able to explain, by themselves,
25(OH)Dconcentrationandcardiovascularoutcomes,andfound muchofthemixedrecordofresponseinRCTsrelatingtoeffects
for the highest vs. lowest vitamin D status groups a highly of calcium and vitamin D on disease risk. We stress that these
significant odds ratio of 0.57 (95% CI: 0.48–0.68). Consistent are not the only reasons a methodologically well designed and
with the emphasis of Bischoff-Ferrari et al.13,14 and many others, executedRCTmayfail.Buttheysufficetoshowclearlywhysuch
a criterion requiring documented values for achieved vitamin D studies can fail—and, indeed, should have been expected to fail.
status, the basis for the Parker analysis, is vitally important This analysis has shown both that many of the existing RCTs
if systematic reviews are to be informative. Incidentally, it may of calcium and vitamin D contain substantial, and sometimes
be worth noting in passing that the studies evaluated by Parker fatal, design flaws—flaws that preclude their adequately addres-
et al. were of the non-concurrent cohort type, which permits sing the research questions they set out to answer. Systematic
low-dose contrasts groups that would usually be unfeasible with reviews that nevertheless include such flawed studies will
the RCT design. inevitably be misleading and should not, we maintain, be used
as a basis for developing nutritional policy.
General Clinical Trial Issues Indirectlywehaveshownalsothatresearchquestionsconcern-
ing nutrient efficacy in humans are intrinsically hard to address.
A probabilistic reason why a group of studies might exhibit the By implication, approaches different from those of EBM would
pattern that seems to characterize the corpus of calcium and seemtobeneeded.29Inanycase,itisinescapablethatconclusions
vitamin D studies is the matter of statistical power. Power is drawn from null-effect studies that contain significant biological
often used to characterize and/or evaluate individual studies, but flaws reveal essentially nothing about nutrient efficacy.
www.landesbioscience.com Dermato-Endocrinology 99
References 13. Bischoff-Ferrari HA, Willett WC, Wong JB, 21. Birkett NJ. Comments on a meta-analysis of the
Giovannucci E, Dietrich T, Dawson-Hughes B. relation between dietary calcium intake and blood
1. Blumberg J, Heaney RP, Huncharek M, Scholl T,
FracturepreventionwithvitaminDsupplementation: pressure.AmJEpidemiol1998;148:223-8,discussion
StampferM,ViethR,etal.Evidence-basedcriteriain
ameta-analysisofrandomizedcontrolledtrials.JAMA 232-3;PMID:9690357
the nutritional context. Nutr Rev 2010; 68:478-84;
2005; 293:2257-64; PMID:15886381; http://dx.doi. 22. CranneyA,HorsleyT,O’DonnellS,WeilerH,PuilL,
PMID:20646225; http://dx.doi.org/10.1111/j.1753-
4887.2010.00307.x org/10.1001/jama.293.18.2257 OoiD,etal.EffectivenessandsafetyofvitaminDin
14. Wactawski-WendeJ,KotchenJM,AndersonGL,Assaf relationtobonehealth.EvidRepTechnolAssess(Full
2. HeaneyRP.Nutrients,endpoints,andtheproblemof AR, Brunner RL, O’Sullivan MJ, et al. Women’s Rep)2007;158:1-235;PMID:18088161
proof.JNutr2008;138:1591-5;PMID:18716155
HealthInitiativeInvestigators.CalciumplusvitaminD 23. ChungM,BalkEM,BrendelM,IpS,LauJ,LeeJ,
3. HeaneyRP,WeaverCM,BlumbergJ.EBN(Evidence-
supplementation and the risk of colorectal cancer. N etal.VitaminDandcalcium:asystematicreviewof
BasedNutrition)Ver.2.0.NutrToday2011;46:22-6;
Engl J Med 2006; 354:684-96; PMID:16481636; healthoutcomes.EvidRepTechnolAssess(FullRep)
http://dx.doi.org/10.1097/NT.0b013e3182076fdf
http://dx.doi.org/10.1056/NEJMoa055222 2009;183:1-420;PMID:20629479
4. Biesalski HK, Aggett PJ, Anton R, Bernstein PS,
15. PapadimitropoulosE,Wells G,SheaB,GillespieW, 24. Armas LAG, Hollis BW, Heaney RP. Vitamin D
BClounmsebnesrugsJC,onHfeeraennecye,RSePp,teemtbearl.1216,t2h01H0oShceinenhteiifmic WeaverB,ZytarukN,etal.OsteoporosisMethodology is much less effectieve t.han vitamin D3 in humans.2J
Group and The Osteoporosis Research Advisory Clin Endocrinol Metab 2004; 89:5387-91; PMID:
substantiationofhealthclaims:evidence-basednutrition. Group.Meta-analysesoftherapiesforpostmenopausal 15531486;hcttp://dx.doi.org/10.1210/jc.2004-0360
Nutrition 2011; 27(Suppl):S1-20; PMID:21700425;
http://dx.doi.org/10.1016/j.nut.2011.04.002 osteoporosis. VIII: Meta-analysis of the efficacy of 25. HeanenyRP,ReckerRR,GroteJ,HorstRL,ArmasLA.
5. WillettWC.TheWHIjoinsMRFIT:arevealinglook vpiotastmmiennoDpautsraelatwmoemnteni.nEnpdreovcernRtinevg2o0s0te2o;p2o3r:o5s6is0-i9n; ehVuitmamanins.JDC(3l)inisEnmdoorcerinpoolteMntettahban20v1i1ta;m96in:E4D4(72)-5i2n;
beneath the covers. Am J Clin Nutr 2010; 91:829- PMID:12202471; http://dx.doi.org/10.1210/er.2001- i PMID:21177785; http://dx.doi.org/10.1210/jc.2010-
30; PMID:20181816; http://dx.doi.org/10.3945/ajcn. 8002 c 2230
2010.29347
16. SheaB,WellsG,CranneyA,ZytarukN,RobsinsonV, 26. Binkley N, Gemar D, Engelke J, Gangnon R,
6. Heaney RP, Rafferty K. Preponderance of the evi- GriffithL,etal.OsteoporosisMethodoloogyGroupand Ramamurthy R, Krueger D, et al. Evaluation of
dence: an example from the issue of calcium intake The Osteoporosis Research Advisory Group. Meta- ergocalciferol or cholecalciferol dosing, 1,600 IU
aPnMdIDb:o1d9y14c6o5m04p;ositihotntp.:/N/duxt.droiR.oervg/1200.0191;116/7j.:13725-39-; aVnIaIl.yMsesetoaf-atnhaelyraspisieosffcoarlcpiuomsBtmsueinpoppleamuseanltaotsitoenopfoorrotshise. dEanidlyocorrino5l0,M00e0tabIU20m11o;n9th6l:y98i1n-8o;ldPeMrIaDdu:2lt1s.28J92C4li9n;
7. 4L8ev8i7n.e2R00J,8H.0a0u1t3h2J.Cx,CuretLB,SibaiBM,CatalanoPM, pRreevve2n0ti0o2n;2o3f:5p5so2s-t9 m;ePnMopIaDu:s1a2l2o0s2t4eo7p0o;rhotstips.://Ednx.ddoocir. httpe://d.x.doi.org/10.1210/jc.2010-0015
27. TrangHM,ColeDEC,RubinLA,PierratosA,SiuS,
8. MNhJRatoctopbkrEs:br/onii/snngdCslxD.RJdJ,,oDeMiL.t,oeaewrlLdg.i/aTs1C1r0Ci9ra.o1El9i0o7,xf5;ec6tAa3/lNZ3cail7,u.E:m6JWGM9ta-oo17sms9p69r;eeMn7vP’0es,Mn7Ht1WIp0Der3aae:3lle9tl7cah2lc0a1e2mI1n06piR1t7siBi5aa-.,; n17. oDeb2nro0agcn0dw/ee21ss;0olot.7n1hs5-s2eH:17einau07sg/3seeho-lrd9ec.s2ie;ar0PtlBiy0Mo,1nm-IH7Deo0an:fr01ra21ipns9rdo1St6Swe.7ion6Cm7iaenlnctai.ukAmemiwinJitthCaklirenatbieNnsfuluotr-f u 2t8. VhDGyir2eda.trnhAotxmRAy.vMJiEtC,avmiAldivnienennNcDeeullttmhrAa1,ot9rCve9iat8mea;fmfp6icbi8niee:8lnDl5tMl34y-iK8nt;h,craPMenMacsdDeIsoDoes:nse9ra7vuld7imt1aAm82M65in2-,
tm2i0vee0n6tIa;ntiv3oen5s4ti:ag6na6dto9r-ts8h.3e;CraiPslkcMiuIomDf:f1rp6alc4ut8su1rve6si3t.a5 mN;inhEttnDpLg:l//sdJuxpM.dpaoleedi-. 18. KGfeorenhpelamttihcewievarayrMiaptr,ieoddnaicotCsfosfsuotslaacteKep--tmAisb,eidlFiitaiytstcethdoercorhnLoeMl-icn,aerZbdoeeinfsiectlireaSnnHcsy-. MGp(RrreaaovcnueLdnpeot.nmionOniasrneadolGfvSEli,otvaMwamlu-ctiaPrntahiuoeDmnrs3aoonaffrnGaCdcCtaucl,careielutcsmiauli.mnROEefrloCdrveOirtslRayemcDopinneTdorapDirlaye,l
9. oBrugc/h1e0r.1H0C56,/GNuEyJaMttoGaH05,5C2o11o8kRJ,H2atalaR,CookDJ, i1n60h2u5m-3a0n;s. PPrMocIdDN:1ait6l23A6c7ad26S;cihUttpS://Adx.2d0o0i.5o;rg1/0120:. RLaEnCceOtR2D00):5;a36ra5n:1d6o2m1i-s8e;dPpMlaIcDeb:1o5-c8o8n5t2ro9l4le;dhtttrpi:a/l/.
LangJD,etal.Effectofcalciumsupplementationon 1073/pnas .0504285102 dx.doi.org/10.1016/S0140-6736(05)63013-9
pm1r9ee9tga6n-;aann2c7ayl5y-:si1ni2s1d1uo3cf-e7dr;an0PhdMyopmIeDritz:ee8nd6s0ioc1no9n3ta1rno;dlhletdptpre:t/er/cidalalxsm..dpoJsAii.aoM:rgAa/ n19. WIannogdaenlgrsissotknTJo,Ef,cPaLerandncioiienvraasKcMu,laJer,tdaBils.oeoaVtsheit.amCSLiinr,cuDJlaatcidqouenfeisc2ie0nP0cF8y,; 29. PSca,arrdkKieoarmnJd,eatHalabaoshlNimc-Bid,iOso,ertDdeurastl.:tosnLyseDtvee,mlsMataiocvfrroevdviatiearimwsiAna,nSdDtrmanaegnteads-
10©. 1JWoav0ceo.k1rm sv0oie0ennw1’R/sjDaamHn,daeL.a1albCt9ha9rsoe6Ili.ixn0ni3Aeti5Za3ct,ih0vCa3er8aau0cclat0eel5ycri5iJus0Atmi3c,1-sMviotcafGDmopiwnaratinDcioJp.atTrn ihatsle:. 20. 11CD11aA76p,:1p5/u0CCc3ucI-Rit1olCe1rF;UPJL,PAAEM.TllIiIEDoOpt:tiN1dP8eA,1mHA8i0loAl3el.on91gd50ie;c7r.7hPa0stSs6t,op1cP:2/ir/a7ydteixor.ndJo,iFb.oeortlgwlm/e1ae0nn. 30. ahH7n0teta:pa1ly:n6/se/5idsy-.x9.MR;dPoai.t.uoPTrrMight/aeI1sD0n2.:1u0201t2r103ie6;6n/46jt.15m:5p2a8r2to;u5br-li3etmah6s;t..t2Pp0NM:/0/u9IdtD.xr1.:2d2R.o00ei01v.o33r21g03/114028.;;
Ann Epidemiol 2003; 13(Suppl):S98-106; PMID: dietary calcium intake and blood pressure: a meta- 1111/j.1753-4887.2011.00469.x
14575942;http://dx.doi.org/10.1016/S1047-2797(03) analysis of published data. Am J Epidemiol 1995;
00046-2 142:935-45;PMID:7572974
11. Osteoporosis Consensus Conference. JAMA 1984;
252:799-802; PMID:6748181; http://dx.doi.org/10.
1001/jama.252.6.799
12. Bischoff-FerrariHA,Dawson-HughesB,StaehelinHB,
Orav JE,StuckAE, Theiler R, etal. Fall prevention
with supplemental and active forms of vitamin D: a
meta-analysis of randomised controlled trials. BMJ
2009;339:b3692;PMID:19797342;http://dx.doi.org/
10.1136/bmj.b3692
100 Dermato-Endocrinology Volume4Issue2