Upload
eugene-h-blackstone
View
214
Download
2
Embed Size (px)
Citation preview
ble biologic markers of the injury that are reliable, spe-cific, sensitive, and quantitative? What are the etiolo-gies of the injury and can they be either avoided or neu-tralized? Are functional changes reflective ofmorphologic injury? What aspects of dysfunction arerelated to true injury and what fraction to a myriad ofother aggravating factors that may be neutralized?
Neurologic injury and cognitive and behavioral dys-function after cardiac surgery are at least as complexphenomena as myocardial injury and functional stun-ning. The centrality of the brain in all that it means tobe human adds further complexity and confounding tothe phenomena. The lesson we have learned from thephenomenon of cardiac injury is that it is possible togain both nonspecific and mechanistic insight into thephenomenon and, thereby, at least partially avoid injuryand neutralize dysfunction.
Are simple analyses adequate?Reduction of cardiac injury and management of car-
diac dysfunction did not come about as the result ofsimple, dichotomous questions such as, “Has the heartbeen injured?” “Is there any cardiac dysfunction?”Instead, insight was gained when we asked, “Howmuch?” “What is the time course?” “What are themechanisms?” “How do quantitative markers relate toquantitative amounts of injury?”
Thus, for me, the most insightful aspect of the articleby van Dijk and colleagues is their discussion of thecurrent state of the analytic approach to neurocognitivedysfunction associated with cardiac surgery (which Iwould classify as the functional analogy of cardiac out-put, compared with structural injury per se). They pointout that psychometric tests are not standardized, whichposes problems of comparability. At the same time,suggesting a specific suite of standardized testing mayplace a straightjacket on neurocognitive research thatcould inhibit development of methodology to yield bet-ter insight into the nature of the phenomenon. Theypoint out that the tests performed generate large quanti-ties of data, but in the end, these data are often con-
SimplismSimplism (noun): The tendency to oversimplify an
issue or a problem by ignoring complexities or compli-cations.1
In this issue of the Journal, van Dijk and his col-leagues2 ask a simple question: What proportion ofpatients have persistent cognitive dysfunction aftercoronary artery bypass grafting with cardiopulmonarybypass? Their pursuit of a single-number simple answerbegan as a meta-analysis of the literature. However,they abandoned formal meta-analysis when theydeemed that compatibility of studies was insufficient.Instead, they found an answer as a simple, weightedaverage from a handful of systematically reviewedpapers.
Simple question, simple answer? Or simplistic ques-tion and simplistic answer?
De ja vu As mortality and morbidity associated with cardiac
injury and dysfunction have retreated in recent years,acute and persistent central nervous system injury anddysfunction after cardiac surgery have emerged as adominating issue. However, the questions being askedabout neurologic events from cardiac surgery parallelthose of a generation ago, when possible injury to theheart was center stage. What is the definition of irre-versible injury? Can the injury be quantified? How canirreversible or irreparable injury be distinguished fromreversible phenomena? Are there blood-borne, accessi-
629
SURGERY FOR ACQUIREDCARDIOVASCULAR DISEASE
EDITORIAL: NEUROLOGIC INJURY FROM CARDIAC SURGERY—AN IMPORTANT BUT ENORMOUSLYCOMPLEX PHENOMENONEugene H. Blackstone, MD
From the Department of Thoracic and Cardiovascular Surgery, TheCleveland Clinic Foundation, Cleveland, Ohio.
Received for publication July 18, 2000; accepted for publication July19, 2000.
Address for reprints: Eugene H. Blackstone, MD, The ClevelandClinic Foundation, 9500 Euclid Ave, Desk F25, Cleveland, OH44195 (E-mail: [email protected]).
J Thorac Cardiovasc Surg 2000;120:629-31Copyright © 2000 by The American Association for Thoracic
Surgery0022-5223/2000 $12.00 + 0 12/1/110252doi:10.1067/mtc.2000.110252
densed into a simple (simplistic), information-losingformat, such as a dichotomous change of one standarddeviation. The authors go on to make a wonderful casefor using more of the information, which is well with-in the capabilities of modern statistics. Yet, astonish-ingly, in the end, they settle for a simplistic dichoto-mous answer to their question, the answer to which allof us know must be, “It depends.”
If we ignore the answer but pay attention to theirinsight, we should be stimulated to perform moreappropriate analyses, as well as to seek new and bettermethods to quantify the injury, whether it be by neuro-logic testing, magnetic resonance imaging, or betterserum markers than S100β.
What is the role of cardiopulmonary bypass?The authors are also correct in wanting to look at
the contribution of such variables as support mecha-nisms in neurologic injury and dysfunction.Certainly, one of the factors that needs to be under-stood better is the contribution of cardiopulmonarybypass. The authors suggest a comparative study ofon- and off-bypass coronary artery bypass grafting.This suggestion fails to take into account the com-plexity of that comparison. Off-bypass coronaryartery bypass grafting does not simply remove a sin-gle variable (cardiopulmonary bypass) from the mix.To the contrary, off-pump coronary artery bypassgrafting introduces other factors that may be associ-ated with neurologic injury and dysfunction, includ-ing the extensive manipulation required to the heartand the use of aortic side-biting clamps.
Two methodologic commentsTwo data-analytic issues are raised in the discussion
that are important for me to reinforce. The first is thenecessity to account for individual pre-test values inmaking group comparisons rather than simply usinggroup means. Failure to take into account individualpatient variation (preoperative test values) when mak-ing group comparisons is not simply wasteful of infor-mation, but it can lead to inaccurate inferences.However, so-called change scores are only one particu-lar form of taking preoperative variation into accountand may not be an optimal choice. This is a sophisti-cated statistical issue that I believe has been incom-pletely addressed at this time.
The second analytic issue addressed by the authors istest-retest learning. It is crucial to account for this fac-tor, but quantifying the degree of correction necessaryis challenging. The systematic direction of this biasworks in such a way as to mask important effects.
Meta-analysisFinally, the authors state they have performed a
systematic literature review, not a meta-analysis.Why is the study not a meta-analysis? A “systematicreview” is an overview of independent publicationsthat uses explicit, reproducible methods and criteriato avoid bias.3 Systematic reviews can include as oneof their components meta-analysis, which is a suiteof mathematical and statistical procedures that inte-grate several independent studies that are deemedcomparable.
The particular class of meta-analysis appropriatefor systematic reviews is the one for which the unit ofobservation is the independent study. Simple poolingof information from multiple sources to obtain a so-called weighted mean, as the authors present, hasbeen around since the 18th century, but the termmeta-analysis, coined by Glass4 in the mid-1970s,connotes a greater expectation. Meta-analysis in-volves meticulous, disciplined, systematic review ofthe literature using clear criteria for and assessmentof study quality, determination of comparabilityamong studies, and extraction of similarly defineddata elements, as is exemplified by the authors. Inaddition to this, many of the following elements arefound frequently in the subsequent quantitative inte-gration of independent publications, that is, the meta-analysis of the information.5
• Quantitative evaluation of the diversity (hetero-geneity, consistency) of results among the differentstudies
• Use of formal quantitative methods to explore thenature of the heterogeneity of results
• Estimation of the overall effect (fixed effect) orhow it varies across studies (random effect),often employing metaregression to identify influ-ential variables such as date of the study, differ-ences in demography among studies, differencesin other characteristics of the study population,size of the study, and in this study, average timeof the post-test
• Sensitivity analyses to determine if the estimates ofan effect are stable (robust)
• Formal evaluation of bias, including publicationbias and study quality bias
In addition, most meta-analysis efforts would haveincluded as wide as possible sample of studies, gener-ally more than the final handful used in this study,accounting as best as possible for clear aspects of het-erogeneity to avoid study selection bias. None of theseanticipated elements of meta-analysis are contained inthis paper, so I would agree with the authors that the
630 Blackstone The Journal of Thoracic andCardiovascular Surgery
October 2000
character of the paper is one of systematic review with-out an accompanying meta-analysis component.
Asking the right questionsThus, in the end, asking how often neurocognitive
dysfunction persists after coronary artery bypassgrafting with cardiopulmonary bypass, and expectinga simple one-number answer to suffice, is simplistic,however desirable. Rather, let this article be a stimu-lus of the development of better quantitative mea-sures of neurologic injury and cognitive and behaviordysfunction, to the better understanding of the rela-tion of neurologic injury and neuropsychiatric func-tion, to the application of better data-analyticmethodology that more fully use the precious infor-
mation that is gathered, and to the development ofmethods to reduce injury and neutralize dysfunction,even in the absence of its complete understanding.
R E F E R E N C E S1. American Heritage Dictionary of the English Language. ed 3.
Boston: Houghton Mifflin; 1992. p. 1683.2. van Dijk D, Keizer AMA, Diephuis JC, Durand C, Vos LJ,
Hijman R. Neurocognitive dysfunction after coronary arterybypass surgery: a systematic review. J Thorac Cardiovasc Surg2000;120:632-9.
3. Egger M, Smith GD. Meta-analysis: potentials and promise. BrMed J 1997;315:1371-4.
4. Glass GV. Primary, secondary, and meta-analysis. EducationalResearcher 1976;5:3-8.
5. Lau J, Ioannidis JPA, Schmid CH. Quantitative synthesis in sys-tematic reviews. Ann Intern Med 1997;127:820-6.
The Journal of Thoracic andCardiovascular SurgeryVolume 120, Number 4
Blackstone 631