Does Age Modify Effectiveness of Statin Therapy in Coronary Artery Disease?

TO THE EDITOR: Chopra and colleagues (1) discuss a need to limit preoperative stress testing, preoperative coronary revascularization, and perioperative -blockade. They use the phrase “time to throttle back,” because a growing body of research—much of it published in the past decade—has shown that these interventions do not improve outcomes in most patients. The authors promote the 2007 American College of Cardiology/American Heart Association (ACC/AHA) guideline on perioperative cardiovascular care for noncardiac surgery as a cost-effective and evidence-based approach. What Chopra and colleagues do not mention is that several of them were authors of the 1996 and 2002 versions of the same ACC/ AHA guideline. Those versions promoted stress testing for large subgroups of patients despite minimal evidence that stress testing, and the coronary revascularization it sometimes spawned, was beneficial. In other words, several authors of the current article were themselves responsible for the “throttling forward” that now makes it necessary to “throttle back.” As I have argued elsewhere (2), the 2007 ACC/AHA guideline is an improvement over the 2002 version, but it repeats the mistake of suggesting interventions for which no supporting evidence exists. For example, limited functional capacity (as indicated by the inability to perform activities requiring 4 metabolic equivalents) is a decision point for stress testing in certain patients, despite the lack of compelling evidence that stress testing improves perioperative outcomes in these patients. Moreover, algorithms for several patient subgroups end with a recommendation to “consider noninvasive testing . . . if it will change management” (3). But without evidence to justify stress testing in these subgroups, how should clinicians determine when testing will change management? An analysis (4) showed that, all too frequently, ACC/AHA practice guidelines make recommendations for which there is no conclusive evidence. The sequence of guidelines on perioperative evaluation exemplifies this problem.

IN RESPONSE: Dr. Brett says that several of our coauthors are responsible for the "throttling forward" of perioperative care because they served on ACC/AHA guideline writing committees.Although earlier iterations of the guidelines may now seem more testingfriendly, dozens of experts from around the world compiled these guidelines from the evidence available at the time.These data were derived largely from retrospective, observational studies that lacked the depth of prospective trials, the advantages of blinding, and the obvious benefits of randomization.In our article, for example, we describe how the Coronary Artery Surgery Study (an analysis by one of our coauthors) (1) was misinterpreted as supporting perioperative revascularization.We candidly discuss how the early evidence-and our interpretation of it-led us astray.This dialogue forms the basis of our call to "throttle back."Guidelines are a synthesis of current evidence and opinion; both must, should, and will change.Our article testifies to our commitment in this regard.
In addition, Dr. Brett states that the guidelines occasionally suggest interventions that lack supportive evidence and that they often end with unclear recommendations.We reiterate that there is still much to learn regarding perioperative practice.The algorithmic approach adopted by the guidelines synthesizes current data and expert opinion into implementable recommendations for end users who may not be familiar with this complex (and controversial) body of evidence.Each recommendation is intended to represent a reasonable approach to a potentially infinite array of clinical situations.Guidelines are not meant to replace clinical judgment, but rather to supplement it.To bypass clinical reasoning or remain mute on important issues for which the evidence is scant or unclear would be a disservice to those who are likely to ask these very questions and need this doctrine the most.
We conclude with a single, unifying thought: Perioperative medicine is far from a perfect science.Rather, it is a discipline in its infancy that remains full of promise but also full of pitfalls.As we discussed elsewhere (2), we must be cautious with what questions we ask and how we choose to answer them.Only through constant analysis of the evidence will we perfect our perioperative paradigm; until then, we advise discretion in perioperative testing and treatments.

Does Age Modify Effectiveness of Statin Therapy in Coronary Artery Disease?
TO THE EDITOR: Hayward and colleagues (1) presented a modeling analysis comparing the total effect and efficiency of a completely risk-based "tailored-treatment" strategy with a traditional National Cholesterol Education Program (NCEP)-based "treat-to-target" strategy.As noted by the authors, treatment benefit in any given person is a function of his or her baseline risk and the relative risk reduction of therapy.The authors then state that ". . .we know of no evidence that any single CAD [coronary artery disease] risk factor or biomarker helps predict the relative effectiveness of starting statin therapy." But there is some evidence that age modifies the effectiveness of statin therapy.Recent large, patient-level meta-analyses (2) show levels of total and non-high-density lipoprotein cholesterol to be much stronger risk factors (on a relative risk basis) in young adults than in older adults (for example, relative risk, 0.44 per 1-mmol/L decrement in total cholesterol level at age 40 to 49 years vs. 0.72 at age 60 to 69 years).This interaction is also noted in statin trials (3): The relative risk reduction is lower in patients younger than 65 years than in patients older than 65 years (interaction P value ϭ 0.01).
In our analysis of risk-based guidelines versus NCEP-based guidelines (4), we included our best estimate of this age interaction based on a custom analysis of low-density lipoprotein cholesterol levels in Framingham data, which closely matches meta-analysis results.Using these age-based estimates, we found (to our surprise) that the NCEP-based guidelines were actually somewhat more effective than a purely risk-based approach, because a risk-based approach tends to preferentially target older persons (consistent with Table 2 in Hayward and colleagues' article).In a sensitivity analysis in which we modeled a constant relative risk reduction across age (similar to Hayward and colleagues' assumptions), the relative efficiency of these approaches was reversed (risk-based therapy was more efficient).This sensitivity analysis is described in the Results section and is fully documented in the online Appendix of our article (4).
Although our risk-based approach was not precisely the same as the tailored-treatment approach advocated by Hayward and colleagues, we suspect that the latter approach, which seems to dominate the treat-to-target approach, would be less efficient if statins are less effective in older persons.If so, this caveat to the authors' findings and conclusions would be important.The meta-analysis by Lewington and colleagues (1) is not very relevant to our study.It combined observational studies of varying quality and completeness that were not treatment trials and thus do not reflect the effect of therapy on outcomes.Furthermore, these studies examined only mortality and did not include other cardiovascular events.

Mark
We are also confused about why Drs.Pletcher and Coxson refer to relative risks associated with total cholesterol level, which is a relatively poor predictor of ischemic heart disease, and did not report the non-statistically significant difference in the hazard ratios of total cholesterol to high-density lipoprotein cholesterol in persons aged 40 to 59 years versus persons aged 60 to 69 years, which was clearly the most powerful predictor in the study by Lewington and colleagues and is the measure used in our study.Furthermore, the meta-analysis from the Cholesterol Treatment Trialists' (CTT) Collaborators (2) did not find an age interaction for low-density lipoprotein cholesterol level.Finally, even if Drs.Pletcher and Coxson's hypothesis (that prediction tools, such as the Reynolds and Framingham risk scores, systematically under-or overpredict risk for CAD mortality by age) is someday found to be true, the appropriate reaction would be to recalibrate these prediction tools, not to change our tailoredtreatment approach.
Drs. Pletcher and Coxson are incorrect that the CTT Collaborators' meta-analysis (2) found statistically significant evidence that the relative effects of statins differ by age.Only 1 of the 10 subgroup analyses in the study had a P value less than 0.05.After the authors appropriately adjusted for multiple comparisons, they found no statistically significant evidence for heterogeneity across subgroups (P ϭ 0.4) and properly concluded that ". . . the proportional reductions in such major vascular events per mmol/L LDL [low-density lipoprotein] cholesterol reduction were similar irrespective of the pretreatment cholesterol concentrations or other characteristics (eg, age, sex, or pre-existing disease) of the study participants [emphasis added]."Furthermore, another comprehensive review of the literature (3) found no evidence that the benefits of statins vary by age for primary prevention (this review was based on a meta-analysis of clinical trials, and our study deals with primary prevention only).
Although we feel that the evidence for this hypothesized agestatin interaction is weak, our proposed tailored-treatment approach is still more effective than the NCEP approach, regardless of whether this interaction is real.Our approach differs markedly from that of Drs.Pletcher and Coxson because they examined a single fixed-dose approach, which we show is almost as bad as the NCEP approach (see second-to-last paragraph of our article, in which an article by Dr. Pletcher and colleagues is referenced directly).What makes a tailored-treatment approach much better than both fixed-dose and treat-to-target approaches is that tailored treatment much more accurately identifies persons who do and do not benefit from more aggressive care (4,5).
Finally and most important, even if the hypothesized age-statin interaction is confirmed, we would still never favor the inherently inferior treat-to-target approach.The more heterogenous treatment effects are across the population, the more important tailored treatment becomes (see sections A and D of the Appendix to our article).If other physicians accept the point estimates in the CTT Collaborators meta-analysis despite these concerns, all that is required is a very slight change in our recommendations: In patients younger than 65 years, use simvastatin, 20 to 40 mg or equivalent, when the 10-year risk for CAD is 9% to 29%, or atorvastatin, 40 mg or equivalent, when the risk is 30% or greater.In patients aged 65 to 75 years, use simvastatin, 20 to 40 mg or equivalent, when the 10-year risk for CAD is 12% to 34%, or atorvastatin, 40 mg or equivalent, when the risk is 35% or greater.

Rodney A. Hayward, MD University of Michigan Medical School Ann Arbor, MI 48109
Potential Conflicts of Interest: None disclosed.

Should Physicians Stop Prescribing Nonsteroidal Anti-inflammatory Drugs for Postoperative Pericardial Effusion?
TO THE EDITOR: We appreciate Meurin and colleagues' (1) initiative on assessing the effectiveness of diclofenac to treat postoperative pericardial effusion.However, before implying that physicians should stop prescribing nonsteroidal anti-inflammatory drugs (NSAIDs) for postoperative pericardial effusion, some points should be considered.
The first is the choice of NSAID.The 2004 European Society of Cardiology guidelines (2) suggest ibuprofen as the preferred NSAID for treating pericarditis because of its less detrimental side effect profile, favorable effect on coronary artery blood flow, and large dose range.The only other study of diclofenac for the prevention of postoperative pericardial effusion was 1 single-center, randomized, prospective trial (3), which showed a favorable effect on the rate of significant pericardial effusion and C-reactive protein concentration in the diclofenac-treated group.However, these differences were not statistically significant.The authors acknowledged that their study was underpowered to detect the beneficial effects of diclofenac.All of this means that before concluding that NSAIDs have no role in the management of postoperative pericardial effusion, use of ibuprofen and other NSAIDs should be studied in adequately powered randomized, controlled trials.
Another set of issues to consider is the dosing frequency, duration of treatment, and the interval between surgery and initiation of NSAID treatment.In their study, Meurin and colleagues used a diclofenac dose of 50 mg twice daily.Because the half-life of diclofenac is 2 hours and the optimum serum drug concentration remains for 4 half-lives, a dosing frequency of every 8 hours may have been more appropriate.The recommended maximum dose of diclofenac is a total daily dose of 150 mg.Treatment with a tapering dose of NSAIDs is recommended for weeks to months, even after the disappearance of pericardial effusion (2).In Meurin and colleagues' study, the treatment duration of only 12.8 days may have provided a lesser therapeutic effect and thus may have unfavorably affected the primary end point.In addition, baseline echocardiography was done and diclofenac treatment was initiated at a mean of 15.9 days after surgery, raising a concern of whether the treatment was delayed rather than being initiated on day 7 after surgery, as per protocol.A median value of the interval between surgery and NSAID initiation would have answered this concern, excluding the outliers.Because of these considerations, it would be inappropriate to conclude diclofenac failure without using the maximum anti-inflammatory dose, the optimum dosing frequency, optimum treatment duration, and timely initiation of the treatment.
Finally, rather than stratifying the grades of pericardial effusion into localized or circumferential, the authors calculated the grades and localization separately (Table 2 in their article).The numbers of grade-4 effusions (23 of 98) and right atrial or ventricular loculated effusions (26 of 98) in the diclofenac group, although not statistically significant, were higher than those in the placebo group (14 of 98 and 17 of 98, respectively).Right-sided loculated effusion may lead to a higher incidence of cardiac tamponade than effusion adjacent to the left heart chambers.Therefore, it is essential to know the percentage of grade-4 effusions that were right-sided and loculated, because the results and implications may have differed.A statistically significant difference in patients with tamponade with baseline grade-3 pericardial effusion between the treatment and placebo groups (Table 4 in their article) was observed but was not discussed further, which raises a question about the benefits of diclofenac in preventing tamponade in patients with baseline grade-3 postoperative pericardial effusion.

IN RESPONSE:
It is very important not to confuse acute pericarditis and the postpericardiotomy syndrome-both of which are clinical syndromes-with postoperative pericardial effusion, which is usually asymptomatic.This confusion is why NSAIDs have been frequently prescribed to treat postoperative pericardial effusion.Indeed, if the European Society of Cardiology (1) recommends ibuprofen, it is to treat acute pericarditis and not postoperative pericardial effusion, and if a long treatment duration is recommended, it is to prevent recurrences and not to help reduce the volume of the effusions (which was the aim of our study) or to prevent the incidence of late tamponades (which usually occur before postoperative day 30).In our study, the poststudy follow-up clearly shows that a longer treatment duration would have been useless, because only 2 patients required pericardial drainage after the end of treatment.
The choice of drug and of its posology can be extensively discussed with multiple references; we chose diclofenac, a classic and effective NSAID, at 100 mg/d, a guideline-recommended dose (2) that is widely used in daily practice (because the cardiovascular risk of the drug increases substantially at a higher dose).
Nowhere did we state that our study was underpowered to detect beneficial effects of diclofenac; the sample size was calculated (172 patients) and largely reached (196 patients).We only suggested that the sample was not large enough to find small beneficial effects (that is, clinically useless in these asymptomatic patients) or to detect groups of patients in which this treatment could have been helpful (for instance, patients with an important inflammatory syndrome).
It is also incorrect to say that a previous study (3) showed a favorable effect on the rate of significant pericardial effusion.This study, which was already quoted in our article, examined prevention (not treatment) of pericardial effusion, found nothing ("there was no statistically significant difference in the size of postoperative pericardial effusion. .."), and was largely underpowered.
The echocardiographic classification we used was shown in 2 large clinical studies (4, 5) (one of which was multicentric [5]) to help predict the occurrence of late tamponade.
Finally, in a study with negative results, such as ours, it is useless and probably harmful to try to find very small postspecified sub-groups in which the NSAID could seem to be effective.For instance, contrary to Drs.Shaikh and Tighe's suggestion, right-sided loculated effusion did not lead to a higher incidence of tamponade in our study.
Medical history is filled with widely applied therapeutic habits that replicate long-standing practices based on theories that have no true scientific background.Thus, challenging physicians' lifelong habits may be good for their patients.Our study, which was properly designed and conducted, is, to our knowledge, the only one dedicated to treatment of postoperative pericardial effusion.It clearly shows the absence of effectiveness of an NSAID.We have no conflict of interest and would have been very happy to find a positive result; however, scientific facts must be admitted.sive coronary angiography" by CT and MRI, the remainder of the text and its conclusions refer less specifically only to "CT" and "MRI."This generalization opens the study to potential misinterpretation.

Philippe Meurin, MD
First, the main strength of cardiac MRI and its main clinical role in coronary artery disease (CAD) have become its ability to provide functional assessment of coronary stenosis through pharmacologic stress imaging rather than providing purely anatomical information.Single-center and multicenter studies have shown MRI to be at least as accurate as the current clinical reference test, single-photon emission CT, in detecting flow-limiting CAD (2).
Second, an accumulating body of evidence underscores the importance of functional assessment of stenosis severity.Patients without detectable ischemia on noninvasive imaging, including MRI, have a very good prognosis, whereas the presence of detectable myocardial ischemia confers a significantly worse prognosis (3).In addition, in patients with anatomically moderate disease, there is a prognostic benefit when percutaneous revascularization is targeted only at functionally significant lesions (4).Given that anatomical assessment of the severity of a coronary artery stenosis has been repeatedly shown to correlate poorly with the hemodynamic effects (5), coronary angiography provides only some of the information that is important in CAD.It is therefore questionable to describe the coronary stenoses detected by angiography as "clinically significant."The predominant view within cardiology, which bases treatment on the presence of coronary stenoses, is beginning to be replaced with an understanding that functional significance needs to be the clinical reference.Computed tomography is currently unable to provide such functional assessment.
Thus, referring to MRI throughout the article solely with reference to coronary magnetic resonance angiography (MRA) and not commenting on the functional information provided by either test seems misleading.Equally, to conclude in the abstract that "[f]or ruling out CAD, CT is more accurate than MRI" and in the Editors' Notes that "[c]omputed tomographic angiography may be a better noninvasive test than MRI for ruling out coronary artery disease" seem unjustified simplifications that ignore the emerging broader role of cardiovascular MRI in the management of CAD.IN RESPONSE: I agree with Dr. Morton and colleagues that "preoccupation with coronary lumenology" (1) prevents one from grasping the whole picture of CAD and that functional tests assessing either perfusion or wall motion during rest and stress enable the best hemodynamic evaluation of coronary artery stenosis (2).Assessment of myocardial perfusion, for example, using MRI or single-photon emission CT, plays a pivotal role in the clinical management of patients with known CAD because revascularization is most beneficial in patients with more severe ischemia (3).

Geraint Morton, MA, MBBS
However, our meta-analysis was designed to analyze noninvasive coronary angiography using CT versus MRI with conventional coronary angiography as the reference standard for detection of stenosis.In this setting, we showed that coronary CT provides a more sensitive and specific assessment of the coronary arteries than MRA.Of interest, we found a per-patient sensitivity for coronary CT angiography of 97.2% (95% CI, 96.2% to 98.0%), which was significantly higher than that reported for myocardial perfusion MRI in a meta-analysis by Nandalur and colleagues (91% [CI, 88% to 94.0%]) (2).Notably, prevalence of CAD was similar in our metaanalysis and Nandalur and colleagues' meta-analysis, at 54.7% and 57.4%, respectively.Thus, published evidence suggests that coronary CT angiography is more accurate than not only coronary MRA but also perfusion MRI in ruling out CAD.
It is important to note that recent studies show the potential of state-of-the-art multislice CT scanners to perform stress and rest myocardial perfusion imaging (4 -6), which may be an important advance toward a more comprehensive cardiac imaging test.Nevertheless, because of its unique advantage of not requiring radiation exposure, MRI is the most promising cardiac imaging test and is also more versatile, allowing assessment of myocardial infarction, fibrosis, function, and metabolism.Thus, if the diagnostic performance of MRI in coronary angiography can be further improved, this test may become the most important imaging tool for noninvasively and comprehensively assessing patients with suspected CAD.Currently, cardiac CT should be considered the foremost noninvasive test for detecting and ruling out coronary artery stenosis in selected patient populations.

IN RESPONSE:
We appreciate the lively discussion with Dr. Goebel and colleagues on their pilot randomized, controlled trial on IVIG in patients with long-standing CRPS.We did not state that CRPS is of psychogenic origin or that patients with CRPS are psychiatrically ill in the classic sense.Although previous studies did not report strong psychosomatic factors, more vigorous and prospective testing might provide different results (1), and the experience of many researchers, which we systematically collected in the national German guideline of CRPS diagnosis and treatment (2), provide some indication that stressful life events (3), latent anxiety, posttraumatic stress disorder, and selfconfidence problems are prevalent in chronic CRPS, or even mimic chronic CRPS.Because CRPS develops after trauma, which often leads to compensation claims (4), goal conflicts might also contribute to chronic CRPS, in particular because treatment requires the patient to actively participate.
Such factors cannot be assessed by routine checklists and usually cannot be treated by unimodal pharmaceutical treatment of chronic CRPS.Therefore, we remain skeptical-not only of IVIGs but also of other agents, such as ketamine, magnesium, or tadalafil.The complete discussion about placebo effects and effect sizes could be resolved only by large, multicenter, randomized, controlled trials.There are plenty of examples of promising phase 2 studies with drugs for chronic pain that did not overcome this obstacle, similar to IVIG for treatment of multiple sclerosis (5).Despite our skepticism, we hope this will not be the case in the planned future studies of IVIG for CRPS.We all need an effective treatment for chronic pain and, in particular, for chronic CRPS.each of cancer of the lung, skin, and hematopoietic system; 4 each of the breast, larynx, and colorectum; 2 each of the endometrium and kidney; and 1 each of the stomach and testis.During followup, 12 patients died (1 of cancer).

Frank Birklein, MD, PhD
Discussion: This study clearly supports the hypothesis that cancer is involved in the pathophysiology of TTC.Together with the 50 patients reported elsewhere (3), our entire cohort comprises 191 patients and is one of the largest populations with TTC reported in the literature.In this population, 23.6% had cancer, which greatly exceeds the expected prevalence of cancer in age-matched populations in the United States (8.2%) (4), Germany (11.2%), and all European countries combined (7.8%) (5).
It is possible that patients with cancer have increased basal sympathetic nervous tone and, therefore, may be more susceptible to TTC with additional stress.In addition, paraneoplastic mediators may directly alter cardiac adrenoreceptors, which would make TTC a paraneoplastic syndrome that occasionally manifests before cancer is diagnosed.
Conclusion: This study does not prove that cancer leads to TTC; however, as long as the association cannot be excluded, the high frequency of cancer should prompt physicians to consider screening patients with TTC.The wide variability of cancer entities reported in TTC (3) necessitates a thorough clinical, laboratory, and technical examination of the patient.Therefore, we strongly suggest the initiation of an international registry with special focus on the prevalence and incidence of cancer in patients with TTC to further elucidate the relationship between these diseases.

CORRECTION Correction: In the Clinic: Breast Cancer Screening and Prevention
In the In the Clinic on breast cancer screening and prevention (1), the authors listed in reference 27 are incorrect.The authors should be listed as "Nelson HD, Tyne K, Naik A, et al" in the reference list and in the footnotes for Tables 2 to 4. This has been corrected in the online version.

Figure . 7 *
Figure.Newly diagnosed cases of cancer, by year.
Potential Conflicts of Interest: Consultancies: Guerbet.Honoraria: University Hospital Schleswig-Holstein 23538 Lu ¨beck, Germany Potential Conflicts of Interest: None disclosed.