Defibrillators post-MI — a note of caution

The use of implantable cardioverter-defibrillators for the prevention of sudden death in patients who are at high risk for but have not yet had cardiac arrest (primary prevention) is gradually expanding. A number of trials have shown mortality benefit in different populations:

  • MADIT: benefit in patients with coronary disease, ejection fraction less than 35%, spontaneous nonsustained ventricular tachycardia and inducible VT with electrophysiologic testing.
  • MUSTT: benefit in patients with similar criteria to MADIT, but EF less than 41%.
  • MADIT II: benefit in patients with prior MI and EF less than 31%. Important study, since EPS no longer required.
  • DEFINITE: patients with non-ischemic cardiomyopathy, VPC’s or NSVT, EF less than 36% had a significant reduction in arrhythmic death but only a trend towards reduction in overall mortality.
  • SCD-HEFT (reported but not yet published): benefit in patients with EF ≤ 35%, both ischemic and non-ischemic, particularly in class II rather than class III CHF.
     

On the negative side, the CABG-PATCH trial, published in 1997, failed to demonstrate any benefit from the implantation of an ICD in patients with ejection fractions under 36% and an abnormal signal-averaged ECG who had just undergone coronary bypass surgery. It has been speculated that active ischemia is an important component of sudden death, and thus patients who had just been revascularized would not benefit as much from an ICD.

Another negative trial has just been published in this week’s NEJM: the DINAMIT trial. In this study from Germany, 674 patients with ejection fractions under 36% who had had an MI less than 40 days previously and demonstrated reduced heart-rate variability or persistent sinus tachycardia (poor prognostic signs related to autonomic system activation), were randomized to therapy with or without an ICD.

Although arrhythmic deaths were significantly reduced by the device, this was almost exactly compensated for by an increase in non-arrhythmic deaths, so the overall death rate was the same in both treatment arms. The authors speculate that in this group of patients, with decreased heart rate variability and a recent MI, the risk of non-arrhythmic, pump-failure death is high, so that ICD therapy ends up converting some arrhythmic deaths to pump failure deaths without affecting the overall outcome.
How do we reconcile these results with the MADIT II trial, in which post-MI patients with reduced LV systolic function benefited from ICD placement? It turns out that most of the MADIT patients had their MI in the remote past. In fact, in a subgroup analysis, the benefit from the ICD was limited to patients whose MI was over 18 months previous to trial entry. This meshes quite well with the data from the current study.

Why remote MI’s should benefit from ICD placement so much more than recent ones is a bit puzzling to me. One possibility is that, as indicated in the CABG PATCH trial, residual ischemia is an important contributor to sudden death risk. In that case, patients with recent MI’s are more likely to have been screened and revascularized when necessary than patients with older events. In MADIT II, patients were not systematically evaluated for ischemia prior to enrollment. Could it be that appropriate screening and treatment of residual ischemia would reduce the potential benefit of defibrillator placement?

D-dimer for the diagnosis of recurrent DVT

Diagnosing recurrent venous thromboembolic disease of the legs is more difficult than diagnosing a first episode of DVT. Patients who have had an initial DVT are often left with some degree of post-phlebitic syndrome, which can mimic the symptoms of recurrent disease. Duplex ultrasound often remains abnormal after a DVT, making the distinction between recurrent disease and old disease problematic.

D-dimer measurement, which is elevated with thromboembolic events, is quite sensitive for the detection of initial DVT. Although not very specific, its sensitivity is such that a negative D-dimer has good negative predictive value. Since imaging studies are problematic for diagnosing recurrent disease, the D-dimer assay should be quite useful in helping to rule out recurrent DVT. This study from Oklahoma, published in the December 7 Annals of Internal Medicine, investigated the utility of a negative D-dimer assay in excluding recurrent venous thromboembolism.

Three hundred consecutive patients with suspected recurrent DVT had D-dimer levels assessed. Those with negative results (<48 mcg/ml) did not undergo any further diagnostic testing. Those with positive results underwent compression ultrasound imaging. For three months after initial presentation, patients were followed up with imaging studies if there were any symptoms of recurrent DVT or PE, and also clinically at three months.

Of the 300 patients, 166 had positive D-dimer studies; one half of these had negative ultrasound studies, one third had positive studies and the remainder were inconclusive.

Of the 300 patients, 134 (45%) had negative D-dimer studies and did not undergo duplex scanning. In this D-dimer negative group, 11 patients returned for symptoms of recurrent thromboembolism. Of these, there were two cases of documented thromboembolism (one DVT and one PE); 4 patients had negative diagnostic tests and 5 had inconclusive studies. There was one death in the D-dimer negative group, a sudden death, which may have been a myocardial infarction. Thus, the rate of documented recurrent thromoembolism was 2/134 (1.5%). Including the sudden death and the 5 inconclusive studies, the recurrence rate was 8/134 (6%).

As the authors state:

The acceptable upper limit for the incidence of venous thromboembolism on follow-up in patients with a negative D-dimer test result remains a clinical judgment in the individual patient. The prognosis on follow-up in our patients with negative D-dimer test results is similar to the prognosis of patients with negative results on combined impedance plethysmography and fibrinogen leg scanning; it is also similar to the prognosis of patients with negative venography results.

These results, though not conclusive, are certainly useful. D-dimer testing is of utility in excluding a first episode of DVT; it is likely to be much more useful in excluding recurrent DVT, since imaging studies and symptoms are more difficult to interpret in this setting.

One aspect of this article that struck me was how often the authors mentioned the specific name of the D-dimer assay used here. It was named not only in the methods, where it is appropriate, but also in the discussion and even twice in the abstract! In the background paragraph, the authors state that the sensitivity of the [assay] … has been reported to be 96% to 100% in patients with suspected first-episode DVT or symptomatic pulmonary embolism”. And they quote two references. In fact, one of these references is a comparison of 13 assays in the diagnosis of DVT. The method used by the authors here is not among the two that were found to be most sensitive!

Better living through electricity?

Atrial fibrillation is associated with increased morbidity and mortality, in part because AF is a marker for other cardiovascular risk factors such as hypertension, valvular disease and left ventricular dysfunction. AF is also deleterious in its own right, due to the rapid, irregular heart rate, loss of atrial systole and atrial thromboembolism. Thus, there is a long history of attempting to restore and maintain normal sinus rhythm through cardioversion and anti-arrhythmic drug therapy.

More recently, several large trials have found no mortality benefit to the rhythm approach over the rate approach (anticoagulating and controlling ventricular rate), and the pendulum has started to swing towards rate control. However, an analysis of one of these trials suggested that there was, in fact, a benefit to sinus rhythm which may have been offset by mortality associated with anti-arrhythmic drug therapy. It is speculated that if sinus rhythm could be maintained without resorting to antiarrhythmics, the rhythm control approach might be beneficial after all. Thus the potential appeal of catheter-based ablation of atrial fibrillation which does not rely so much on anti-arrhythmic drugs.

In today’s NEJM, Hsu et al from Bordeaux, France investigate catheter ablation of atrial fibrillation in congestive heart failure. They studied 58 consecutive patients with at least class II CHF and echocardiographically documented ejection fractions less than 45%, who were undergoing catheter-based ablation of AF. Ejection fraction, symptoms and exercise capacity before and up to one year after the procedure were studied. These patients were also compared to 58 matched controls with normal ejection fractions undergoing the same procedure.

The authors observed substantial improvement in ejection fraction after ablation, from a mean of 36% before the procedure to 57% at one year post-procedure. NYHA class improved from 2.3 to about 1.5. Bicycle ergometer exercise time increased from 11 minutes to 14 minutes (in the control group without CHF, it increased from 14 to 16 minutes).

The authors also looked at their results among patients with adequate vs inadequate rate control before the procedure (average HR 72 vs 103). The ejection fraction improved by 23% in patients with poor rate control and by 17% in those with good rate control. In the absence of structural heart disease, EF increased by 24%; in patients with structural heart disease it increased by 16%.

The authors conclude:

Restoration and maintenance of sinus rhythm by catheter ablation without the use of drugs in patients with congestive heart failure and atrial fibrillation significantly improve cardiac function, symptoms, exercise capacity and quality of life.

I have some reservations about this study. First of all, it is not clear to me how exactly to evaluate a comparison of an ejection fraction in atrial fibrillation (the baseline) with one obtained in normal sinus rhythm. The EF is difficult to measure accurately in fibrillation; furthermore, the decrease in EF in atrial fibrillation which is due in part to tachycardia, in part to the absence of atrial contraction may not exactly reflect the actual decrease in LV contractility. Thus, I am not as impressed by the large increase in ejection fraction as I would have been if the before and after measurements had been taken in sinus rhythm.

Second, the rise in ejection fraction was less among patients with inadequate rate control before the procedure, but the authors do not state how the functional improvements (exercise time and functional status) fared in these two groups. Were they substantially less in those patients with adequate rate control?

Finally, as the authors themselves note, this was not a trial designed to evaluate mortality. But they go on to state that

Since a reduced left ventricular ejection fraction is an important predictor of mortality, the significant improvement in left ventricular function after ablation could be important in improving survival.

I would emphasize the word could here.

Note: there are three other atrial fibrillation articles in this issue. In an editorial, Stevenson and Stevenson summarize nicely the status of catheter ablation for atrial fibrillation and the potential problems and future approaches. Page reviews the approach to newly diagnosed atrial fibrillation. And there is an article on outpatient treatment of recent-onset AF with the “pill-in-the-pocket” approach.

Authors’ replies

Dr. Lucy Yardley, corresponding author for the paper on vestibular rehabilitation, has posted comments directly addressing my critiques.

Dr. Robert Belshe, corresponding author of the first of the two NEJM articles on intradermal flu vaccination, emailed several corrections to statements I had made. These are important points.

Dr. Mathew Hotopf, corresponding author of the article on medical decision making competence, posted a detailed reply that greatly clarifies points that I had raised.

EUS vs CT for pancreatic cancer

The November 16 Annals has an article from Indiana University comparing endoscopic ultrasound with multidetector CT scanning for the evaluation of suspected pancreatic cancer.

104 patients with suspected non-metastatic pancreatic malignancy underwent both EUS and high-resolution multidetector CT scanning. Of these patients, 41 were managed medically, and 63 underwent surgery. The two modalities were then compared for their accuracy in determining resectability, staging and detection of tumors.

The original hypothesis was that EUS would be better at determining resectability (in particular recognizing unresectable tumors) than CT scanning, and resectability was the study’s primary endpoint. There was no significant difference between the two modalities for this endpoint: resectable tumors were correctly identified by EUS vs CT in 88% vs. 92%. Non-resectable tumors were accurately identified by EUS vs CT in 68% vs 64%. The authors conclude that “if multidetector CT detects a pancreatic mass that seems to be resectable in an appropriate surgical candidate with suspected cancer, preoperative endoscopic ultrasonography does not seem to be necessary unless tissue confirmation of suspected malignancy is desired”.

Secondary endpoints were staging and the actual detection of tumors. In terms of overall staging, endoscopic US was superior to CT scanning, due to better recognition of T3 disease. For nodal staging, there was no significant difference.

Interestingly, EUS was superior to CT scanning in the detection of small tumors. Among the 63 patients who underwent surgery, 53 were found to have cancer. Of these 53, endoscopic ultrasound failed to diagnose 2 tumors less than 25 mm in size; CT scanning missed 10 tumors, 9 of which were less than 25 mm and one greater than 25 mm. Most of these were in the head of the pancreas. The authors do not specifically explain how 10 patients who had no pancreatic masses on CT scanning came to be evaluated for suspected pancreatic cancer, but 4 of them had had biliary stents placed, so presumably biliary obstruction was the presenting symptom for many of these CT-negative cases.

Thus, endoscopic ultrasound was not superior to CT scanning in determining resectability of pancreatic tumors but was superior in picking up small (less than 25 mm) masses that turned out to be cancers at surgery. This study, supported by grants from the American Society of Gastrointestinal Endoscopy, does not make endoscopic ultrasound an obligatory part of the resectability work-up, which was presumably the original intent. It certainly does raise the question of the role of EUS in the work-up of patients with possible pancreatic neoplasms, however, since it was able to pick up small tumors that escaped CT detection.

Stretching the flu vaccine?

In this flu-vaccine shortage year, any means of extending the vaccine supply is worth examining. There is evidence that giving vaccines intradermally rather than intramuscularly is more effective. In last week’s NEJM are two articles examining the effect of administering a lower dose of the flu vaccine intradermally. There is also a letter to the editor looking at the effect of simply giving a lower dose of the vaccine by the traditional i.m. route.

In the first study, sponsored by Glaxo SmithKline, 238 subjects were randomized to either a standard 0.5 ml i.m. dose of flu vaccine or a 0.1 ml intradermal dose of a candidate vaccine that was double concentrated (yielding an intradermal vaccine with 40% the concentration of antigen of the standard vaccine). 130 subjects were aged 18-60 years, 108 were over 60 years of age (average age: 69). In the younger group, the antibody responses were almost the same and were adequately protective in the intradermal group. In the group of subjects over 60 years old, the intradermal route yielded antibody titers that were generally lower, significantly lower for one strain and probably only incompletely protective.

In the second study, suppored by NIH grants, 100 subjects 18-40 years of age were randomized to receive either a standard 0.5 ml i.m. dose of a standard flu vaccine, or 0.1 ml of the same vaccine given intradermally. The intradermal route yielded, overall, a similar (for some strains better) antibody response as the i.m. route.

In the letter to the editor, a group from Canada gave healthy volunteers aged 18-40 years old either a full dose or a 1/10 dose of an influenza vaccine by the intramuscular route. Those given the 1/10 dose achieved protective antibody levels (although lower titers for most strains).

These studies indicate that the intradermal route produces better immunogenicity than the i.m. route. This could be used to enhance the immune response in populations that mount a less effective response (such as the elderly). It could also be used to stretch the vaccine supply in case of future shortages, particularly in the younger population. The practical effectiveness of these approaches will depend, at least in part, on FDA approval of different dosages and routes. It seems unlikely that a manufacturer of a conventional influenza vaccine will apply for approval of one-fifth of the dose of that same vaccine, administered intradermally. The Glaxo SmithKline approach of manufacturing a different product is more likely to be financially attractive.
 

Six articles on postmarketing drug surveillance from JAMA

The pharmaceutical industry and the FDA are being battered by testimony and discussion in the wake of Vioxx. Presumably this will soon die down: news stories appear, flare up and soon fizzle out when their energy (and the public’s interest) is spent. For now, however, the drug surveillance story is still burning bright. David Graham, the gaunt FDA whistle-blower, is enjoying his Warholian fifteen minutes of fame. And this week’s online edition of JAMA pre-publishes six pieces that will appear in next week’s print edition. These articles are about Baycol (cerivastatin), a drug no longer in the news but relevant to the topic. Very briefly summarized:

  • The FDA’s Graham et al examine the incidence of hospitalized rhabdomyolysis in patients treated with lipid-lowering drugs, an inception cohort study of 252,460 patients treated with various statins and fibrates between 1998 and 2001. The number of patients needed to treat for one year to produce one case of severe rhabdomyolysis was roughly 23,000 for monotherapy with atorvastatin, pravastatin or simvastatin; 3500 for monotherapy with a fibrate, 2000 for monotherapy with cerivastatin, 1500 for fibrate plus a statin other than cerivastatin and only about 10 for fibrate plus cerivastatin.
     
  • Psaty et al discuss the potential for conflict of interest in the evaluation of suspected adverse drug reactions. They discuss the FDA’s regulatory process, noting that: “…the proportion of new molecular entities that are first introduced in the United States has increased from 2-3 % in the early 1980s to 60% in 1998” and “in contrast to the highly structured premarketing evaluation, postmarketing surveillance has little structure” (and is underfunded).
     
    They describe the information that was publicly available on Baycol, mainly from SADR’s, and contrast this with the information that was available to Bayer but not to the public (which they derive from litigation documents, in which they were expert witnesses for the plaintiffs). They suggest that there was information posessed by the company which should have been made public, but which represented a conflict of interest for Bayer.
     
    The authors argue that, given the speed with which new agents enter the market, the public is at increased risk due to the lack of resources available for post-marketing surveillance and the conflict of interest inherent in the pharmaceutical companies’ positition. This should be dealt with by strengthening the post-marketing surveillance structures and possibly creating an independent agency to deal with this.
     
  • Pierkowski responds to this article on behalf of Bayer, emphasizing that Psaty et al were expert witnesses against Bayer. He argues that both the FDA and Bayer took appropriate actions at the appropriate times.
     
  • Psaty et al then respond to Pierkowski’s points, arguing that whether or not Bayer took appropriate action is less important than whether or not the post-marketing system needs to be dramatically improved.
    “While Piorkowski, as an attorney representing Bayer, properly defends some of the company’s specific actions, the purpose of our article was to raise an important public health issue. For us, the cerivastatin-rhabdomyolysis case report served as an illustration. We were primarily concerned to demonstrate how the current postmarketing surveillance system and the current FDA regulations may not, under certain circumstances, be adequate to protect the health of the public.”
  • Strom reviews the whole post-marketing surveillance system. His manuscript was submitted to JAMA by Bayer. He notes that much data is supplied by the SADR system, but the data itself is, necessarily, of limited quality and must be considered hypothesis generating. It needs to be followed up by appropriate hypothesis testing, which is where the system can and should be improved. He believes that, from a regulatory standpoint, the FDA is up to the task if given the appropriate resources. From a scientific standpoint, other organizations (such as the publicly funded Centers for Education and Research in Therapeutics) need to be developed and strengthened in order to analyze and respond to hypotheses generated from the SADR’s. This cannot be left up to the pharmaceutical industry, given its inherent conflict of interest.
     
  • Finally, in an excellent editorial, JAMA editors Fontanarosa, Rennie and DeAngelis sum it all up nicely. If you only read one of the six articles, I would suggest reading this one. The authors provide the background to the articles published in this issue, and make several suggestions, including decoupling the drug-approval and post-marketing surveillance processes, since “it is unreasonable to expect that the same agency that was responsible for approval of drug licensing and labeling would also be committed to actively seek evidence to prove itself wrong”. They conclude:
    “The postmarketing surveillance system requires a long overdue major restructuring. Until that occurs—as indicated by the articles in this issue of JAMA, as epitomized by recent evidence of serious harms from widely used and heavily promoted medications, as demonstrated by the influence of industry over postmarketing data, and as illustrated by the lengths to which some manufacturers will go to protect their interests—the United States will still be far short of having an effective, vigilant, and trustworthy system of postmarketing surveillance to protect the public.”

That being the case, what is the practitioner to do? Assuming that there will continue to be delays before adverse drug effects are discovered and publicized, prudence before prescribing the latest “hot” drugs seems warranted, more than ever. About a year ago, the pharmaceutical rep promoting one of the newer statins asked me when I would be comfortable prescribing this particular medication. I said I would wait at least a year. Now that same representative is reminding me that it’s been a year. I’m still not ready.

The vitamin E study

I just got around to looking at the vitamin E study that was publicized at the AHA meetings two weeks ago and published online at the Annals website (to be published in print in January).

It is a meta-analysis of randomized trials of vitamin E, that looks at overall mortality. Trials were analyzed by vitamin E dosage. There were 8 low-dose trials where the vitamin E dose was less than 400 IU daily, with a total of 95,000 subjects; there was no effect on mortality. Pooling the results of the 11 trials (about 41,000 subjects) where the dose was 400 IU or greater, there was an increase in mortality in the vitamin E group. The actual calculated number was 63 deaths per 10,000 persons.

My take on this:

The actual derived number is not meaningful, since results are pooled across a heterogeneous group of subjects, methods and doses. The number 63 per 10,000 should not be taken as is, especially since the confidence interval is 6-119!

On the other hand, the graphic representation of the 19 trials is highly instructive (a picture really is worth a lot of words). Even if one doesn’t place any stock in the pooling of results, just looking at the graphic summary of the trials is very convincing. I believe this is the biggest value of metanalyses: providing a graphic summary of multiple studies.

Since the inflection point for increased risk seems to be around 400 IU daily, and since that is a rather common vitamin E dose, it’s not clear exactly what to think about that dose. But I do feel fairly comfortable telling people who ask about 400 IU of vitamin E or more that I wouldn’t recommend it for prevention and would consider stopping it if they are taking this dose.

Ultrasound enhanced tPA for stroke

An article in yesterday’s NEJM on Ultrasound-enhanced systemic thrombolysis for acute ischemic stroke was the first I had seen on thrombolysis for CVA in quite a while (the last major trial on t-PA for stroke published in the NEJM, which I reviewed here, was 9 years ago). And a quick PubMed search yielded this study from Cleveland, which indicates that only about 2% of strokes nationwide receive t-PA, mainly because of the 3 hour time requirement from onset of symptoms.

Ultrasound energy has been shown to facilitate the activity of fibrinolytic agents. Experimental work with lower frequency (Kilohertz) ultrasound in conjunction with t-PA had previously shown an increased risk of hemorrhage. This effect has not been noted with higher frequency (MHz) transcranial diagnostic ultrasound which has been used to look at the patency of cerebral arteries. The current phase II trial was designed to examine the effects of continuous, high frequency transcranial doppler monitoring, in conjunction with t-PA, on hemorrhage, patency and recovery rates.

126 stroke patients who presented early with evidence of abnormal flow through the middle cerebral artery were randomized into two groups of 63 patients. Both groups received t-PA, and both groups had transcranial doppler measurements that checked the flow at 0, 30, 60, 90 and 120 minutes. Patients in the target, ultrasound group had diagnostic ultrasound insonation that was continuous for the two hours; patients in the placebo group only received the diagnostic ultrasound at the prespecified times.

Patients in the target ultrasound group did better, in terms of both recanalization and clinical improvement (NS, however), without any difference in intracranial hemorrhage:

Continuous ultrasound Placebo ultrasound
Complete recanalization within 2 hours 46% 18% (p<0.001)
Clinical recovery within 2 hours 29% 21% (NS)
Favorable outcome at 3 months (modified Rankin score of 0 or 1) 42% 29% (NS)

These results are encouraging and presumably more studies will be undertaken to see what the optimal ultrasound approach should be (energy level, duration of therapy). If this pans out, the benefit of thrombolytic therapy would be more convincing than it is at present, and might bolster enthusiasm for its use. The addition of another technological requirement to stroke management would complicate matters somewhat, and might make a more effective case for centralizing stroke treatment.

Most fascinating is a report, cited by Medpundit, that indicates a possible beneficial effect of ultrasound alone in the management of stroke! This comes from an anecdotal report by a physician in Scotland.

Time to reconsider atenolol?

In the November 3 Lancet is a meta-analysis entitled “Atenolol for hypertension: is it a wise choice?“.

When compared to placebo for the treatment of hypertension (4 trials with about 3,000 patients total), there was no benefit for atenolol in terms of mortality, cardiovascular mortality or MI. There was a benefit for stroke, but this was mainly from one trial in which atenolol was not used as monotherapy in most patients.

In 3 trials with about 7,000 patients that compared atenolol to other anti-hypertensives, atenolol fared worse than the other medications for all parameters except for MI, where it was about the same.

The authors speculate that the non-lipophilic nature of atenolol (unlike metoprolol and others) might play a role, as might a lesser or absent influence on LVH, compared with other medications. They note that there have been no good trials comparing different beta-blockers in hypertension, thus this need not apply to other beta-blockers. And, of course, beta-blockers have an important role to play in the setting of ischemic heart disease, CHF and arrhythmias.

I’m not a big fan of meta-analyses, in general, but this one is intriguing, and does make atenolol seem much less attractive for hypertension (and perhaps for other indications as well).