A 75-Year-Old Man With Congestive Heart Failure

DR DELBANCO: Mr C is a 75-year-old retired fisherman who came to the United States from the West Indies more than 10 years ago. He and his wife live in Boston, Mass, with their daughter, who has been an important source of support for him. He has Medicare and supplemental commercial insurance.

In 1988, Mr C developed exertional dyspnea and was diagnosed as having congestive heart failure and chronic obstructive pulmonary disease (COPD), with a forced expiratory volume in 1 second of less than 0.5 L. Chest radiographs showed a large cardiac silhouette, with hyperlucent lung fields and flat diaphragms. Electrocardiogram results showed poor R-wave progression and were consistent with left ventricular hypertrophy and left atrial enlargement. Echocardiogram results showed severe left ventricular dilatation, diffuse hypokinesis, ejection fraction of less than 0.20, and mild right heart dilatation, with 3+ mitral regurgitation and 2+ tricuspid regurgitation. He also had chronic mild renal insufficiency, with a creatinine level of about 221 µmol/L (2.5 mg/dL), and a history of gout, with erosive arthritis, nephrolithiasis, and uric acid stones. There is no known family history of heart disease. Cardiac risk factors include long-standing mild hypertension and 50 pack-years of cigarette smoking, ending in 1988.

Mr C was treated initially with captopril, furosemide, digoxin, aspirin, and isosorbide. From 1990 to 1995, he was relatively stable, requiring hospitalization for heart failure once or twice a year. A search for the cause of his dilated cardiomyopathy revealed no history of alcohol abuse and normal serum iron, iron binding capacity, ferritin, and thyroid function test results. There was no history or sign of malnutrition. In 1993, he developed diabetes mellitus and was treated with dietary restrictions and oral hypoglycemic agents. In 1996, Mr C developed episodes of intermittent atrial fibrillation and flutter. While in sinus rhythm, he was noted to have second-degree atrioventricular (AV) block, with a Wenckebach pattern. Echocardiogram results showed increasing biventricular dilatation, with a left atrial size of 5.2 cm. Aspirin was replaced by warfarin. Mr C's renal insufficiency worsened, hyperkalemia developed, and captopril was discontinued. Therapy with hydralazine hydrochloride and amlodipine was initiated.

In the past 3 years, Mr C has been hospitalized every 3 to 4 months. His functional status is significantly limited. He can walk slowly around his apartment but cannot climb stairs. He experiences chest pain a few times per week, while walking or at rest, that usually responds to sublingual nitroglycerin. His dyspnea worsens when his weight increases by more than 0.9 kg. He uses oxygen both at rest and with physical activity. Attentive care by his family members and visiting nurse services makes it possible for him to continue living at home.

On physical examination, Mr C appears chronically ill and weak. He does not cough. He is normotensive, with distant heart sounds and a soft left sternal border systolic murmur that radiates to the left axilla. There is no hepatojugular reflux or peripheral edema. Recent electrocardiogram results showed a sinus rhythm, with ventricular ectopy. Recent 24-hour cardiac monitoring showed asymptomatic runs of ventricular tachycardia. An arterial blood gas study revealed a partial pressure of oxygen of 69 mm Hg and a partial pressure of carbon dioxide of 51 mm Hg, with venous carbon dioxide of 36 mg/dL.

Mr C's current medications include hydralazine hydrochloride, 100 mg twice daily; amlodipine, 5 mg/d; extended-release isosorbide, 90 mg/d; digoxin, 0.125 mg/d; bumetanide, 2 mg twice daily; metolazone, 2.5 mg as needed when his weight increases by more than 0.9 kg; warfarin, 5 mg/d; glyburide, 10 mg/d; colchicine, 0.6 mg/d; allopurinol, 100 mg/d; several bronchodilators that he inhales intermittently; oxygen; and stool softeners.

[Editors' note: The patient spoke little English, and during the interview his daughter spoke for him.]

My dad started having short breath. He ignored the symptoms until they became really bad. Then he went to the doctor, and the doctor told him he was sick and gave him the medication. The problem is that in our country, the medications are not as improved as in the United States.

Heart failure to me sounds basically fatal. When I hear the words heart failure, it scares me. We've been living with it for a number of years now, so it's not as scary as when I first heard the words and thought, well, this person's going to drop any second. However, over the years, I've come to the understanding that it can happen. But they can also live with it. I'm not as afraid of the words as I used to be.

Nurses come to the house. Generally, I've learned from them what to look for when he is not doing too well. For the most part, the insurance has been reasonably good. However, I have a problem with them sometimes, when they are ready to cut off nursing care. As a result, I end up taking my dad to the emergency room. Sometimes, they want to give him 2 or 3 or 5 visits from the nurses. He needs more than 5 visits. Because the insurance is not there looking at my dad, I think they're too quick to stop his care. Then we end up in the emergency room. If the insurance company would think about it, if home health aides or nurses came to the house, it would save [the insurance company] lots of money.

Mr C's congestive heart failure and COPD are both severe, which limits his activity to walking around. He becomes out of breath climbing any stairs, even carrying a small canister of oxygen with 2 or 3 L flowing. He frequently gets shortness of breath when his weight goes up by even 2 lb. He can't really do much, and he is taking a long list of medications. His daughter has been the major caretaker in helping arrange all the medications. The severity of his disease affects not only his life but also his daughter's life.

Treating Mr C has been challenging for many reasons. He has comorbid conditions that limit our options for his medications. Because of severe COPD, we cannot really use β-blockers, even though we know his congestive heart failure is most likely due to coronary artery disease. The question of what to do about his arrhythmia keeps haunting me. Should I give him another medication, with his already-complicated regimen? I feel like that seems to be treating my mind more than treating his disease.

I'm not sure if Mr C realizes how risky this condition can be. He's always very polite, obliging, and respectful to his physicians, although he doesn't participate in the decision making that much. He just says, "Whatever you say, doctor." Even when he comes in very sick, his most frequent response to the question, "How are you doing, Mr C?" is "Not too bad," or "Not too good." That makes it challenging to detect any early signs of deterioration in his underlying condition.

What are the etiology, prevalence, and societal burden from congestive heart failure? How does one make the diagnosis in 1999? What treatments are available, what are the risks and benefits, and how should clinicians evaluate them? What are the roles for angiotensin-converting enzyme (ACE) inhibitors, β-blockers, and inotropic agents? What are your recommendations for Mr C?

DR GUYATT: Following the tenets of evidence-based medicine,¹ my discussion will highlight results from randomized trials and insights from controlled trial methods that can guide us in managing patients with heart failure. Although asked to focus on Mr C's heart failure, as a general internist I note that Mr C suffers from severe chronic airflow limitation (CAL), with a forced expiratory volume in 1 second of less than 0.5 L. This degree of airflow obstruction is sufficient to explain Mr C's exertional dyspnea even without concomitant heart failure. Thus, deciding on the extent to which Mr C is benefiting symptomatically from his respiratory and cardiac medications will be a challenge.

Mr C uses home oxygen. Randomized control trials (RCTs) have demonstrated that home oxygen reduces the mortality of hypoxemic patients with CAL² and that continuous oxygen results in larger mortality reductions than nocturnal oxygen alone.³ These trials enrolled patients with evidence of cor pulmonale whose resting arterial partial pressure of oxygen was less than 60 mm Hg. The severity of Mr C's CAL and heart failure suggests that he almost certainly meets these criteria.

Mr C is using a trio of inhalers often prescribed to patients with CAL: a β-agonist, an anticholinergic agent, and a steroid. While this combination is standard management in numerous settings, many patients will not receive incremental benefit from using more than 1 inhaled drug.⁴ I will revisit this issue in discussing its implications for Mr C's heart failure.

Studies of heart failure incidence are limited by lack of agreement regarding definition, as well as variability in methods and criteria and a paucity of population-based studies. Estimates of heart failure incidence vary from 1.0 to 5.0 per thousand persons per year, increasing substantially with older age to a level as high as 40 cases per thousand in those older than 75 years.⁵ This corresponds to a prevalence of 1% in those aged 50 to 59 years and 10% in those older than 80 years.⁶ Only about 5% of symptomatic patients have the severe heart failure from which Mr C suffers.⁷

In more industrialized countries, ischemic coronary disease and hypertension account for most cases of heart failure.^{8 - 9} Valvular disease and viral, idiopathic, and alcoholic cardiomyopathy are also not rare. Clinicians appropriately bear in mind uncommon but treatable causes, such as thyroid disease and hemochromatosis. Consistent with Mr C's experience of multiple medications, hospitalizations, and home care requirements, the yearly cost of treating heart failure patients may be as high as $22,000.¹⁰ Table 1 lists approximate charges for medications for congestive heart failure.

Classic heart failure symptoms of dyspnea, orthopnea, and paroxysmal nocturnal dyspnea lack specificity; many patients with other conditions report identical symptoms.¹¹ The same is true for the most easily detected physical sign, peripheral edema.⁵ This probably explains the findings of 2 studies that examined the accuracy of the diagnosis of heart failure in primary care and found false-positive rates of 47%¹² and 55%.¹³ Physical findings of a third heart sound, elevated jugular veins, and pulmonary crackles are more specific,¹¹ although the last of these is prominent in 1 easily confused condition, pulmonary fibrosis. The diagnosis may be more difficult in older patients with heart failure, of whom 30% to 40% have diastolic dysfunction.¹⁴ The chest radiograph is helpful, despite the poor relationship between heart size and left ventricular function and between upper-lobe distention and pulmonary capillary wedge pressure.¹⁵ The limitations of history taking, physical examination, and chest radiographs lead experts to recommend routine echocardiography in patients with possible heart failure.

Dr M finds herself in a quandary with Mr C's family. Consistent with her understanding of his prognosis, she has presented a grim picture, yet Mr C pulls through his exacerbations repeatedly. A clear picture of the prognosis for heart failure is difficult because studies almost invariably classify patients according to New York Heart Association (NYHA) functional class, yet the between-observer agreement in NYHA class is poor.¹⁶ Evidence suggests that neither the incidence nor the prognosis of heart failure has changed dramatically over the last 2 decades.¹⁷ Despite limitations of the studies, 1-year survival rates in older patients with heart failure as severe as Mr C's are no higher than 60%.¹⁸ Mr C has beaten the odds, and we hope he continues to do so.

We offer our patients treatment for 2 reasons: to help them live longer or to make them feel better. Uncontrolled observations, or observations from studies of weak design, may provide misleading estimates of treatment efficacy. Ten years ago, we were offering heart failure patients prophylactic antiarrhythmic agents that we now know led to premature death and withholding β-blockers that make them live both longer and better lives. The cardiology community was divided between vociferous advocates of digoxin and equally vociferous opponents. Only rigorously conducted RCTs (ie, double-blind, achieving complete patient follow-up, and conducting an intention-to-treat analysis¹⁹ ) have opened our eyes to the true benefits and harms of these interventions.

Randomized controlled trials are inevitably limited by random error or noise and by the particular patient population and intervention that the investigators choose to study. Meta-analyses that pool results across RCTs provide a powerful tool to help deal with both these limitations. To avoid bias, meta-analyses must themselves be rigorously conducted. For these systematic reviews, avoiding bias requires eligibility criteria that are both explicit and appropriate (ie, across the range of pooled patients, interventions, outcomes, and methods, we expect more or less the same treatment effect) as well as a comprehensive and reproducible search for relevant studies.²⁰ Furthermore, investigators must evaluate the methodological quality of the original studies; poorly designed studies will yield biased estimates, irrespective of the rigor of the meta-analytic methods. Finally, meta-analysts must look carefully to see if study results are similar across different patients, interventions, measurements of outcome, and study methods.²¹

Meta-analyses are limited by the inclusion criteria of the RCTs from which they draw their data. For example, almost all RCTs studying heart failure would exclude Mr C because of his severe CAL. When deciding whether to apply the results of RCTs to groups of patients who did not participate, we should ask if there is a compelling biological reason why the treatment effect might differ in these groups.^{22 - 23} In addition to his CAL, Mr C is older than most patients in clinical trials and has more severe renal failure. While clinicians must make their own judgments, the biology associated with these differences is, in my view, unlikely to lead to major differences in the effect of the relevant interventions.

Because they must always trade off benefit and harm, clinicians cannot be satisfied with knowing whether treatments work. They also must attend to the magnitude of the benefit patients can expect. The relative risk reduction (RRR) provides 1 way of expressing the extent to which treatments reduce the relative risk of an adverse outcome (for instance, they may cut the risk by half or by one quarter). The RRR may, however, be misleading. In patients with a 2% risk of dying, a treatment with an RRR of 50% will reduce the risk to 1%. The situation is very different when the patient's risk of dying is 40%, when the same treatment will cut the risk to 20%. Thus, the first patient experiences an absolute benefit of 1% and the second, 20%. A compelling way to present such results uses the number needed to treat (NNT), the number of patients we must treat for a particular time to prevent an adverse outcome, such as death, stroke, or myocardial infarction. For example, we would need to treat 100 of the low-risk patients described to save a life, but only 5 of the high-risk patients.²⁴

A rigorously conducted systematic review of RCTs has examined the effect of ACE inhibitors in heart failure.²⁵ The investigators found that ACE inhibitors reduce mortality by approximately a quarter (odds ratio [OR], 0.77; 95% confidence interval [CI], 0.67-0.88). The effect was consistent across all NYHA functional classes and all ACE inhibitors. A conservative estimate of 1-year mortality for patients such as Mr C, 40%, implies the need to treat approximately 10 patients for a year with an ACE inhibitor to save a single life. Clinicians can be confident in the applicability of these findings if they prescribe the relatively large dosages of ACE inhibitors used in the RCTs (captopril, 100 to 150 mg/d; enalapril, 20 mg/d).²⁶ Unpublished results from the first large RCT comparing different dosages of an ACE inhibitor suggest a benefit of larger dosages on the combined end points of death and deterioration due to heart failure.²⁷

Dr M prescribed an ACE inhibitor for Mr C, but the patient experienced increasing renal dysfunction and hyperkalemia. Decreasing the dosage of diuretics will sometimes ameliorate or prevent renal dysfunction accompanying an ACE inhibitor, but the combination of increasing creatinine and serum potassium levels suggests this is unlikely in Mr C's case. Given Mr C's intolerance of an ACE inhibitor, what are the options? A single RCT, the ELITE trial, has suggested that an angiotensin II receptor blocker, losartan potassium, may lead to a larger mortality reduction than an ACE inhibitor, captopril (RRR, 46%; 95% CI, 5%-69%).²⁸ These findings were unexpected and require replication. Moreover, the ELITE trial found a similar incidence of renal dysfunction with both agents. Nevertheless, Dr M may wish to consider a trial of losartan for Mr C.

Because Mr C could not tolerate an ACE inhibitor, Dr M chose an alternative, the combination of hydralazine and an oral nitrate. The evidence supporting this combination comes from a single RCT in which 34.3% of 273 heart failure patients randomized to placebo and 25.6% of 186 patients treated with hydralazine and nitrates were dead after 2 years of follow-up.²⁹ The associated RRR of approximately one third was on the borderline of conventional levels of statistical significance, and the usually preferable survival analysis failed to cross the conventional P=.05 boundary. A subsequent study demonstrated a mortality reduction with an ACE inhibitor compared with the hydralazine-nitrate combination.³⁰

Diuretics are a mainstay of treatment in heart failure patients, and their dramatic impact on symptoms have obviated the necessity for RCTs. Loop diuretics are usually required, though their effectiveness decreases as heart failure severity increases, particularly in patients with decreased renal function.^{31 - 32} When loop diuretic resistance becomes sufficiently severe, patients require the addition of a thiazide or spironolactone.³³ Metolazone is popular in North America, though its continuous use tends to be associated with hyponatremia, hypochloremic alkalosis, and hypokalemia.³⁴ It is best used in the way Dr M has recommended to Mr C, as intermittent treatment in response to weight gain caused by fluid retention.

After 2 centuries of digoxin use, RCTs have finally yielded a clear picture of the drug's effect in heart failure patients in sinus rhythm. A recent large RCT, the Digitalis Investigation Group (DIG) study,³⁵ has confirmed the findings of a meta-analysis of small RCTs prior to 1990³⁶ and 2 subsequent RCTs of digoxin withdrawal^{37 - 38} that suggested that digoxin can reduce the incidence of heart failure exacerbations. After enrolling more than 6500 patients and achieving an average follow-up of 37 months, the DIG study found nearly equal mortality in treated and control patients, but a reduction in heart failure hospitalizations with digoxin (26.8% vs 34.7%; RRR, 0.72; 95% CI, 0.66-0.79). Given his hospitalization history, Mr C's risk of admission for heart failure in the next year might be as high as 90%. Were this so, one would need to treat only 3 patients like Mr C for a year with digoxin to prevent a hospitalization. Digoxin should be used cautiously in patients with a tendency toward bradycardia or in those who, like Mr C, have some degree of AV heart block.

While acknowledging the possible contraindication to β-blocker therapy for Mr C because of his CAL, Dr M asked about the role of β-adrenergic blockers in heart failure. In the last 3 years, 5 meta-analyses of β-blockers in heart failure have been published.^{39 - 43} While differing in details of methods and results, all 5 are sound and yield consistent findings. While increasing the ejection fraction, β-blockers reduce the RR of dying by about 30%, with a lower boundary of the 95% CI at 10% to 15%. Patients with ischemic and nonischemic cardiomyopathy can expect similar benefits. In patients with severe heart failure, of whom 40% might die in a year, the clinician must treat 8 to 9 patients with β-blockers for 1 year to save a life. Like digoxin, β-blockers can exacerbate tendencies to bradycardia or increase the level of AV block in those with AV nodal disease.

Dr M's biggest concern about care for Mr C is the possible consequences of the frequent nonsustained ventricular tachycardia (NSVT) she has observed on his Holter monitor. Her concern is well justified. In the Framingham Heart Study, approximately 50% of all deaths in heart failure patients occurred within 1 hour of onset of symptoms,⁴⁴ and sudden death continues to account for half the mortality in recent trials of pharmacological therapy in heart failure.⁴⁵

In the past, cardiologists and internists often responded to Holter monitor findings of NSVT in patients with coronary artery disease or heart failure by prescribing class 1 antiarrhythmic agents that include quinidine, procainamide hydrochloride, disopyramide, mexiletine hydrochloride, encainide hydrochloride, flecainide acetate, and moricizine. Despite their ability to obliterate nonlethal arrhythmias, individual trials of the latter 3 drugs have shown a paradoxical statistically significant increase in sudden death.^{46 - 47} Moreover, a rigorously conducted meta-analysis of all 51 RCTs of class 1 agents confirmed an increase in mortality (OR, 1.14; 95% CI, 1.01-1.28).⁴⁸

In 1997, the investigators of all 13 RCTs of a class 3 agent, amiodarone, reported a meta-analysis that combined individual patient data from their trials.⁴⁹ Of these studies, 5 were undertaken in heart failure patients and 8 in patients who had myocardial infarction. Their pooled analysis showed an RRR for sudden death of 29% (95% CI, 15%-41%) and for total mortality of 15% (95% CI, −2% to 29%), with consistent effects in the post–myocardial infarction and heart failure trials.

A conservative estimate of Mr C's risk of a lethal arrhythmia in the next year of 20% suggests that we would need to treat approximately 17 similar patients for a year to prevent a single death. In treating 100 patients, we would prevent approximately 6 arrhythmic deaths, while causing 6 patients to develop hypothyroidism, 1 patient to develop hyperthyroidism, 2 patients to become bradycardic, and 1 patient to develop pulmonary infiltrates.⁴⁹ Given his lung disease, the last risk may be particularly relevant to Mr C.

The results of an RCT suggest that prophylaxis with an automatic implanted cardiac defibrillator can reduce mortality in patients with coronary artery disease and NSVT compared with drug therapy.⁵⁰ While Mr C's age and other medical problems argue against therapy with an automatic implanted cardiac defibrillator, it offers another option that Dr M might consider.

Dr M's final question concerned anticoagulation. Mr C began to take warfarin after a Holter monitor showed intermittent atrial fibrillation. Given his age and the severity of his heart failure, a conservative estimate of Mr C's 1-year risk of an embolic stroke is 8%.⁵¹ A systematic review of 5 RCTs of warfarin in patients with nonvalvular atrial fibrillation showed a 68% RRR, with a CI of 50% to 79%. Mr C has now been taking warfarin for several years without a major hemorrhage, so his risk of serious bleeding in the next year is 1% or less.⁵² Thus, in 100 patients like Mr C treated for 1 year, 1 patient will have a serious hemorrhage and 5 to 6 patients will avoid a stroke they otherwise would have had.

Were a patient in stable sinus rhythm, which it seems Mr C has been for some time, the decision about anticoagulation becomes more difficult because of the lower risk of thromboembolism and, more important, the lack of an RCT directly addressing this question.⁵³ A decision to treat such patients therefore depends on observational studies and generalization from trials in patients with atrial fibrillation. A decision analysis addressing this issue suggested that patients like Mr C, who find little problem with taking warfarin and whose hemorrhage risk is low, benefit from anticoagulation.⁵⁴ A conservative "do no harm" approach would withhold anticoagulants pending the conduct of an RCT directly addressing this issue.

Heart transplant provides a last resort option for some patients with severe heart failure. Mr C's severe lung disease and age rule out this possibility.

The last decade has seen a growing interest in exercise rehabilitation in patients with heart failure. A number of observational studies and a few small RCTs suggest that exercise in these patients is safe and can improve exercise capacity.⁵⁵ Larger RCTs are needed to examine the impact on quality of life, and much larger trials are needed to determine effects on hospitalizations or mortality. For Mr C to participate, a dedicated heart failure rehabilitation program would have to be both available and covered by his insurance plan.

As Dr M has found with Mr C, organizing care for a patient with severe heart failure is both challenging and crucial. Rich and colleagues⁵⁶ conducted an RCT of a nurse-directed multidisciplinary intervention that included patient education, dietary instruction, social service consultation, medication simplification, and intensive follow-up, including use of home health services, individualized home visits, and ongoing telephone contact. The control group, which received conventional community care through primary care physicians, showed poorer levels of compliance and rates of hospitalization sufficiently higher that the intervention actually saved $460 per patient during the 90 days of follow-up.⁵⁷ The high readmission rates in both groups (54% in treated patients and 64% in controls) attest to the challenges of managing patients with severe heart failure.

These results emphasize the importance of compliance-enhancing maneuvers and intensive follow-up in patients like Mr C. A recent systematic review of 116 RCTs of compliance-enhancing maneuvers further reinforces the message.⁵⁸ No single category of intervention was superior, but comprehensive interventions combining education (individual or group; oral, written, or audiovisual presentations), behavioral (packaging, mail, or telephone reminders) and affective (home visits and family support) components were more effective than single-focus interventions. The recent reduction in the frequency of Mr C's hospital admissions may represent successful efforts to enhance his family's participation in Mr C's care, and, as his daughter alludes to, the impact of home care visits in rapidly dealing with new or recurrent problems.

In my discussion, I have placed great emphasis on RCTs and meta-analyses of RCTs. Heterogeneity of patient response limits the usefulness of positive RCTs: is our patient a nonresponder to an intervention that benefits only some of the target population? For interventions directed at preventing distant events, such as death or hospitalization, identifying individual responders and nonresponders is never possible, and we must treat the entire population. However, when the question concerns symptomatic benefit or possible adverse effects, individualizing therapy becomes a possibility.

Clinicians traditionally respond to the latter situations with trials that institute or withhold therapy. Dr M conducted such a trial when, in response to increasing creatinine level and hyperkalemia, she withdrew the ACE inhibitor and found that renal function and electrolytes stabilized. She and the patient's daughter have used metolazone in a similar fashion, demonstrating its helpful diuretic effect. Often, however, bias may intrude in deciding whether there is a causal connection between the intervention and the target outcome. Particularly when symptoms are the issue, natural history, expectations, placebo effects, and the patient's desire to please may all contribute to incorrect inferences.⁵⁹

My colleagues and I have refined a method for making strong inferences about intervention effects in individual patients.⁶⁰ The approach, which we call N-of-1 RCTs, involves pairs of treatment periods in which the patient receives active treatment during one period of each pair and a matching placebo during the other period (Figure 1). The treatment target, most often a symptom or group of symptoms, is monitored quantitatively during each period. Pairs of treatment periods are repeated until both the patient and clinician become convinced of the treatment's effect or lack of effect. Using the full method is challenging, but individual components can be more easily implemented. Repeatedly introducing and withdrawing a treatment and quantitating patient symptoms will strengthen inferences about the intervention's impact.

How might such an approach benefit Mr C? Many, perhaps most, patients with CAL do not gain incremental benefit from adding a β-agonist to inhaled ipratropium,⁴ a consideration that may be particularly compelling in patients with heart failure who benefit from β-blockade. Thus, Dr M might consider an N-of-1 trial of withdrawal of the β-agonist inhaler. Indeed, N-of-1 RCTs have proved useful in patients with CAL,⁶¹ and an RCT of N-of-1 RCTs in patients with CAL has demonstrated reduction in the use of theophylline, from which patients were not benefiting.⁶²

If Mr C experienced no adverse effects of withdrawal of a β-agonist, Dr M could consider a trial of β-blockers. The severity of Mr C's CAL and his AV node block argue against a trial of β-blockers. Consider, however, a heart failure patient with milder CAL and no additional contraindications to β-blockers who shows no benefit from inhaled β-agonists. Many such patients tolerate cardioselective β-blockade,⁶³ and the physician should consider a trial of β-blockers.

Presumably, Mr C is taking amlodipine for his angina. Dr M has chosen the calcium channel blocker wisely; in contrast to other calcium antagonists,⁶⁴ it is unlikely that amlodipine has adverse effects in heart failure patients.⁶⁵ Nevertheless, the presentation of Mr C's case left me unconvinced that Mr C's angina is better controlled as a result of amlodipine administration, raising the possibility of an N-of-1 trial of amlodipine withdrawal. The N-of-1 approach could also be used if Dr M chose to conduct a trial of losartan.

With 2 serious chronic illnesses and other comorbidity, Mr C's management is a major challenge. His assets, however, include a loving and supportive family who are very willing to do what they can to help with Mr C's care. Possible modifications in management that Dr M might consider include carefully monitored withdrawal of some of his medications, a similar careful trial of losartan, and the addition of amiodarone.

AN INTERNIST: One problem we face, and it's beautifully illustrated by this patient, is that most of the evidence-based data that you've provided are based on studies of single agents. That's not the way they're given to these patients. The whole issue of drug interactions really complicates management in this setting. How do you sort that out? How do you address the next layer of complexity as you begin to look at evidence-based medicine as it applies to an individual patient?

DR GUYATT: It's not quite true to say that most of these trials have single agents. If we look at most of the β-blocker trials, patients are already taking ACE inhibitors and/or diuretics, and the issue is the addition of β-blockers. If you look at the trials of amiodarone, most of the heart failure patients were already on multiple medications. Nevertheless, your point is still a good one. Consider trials of mortality after myocardial infarction, in which aspirin, β-blockers, and ACE inhibitors were used in sequence. Once patients are already on aspirin, they get incremental benefit from β-blockers. Once they're already on aspirin and β-blockers, the additional effect of adding ACE inhibitors is similar to the effect on patients who are not already on aspirin and β-blockers. Still, we must take particular care about adverse effects. For instance, Mr C has COPD and AV node block. His AV node block could be made worse by digoxin. It could be further worsened by a β-blocker.

A PHYSICIAN: One problem with evidence-based medicine is that the evidence keeps changing. Type 1 agents were good; now they're bad. For example, calcium channel blockers were good; now they're bad. Would you comment on their use for Mr C?

DR GUYATT: Calcium antagonists are another great example of the limitations of observational studies and physiologic rationale. The trials were mounted with every expectation that calcium antagonists would be beneficial, yet there's a suggestion that they are harmful. The 1 agent that may be different in patients with heart failure is amlodipine. One randomized trial suggests that amlodipine may even be beneficial.⁶³ If one chooses a calcium antagonist for a patient with heart failure (I presume Mr C is on a calcium antagonist because it's desirable in terms of his angina), amlodipine is a good choice.

DR M, THE PRIMARY PHYSICIAN: Mr C also has severe 4-chamber dilatation. In the absence of documented persistent atrial fibrillation, should he be anticoagulated?

DR GUYATT: Investigators have not yet undertaken randomized trials in which patients with 4-chamber dilatation in sinus rhythm are randomized to receive anticoagulants or not receive anticoagulants. As a result, we are much less confident about management recommendations in such patients. One decision analysis noted that the biggest risk of bleeding is in the first year. Mr C doesn't fall and has gotten through a year. So, he's at a relatively low risk of bleeding. The decision analysis concluded that patients like him who don't find it too much of a hassle to take anticoagulants and have them monitored are probably likely to have a net benefit from taking anticoagulants. In the absence of RCTs, however, the evidence is relatively weak. A conservative approach would say we don't have randomized trial evidence in this situation and that we should be more hesitant to prescribe anticoagulants as a result.

A PHYSICIAN: Would you offer Mr C a dobutamine holiday or other inotropic support? What data exist for those treatments?

DR GUYATT: I know of no adequate data in this area. Here's where your N-of-1 method can come in. When we're talking about symptoms, we can focus on our patient, and if we do that in a rigorous way, with careful monitoring of symptoms, we might be able to make inferences about the symptomatic benefit of such interventions. We won't be able to say anything about mortality, but we might be able to say something about whether they make Mr C feel better. The results of trials of inotropic agents, which showed increased mortality, should dampen our enthusiasm for this experimental approach.^{64 - 68}

AN INTERNIST: What about testing medicines that take a long time to have an impact—or to wear off? How do you think about the placebo effect in N-of-1 trials?

DR GUYATT: Although N-of-1 trials are helpful, they're not the solution for all problems. You're quite right in implying that you would be much more reluctant to use N-of-1 methods in a medication that took a long time to work, or if you withdrew it and its effects took a long time to wear off. Second, you're right that part of the effect we have when we give medications is the placebo effect. That may be a very good thing—it actually makes people feel better. The question addressed in an N-of-1 trial, in which you use a placebo, is, "What is the biologic effect of the medication?" That is often a very important question. It certainly could also be argued, why subject our patients to the potential downside, harm, risks, and side effects associated with a therapy if there's no potential positive biologic effect beyond the placebo effect?