|Home | About | Journals | Submit | Contact Us | Français|
Patients trying life-preserving agents such as beta-blockers may be discouraged by listings of harmful effects provided in good faith by doctors, drug information sheets, and media. We systematically review the world experience of side-effect information in blinded, placebo-controlled beta-blockade in heart failure. We present information for a physician advising a patient experiencing an unwanted symptom and suspecting the drug.
We searched Medline for double-blinded randomized trials of beta-blocker versus placebo in heart failure reporting side-effects. We calculated, per 100 patients reporting the symptom on beta-blockade, how many would have experienced it on placebo: the “proportion of symptoms non-pharmacological”.
28 of the 33 classically-described side-effects are not significantly more common on beta-blockers than placebo.
Of the 100 patients developing dizziness on beta-blockers, 81 (95% CI 73–89) would have developed it on placebo. For diarrhoea this proportion is 82/100 (70–95), and hyperglycaemia 83/100 (68–98). For only two side-effects is this under half (i.e. predominantly due to beta-blocker): bradycardia (33/100, CI 21–44) and intermittent claudication (41/100, 2–81).
At least 6 so-called side-effects are less common on beta-blocker than placebo, including depression (reduced by 35%, p < 0.01) and insomnia (by 27%, p = 0.01).
Clinicians might reconsider whether it is scientifically and ethically correct to warn a patient that a drug might cause them a certain side-effect, when randomized controlled trials show no significant increase, or indeed a significant reduction. A better informed consultation could, in patients taking beta-blockers, alleviate suffering. In patients who might otherwise not take the drug, it might prevent deaths.
Beta-blockers greatly improve survival in chronic heart failure [1,2] but only in patients who take them. Uptake is poor  and not improving . One contributor is perception amongst patients and physicians of side-effects.
Patient experience of side-effects may be an unreliable guide to the true pharmacological induction of side-effects by the drug because, in real life, drugs are not administered in an information vacuum. Instead, well-intentioned forewarning of potential side-effects by clinicians conditions patients to believe any new symptom is a drug side-effect, even if it is really due to heart failure itself or a coincidental problem. Side-effects described in drug information leaflets and official online information  are listed in Table 1.
Information not available to patients, and not realistically available to front line doctors, is careful delineation of which side-effects have reliable evidence of induction by beta-blockers, in patients with heart failure. Even more helpful would be to be ability to discuss quantitatively, with patients reporting a listed side-effect, the probability that they are experiencing a genuine consequence of the drug rather than spontaneous phenomena or nocebo (the undesirable flip-side of placebo).
In this study we examine the only reliable source of this information: double-blinded randomized controlled trials (RCTs) in which doctors and patients reported side-effects without knowing whether the patient was taking the drug or placebo. The increment in side-effects genuinely caused by the drug can therefore be seen.
This systematic review is the first to present reliable values, for each listed side-effect, of the proportion of heart failure patients who report it in which the beta-blocker is truly the cause, using comprehensive data from RCTs of placebo versus beta-blocker therapy in systolic heart failure. It extends upon previous analyses  by covering a more extensive set of studies, by intentionally using an analysis that is easily replicated by a physician later wanting to add data of trials not currently available, and by deliberately presenting information in a format as helpful as possible to clinicians and patients handling the situation of a suspected side-effect.
We carried out a systematic review of studies that evaluated side-effects noted in RCTs in patients with systolic heart failure in accordance with published guidelines . Between November 2010 and 8 November 2011 we searched Medline (1950 to present) using the keywords beta-blocker, systolic heart failure, randomized controlled trial and RCT, accessible in English. We excluded letters, abstracts, systematic reviews and meta-analyses. Two independent reviewers (NZ, AB) screened the titles, assessing for eligibility by abstract review to determine satisfaction of the selection criteria, before retrieving selected titles. Any disagreements were resolved through consensus or consultation with a third reviewer (DF). Reference lists of the retrieved articles were searched for additional publications. Follow-up length was not specified by our review protocol. Each article was fully reviewed for data including total number of patients in each arm, and number of patients exhibiting each side-effect in the beta-blocker and the placebo arm separately. The authors of this manuscript have certified that they comply with the Principles of Ethical Publishing in the International Journal of Cardiology.
We identified parallel-group RCTs comparing a single beta-blocker versus placebo. Trials were excluded if they were not randomized double-blinded studies; had cross-over design; did not report side-effect data separately for beta-blocker and placebo arms; or if other (heart failure) medications were selectively introduced as part of study protocol.
We listed side-effects, 33 in total, that were assessed for presence in more than 100 patients across all eligible studies. We have used the phrase “listed side-effect” to indicate a symptom listed by at least one RCT as a side-effect, regardless of whether our final results indicate that the drug increased, did not affect, or even decreased the rate. Because some studies reported side-effects that others did not, the denominator number of patients differed between side-effects. Funnel plots were constructed for 10 of the listed side-effects which in the majority show an acceptable distribution. These results are shown in Supplementary Fig. 1.
Tests for heterogeneity were employed, and because of heterogeneity between trials in some side-effects, a random-effects, rather than fixed-effects, model was used to evaluate the difference in event rate between arms in the trials. Comprehensive Meta Analysis software package, Version 2.2.057, was used. The z-score and confidence interval of difference were calculated by standard methods.
In parallel a simple analysis was employed. This secondary analysis, presented as a supplemental file, is designed to be conceptually simple and arithmetically straightforward so that it could be easily be followed and replicated by anyone wishing to recheck the findings or update it with future primary data, or explain it to a patient. It is based on simple addition of numbers of patients enrolled in corresponding arms, and numbers with side-effects. The excess event rate in the drug arms was calculated by straightforward subtraction: rate in beta blocker arms minus rate in placebo arms. A negative value indicated that the symptom was less frequent in the drug arms than the placebo arms.
We calculated per 100 beta-blocker recipients describing that symptom, how many would have had that symptom on placebo. This “proportion of symptoms non-pharmacological” was calculated for each listed side-effect. Fig. 1 shows an example of how this is calculated, and offers one simple way of displaying the results. The spreadsheet used to produce Fig. 1 is freely downloadable in the Online-only supplement. The result is shown in the final column of Table 3.
Total adverse events and withdrawals due to supposed adverse side-effects or (if those specific data were not available) total withdrawals were also calculated.
Beta-blockers included carvedilol and bucindolol (non-selective) and metoprolol, bisoprolol and nebivolol (selective). We performed sensitivity analyses to assess for inherent differences in side-effect profiles between these two subclasses.
Run-in periods (identified in Table 2) were excluded because the data including the whole cohort prior to a run-in period was almost never disclosed. Most studies present the results of an up-titration phase and a subsequent maintenance phase together but in some, where side-effects were described separately, we have included only those seen in the maintenance phase, to avoid the risk of double-counting.
Chi-squared analysis with threshold of p < 0.05 was used to determine whether there was a significant difference in rate between drug and placebo arms.
Of the 319 studies screened, thirteen RCTs were included in the final analysis [1,2,8–18] (Fig. 2). Table 2 shows a summary of all these trials. These include up to 7836 patients in the beta-blocker arm, and 7547 patients in the placebo arm for the 33 listed side-effects. Only two studies describe that side-effects were actively discussed at each follow-up visit [10,16] otherwise it is unclear whether the occurrence of side-effects was solicited or gathered through spontaneous self-reporting. The studies that listed depression as a side-effect did not report how this was validated.
For 3 side-effects (worsening heart failure, oedema and bradycardia) there was significant heterogeneity between studies so, for all side-effects, a random-effects model was employed. The p values listed are from this principal analysis. A corresponding table showing the results from the secondary analysis is shown in the Online Supplementary Table 1.
There was no significant difference in the frequency of side-effects between arms for most (21 of 33) listed side-effects. Examples of listed side-effects in this group included impotence, weight gain, postural hypotension and headache; they are shaded yellow in Table 3.
Depression (p < 0.01) and insomnia (p = 0.01), along with worsening heart failure (p < 0.01), palpitations (p < 0.01), chest pain (p < 0.01) and tachycardia (p < 0.01), were significantly less frequent in the beta blocker arm in the principal analysis and are shaded green in Table 3. Broadly similar results were seen in the secondary analysis (Online Supplement Table). Nausea was less common in the beta-blocker arm, in the secondary analysis, but the formal meta-analysis showed no significant difference between arms; this was the only listed side-effect with a discrepancy in category (colour code) between the two analyses.
Only 5 of 33 listed side-effects are significantly more common in the beta-blocker arm, shown with red p-values in Table 3. These are considered further in Fig. 3, which shows that, even for these 5 side-effects, in many patients it is not the drug that is the cause.
Out of 100 patients on beta-blockers developing dizziness, a commonly discussed side-effect of beta-blockade, 81 (95% confidence intervals 73–89 patients, p < 0.01), would have had that symptom on placebo; either arising naturally or from the warnings given (i.e. beta-blockers are only causative in 19 from every 100 patients). For diarrhoea this proportion was 82 from 100 patients (95% CI 70–95, p < 0.01). For hyperglycaemia this proportion was 83 (CI 68–98, p < 0.01), for intermittent claudication 41 (CI 2–81, p < 0.01), and for bradycardia 33 (CI 21–44, p < 0.01).
Forest Plots for 5 frequently discussed side-effects can be viewed in Fig. 4.
Drug withdrawal data were formally presented in 6 out of 13 trials. Randomization to beta-blocker did not significantly increase the risk of withdrawal, with a 20% relative reduction compared with placebo (Fig. 5A).
Serious adverse events were formally presented in 2 out of 13 trials and occurred in 22.1% of patients receiving beta-blocker and 25.6% of patients receiving placebo: randomization to beta blocker therefore decreased the risk of serious adverse events by 16% (95% CI 4–27% p = 0.01, Fig. 5B).
For only 10 of the 33 listed side-effects were sufficient data available from trials of both selective and non-selective beta-blockers (Supplementary Table 2). Results were similar between types of beta-blockers for dizziness, bradycardia, chest pain, dyspnoea, fatigue, headache, asthenia, and hypotension. Although both classes of beta blocker resulted in significantly fewer cardiac failure events in the active drug arm, the magnitude of this difference was significantly greater in the non-selective trials. The only listed side-effect with any suggestion of difference between the two types of beta-blockers was oedema with a significant increase in the drug arm of the selective beta-blocker trials (p < 0.01), and a borderline significant decrease in the non-selective trials (p = 0.06), leading to overall no significant difference between placebo and beta-blocker amongst all trials.
Medicine has become a science through recognising the existence of spontaneous variability in disease over time, and the resulting need to design and conduct an RCT to reliably establish whether a treatment works. This recognition should also be applied to information given to patients about side-effects.
Published RCTs indirectly give valuable information that is currently rarely discussed. For any listed side-effect it is easy for a physician to calculate, from one RCT or a set of RCTs, the proportion of those who suffered a symptom who would have suffered it even without the medication (the PSN). The downloadable spreadsheet in the Online Supplement assists with this, and provides a visual display that may support discussions with patients who find the PSN difficult to understand in isolation.
The majority of adverse effects reported with beta-blockers in heart failure are not caused by the beta-blockers per se, but arise either from the disease itself, another coincident problem, or from the power of suggestion — the nocebo phenomenon .
Although beta-blockers reduce mortality by about 35% in heart failure [1,2,14], utilization is disappointing, with sustained usage being only ~ 50% . Even when beta-blockers are started, willingness to achieve target doses is disappointingly low , presumably because of fears of provoking side-effects by uptitration, despite randomized double-blind experience contradicting this . Asymmetric discussion of the benefits and risks of therapy can be particularly damaging during uptitration. Each clinic visit risks a fresh enquiry about side-effects, implying to the patient that future symptoms might be assumed to be the result of the drug. Albeit innocent, this sows the seeds for future abandonment.
In this analysis there were only 2 of 33 listed side-effects (bradycardia and intermittent claudication) for which the majority of sufferers had that side-effect genuinely caused by the drug.
In another 3 of the 33 side-effects, the drug was genuinely responsible in less than 1 in 4 sufferers. For example, in patients with dizziness, a symptom commonly attributed to beta-blockade, PSN was 81/100, i.e. beta-blockers were found to be genuinely responsible in only 19 out of 100 cases.
In the other 28 of 33 side-effects, there is no evidence that the drug makes them more common. These include fatigue, commonly attributed to beta-blockade. Fatigue is very prevalent in the general healthy population, being reported in 39% of un-medicated volunteers : it is unlikely to be less prevalent in heart failure.
Unscientific claims of therapeutic efficacy, that do not account for the placebo effect, are rightly condemned and punished. Yet the same scrutiny is not yet applied to incorrect assertions of harmful effects. It would never be permitted to claim that a drug decreased an undesirable outcome when in fact the data showed that it increased it: it should correspondingly not be permitted to warn that a drug increases a side-effect when the data show that it decreases it.
The origin of the extensive list of side-effects may have been un-blinded, early-phase experience where investigators were asked to document any adverse events occurring alongside medication use, with the focus on comprehensiveness rather than meaningfulness. Such a list can only be interpreted alongside a parallel list of side-effects of similar patients taking placebo having been given the same warning information, but there is no organized process for formulating this.
Incautious reports of side-effects can end up being re-reported and advertised in lay media uncritically, discouraging initiation and encouraging discontinuation, ultimately causing deaths. This unjustifiable asymmetry is most reliably resolved using double-blinded data from RCTs.
Well-intentioned individuals designing best-practice recommendations, speculating in isolation from day-to-day experience, may think it superficially reasonable to insist on providing patients with all information on potential consequences of proposed therapy . However if there is no standard on what is considered a potential consequence, then any list must necessarily be infinitely long since no proposed addition can ever be refused.
Individual patients may have preferences on how they wish to be managed on a spectrum between paternalistic medicine (told little) and a fully informed approach (told everything) , but few want information on side-effects that is invalid or – even worse – backwards.
Physicians personally involved in proposing highly effective therapies to psychologically-vulnerable patients with lethal diseases might be forgiven for not recommending that they carefully scrutinise side-effects lists which could be argued to contain more untruth than truth. Instead, physicians might focus the patient's attention on a much smaller core of reliable information, together with information on the proportion of side-effects that are non-pharmacological. Thus adverse effects, when later experienced, are not automatically assumed to be caused by the drug. Fig. 6 shows a starting proposal for a replacement side-effect list for heart failure patients starting beta-blocker therapy.
Physicians may feel that in individual cases it can be clear that a side-effect is due to the medication, and they may be right. On the other hand it was notable in the COPERNICUS study that whilst 38% of patients on beta-blocker required down-titration due to side-effects, 33% of patients on placebo also “required” this down-titration. As physicians we may be no less susceptible to the illusion of control than other disciplines . PSN values, such as those shown here, are a starting point for a rational open discussion with a patient.
Some physicians also consider the legal dimension, and sometimes feel forced to make decisions to avoid litigation . If, on the one hand, the law compels the clinician to advise a patient that a drug could cause him or her (for example) depression, whilst at the same time the randomized blinded data show a significant reduction in depression, the clinician is in a quandary. We may need clearer language so that the scientific, clinical, and legal fields can share the same conversation. One possible form of words to resolve this might be that taking this medication “reduces the chances” of depression. This highlights that depression is still possible but truthfully reports that its likelihood is reduced. The patient would correctly understand that whilst the drug may still cause depression, if it does so it prevents it in more patients.
This study only looked into trials of beta-blockers in heart failure rather than hypertension, angina or arrhythmia because it is in heart failure where the absolute survival benefits of beta-blockers are most pronounced, and inadvertent discouragement of therapies is most prone to increase mortality. Therefore these data are only applicable to heart failure. For other disease states, separate analysis of the appropriate RCTs is needed. Although a single analysis for all diseases of all RCTs of beta-blocker versus placebo would have more patients, it would be difficult to apply the findings confidently to heart failure because, just as therapeutic efficacy of drugs may differ between disease states, side-effects may also differ.
It is possible that there are some side-effects that are caused in some patients and prevented in others, leaving no net effect in rate, since RCTs are only able to identify net effect reliably. There is no reliable method to identify in advance (a) which patients will be symptom-free without therapy but get symptoms from therapy, and (b) which patients will have symptoms without therapy but be symptom-free with therapy. Without such a method, the net effect is the main clinically useful value for side-effects — just as it is for benefits.
Provision of information and protocol for eliciting the presence of new symptoms was not standardized across the trials. Both are known to affect the frequency of side-effects . Ideally an analysis would correct for this, frequency of visits and information given prior to enrolment regarding potential side-effects. Unfortunately the trial reports did not consistently indicate how side-effect information was collected.
The underlying RCTs recruited willing volunteers. In general these are highly motivated, and perhaps less likely to experience or report spontaneous events as potentially drug-related. Therefore these figures may not be representative of clinical reality where patients enter the heart failure population unwillingly. However, this means that real-life clinical populations would have higher PSN values.
Trial design differed between studies, most notably with regards to a run-in period in which all receive placebo and some (6–12% in these studies) drop out for various reasons including perceived side-effects. Again this means that real-life PSN may be even higher.
There is potential for bias towards greater side-effect identification in the beta-blocker arm, because they lived longer and so had more opportunity to report symptoms. Yet again, this means that real-life PSN might be even higher.
Almost all commonly listed side-effects attributed to life-saving beta-blocker therapy in heart failure are no commoner in patients on beta-blockers than their counterparts receiving placebo. In only 2 of 33 listed side-effects are the majority of patients experiencing a genuine adverse effect of the drug. For a further 3, although commoner in patients taking beta-blockers, the majority of patients experience either the natural history of their disease, or the psychological effect of being given incorrect or incomplete drug information.
For 28 of the 33 listed side-effects, the randomized controlled trials in heart failure do not scientifically support their listing for heart failure patients. For at least 6, the warnings may actually be backwards, because the drug is demonstrably preventative rather than causative.
The current convention of providing misinformation may be impeding uptake and uptitration of these agents which would otherwise prevent deaths.
The following are the supplementary data related to this article.
Simple analysis, replicable by any reader, to calculate the proportion of side-effects on beta-blocker that are caused by being on beta-blocker (green denotes significantly more common in the placebo arm, amber denotes no significant difference between arms, and red significantly more common in the active drug arm).
Differences in side-effect profiles of certain listed side-effects between selective and non-selective beta-blockers.
No other persons have made substantial contributions to this manuscript.
Funding sources: AJB and DPF acknowledge support from the British Heart Foundation (FS/10/038).
This is an open-access article distributed under the terms of the Creative Commons Attribution-NonCommercial-ShareAlike License, which permits non-commercial use, distribution, and reproduction in any medium, provided the original author and source are credited.