Data sources and searches

The search strategies (S1 Text) were developed without language restrictions, and the databases EMBASE, MEDLINE, and Web of Science were searched from inception to November 23, 2020. A search strategy for general drug terms, individual drug names (identified by the international nonproprietary name (INN) in the British National Formulary (BNF) and US National Library of Medicines) were used, and drugs with name changes were accounted for [17]. To facilitate the search strategy, a macro was developed to automate searches (Pulover’s macro creator version 5.2.8). Reference lists of eligible reports were reviewed, and authors were contacted to supplement incomplete papers. This study is reported as per the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guideline (S1 Checklist). The study protocol is publicly available on PROSPERO (CRD42020168697).

Study selection

RCTs comparing any drug with placebo, reporting OH as an adverse effect or outcome in adults (≥18 years), were included. Studies in pregnancy and anaesthetic or surgical contexts were excluded. Studies on the drug treatment of OH (such as midodrine, fludrocortisone, droxidopa, erythropoietin, pyridostigmine) and investigational or withdrawn drugs were excluded.

Data extraction and quality assessment

Using a standardised form, 1 reviewer (CB) screened all titles, abstracts, and full-text articles reporting potentially eligible studies. A second reviewer (DN) screened 10% of titles and abstracts and 3 reviewers (DN, HH, and SM) screened 10% of all full-text articles. Disagreements were resolved by discussion with an adjudicator (KW) when necessary. Online systematic review software (Rayyan, QCRI) was used to facilitate literature screening.

CB, HH, and SM independently extracted data from each article. Data were extracted on study characteristics, drug, dose, participant characteristics (age, sex, comorbidities), study setting, pharmaceutical sponsorship, and reporting of OH. Reporting of prevalent OH while taking the medication or placebo was grouped into 6 categories following expert consensus: “measured and validated” (a documented postural BP examination performed using a threshold of ≥20 mm Hg systolic/≥10 mm Hg diastolic reduction); “measured” (a documented postural BP examination performed without a specified threshold); “BP examined” (a documented BP examination only—but implied as postural, since the study reports OH); “physical examination”; “vital signs examination”; “symptom report” or “unclear.” Where possible, authors were contacted to obtain further information on reporting of OH. Reports of OH induced through orthostatic stress (e.g., tilt table testing) were excluded. We included RCTs comparing a drug to placebo, which reported OH according to the top 3 categories we regarded as acceptable (“measured and validated”; “measured”; “BP examined”).

Using the revised Cochrane risk-of-bias tool for randomised trials (RoB 2) [18,19], CB assessed all articles for risk of bias, and 3 reviewers (DN, HH, and SM) independently assessed 10% of articles. Studies were assigned an overall score: “low,” “some concerns,” or “high.”

The Cohen’s κ statistic addressed inter-rater agreement regarding eligibility. For duplicate studies or pooled analyses, we included the report with the most complete data. All RCTs comparing a drug versus placebo with acceptable reporting of OH were grouped according to drug class. Where at least 3 studies were available, these groups were then pooled.

Data synthesis and analysis

Summary odds ratios (ORs) were estimated for OH (whether a patient had OH or not while taking the medication or placebo) as a dichotomous outcome using fixed effects Mantel–Haenszel statistics. Zero total event trials were included in the meta-analysis since this can move the pooled estimate of treatment effect closer to nil, decrease its confidence interval, and decrease between-study heterogeneity [20]. Heterogeneity was assessed using the I² statistic with an I² >30% representing substantial heterogeneity [21].

Subgroup analyses were conducted if there were 3 or more studies available in a given subgroup: (1) trials reporting “measured and validated” OH outcomes only; (2) low risk of bias (excluding studies with an overall score of “high”); (3) drug dose (low versus high); (4) older patients ≥65 years; and (5) populations at greater risk of OH (this included people with cardiovascular conditions, diabetes, and older people at high risk of falls). Review Manager software version 5.4 was used (Cochrane).

The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) tool [22] was used to summarise the certainty of evidence. When there were at least 5 studies for meta-analysis, publication bias was assessed by visual assessment of funnel plot asymmetry.

Vasodilators

Calcium channel blockers (CCBs).

There were 5 eligible RCTs including 721 patients comparing a CCB to placebo. The mean age among patients was 62.6 years, and 44.0% were female. The largest trial included older patients (mean age 72.4 years) (S1 Table). CCB use was associated with 11% lower odds of OH compared to placebo; however, this was nonsignificant, and a more substantial decrease or increase in odds is also possible (OR 0.89 [95% CI 0.49, 1.65]) (Fig 2). Subgroup analyses on trials with “measured and validated” outcomes only, at low risk of bias, in older patients and patients at greater risk of OH due to cardiovascular conditions showed similar results (S2 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Meta-analysis results for vasodilators vs placebo.

ACE, angiotensin-converting enzyme; ARB, angiotension receptor blocker; CCB, calcium channel blocker; SGLT-2, sodium–glucose cotransporter-2; SSRI, selective serotonin reuptake inhibitor.

https://doi.org/10.1371/journal.pmed.1003821.g002

Sympathetic inhibitors

Alpha-adrenoreceptor blockers.

There were 17 studies with 18 eligible trials including 7,250 patients comparing an alpha-adrenoreceptor blocker to placebo. The mean age of patients was 55.9 years, and 92.3% were males. The majority were treated for benign prostatic hypertrophy (BPH) or lower urinary tract symptoms (LUTS) at the maximum recommended dose (S1 Table). Alpha-adrenoreceptor blockers were associated with higher odds of OH compared to placebo (OR 1.87 [95% CI 1.33, 2.64]) (Fig 3). Subgroup analysis in trials with “measured and validated” OH outcomes and in populations with urological conditions (excluding healthy volunteers) showed similar results (S2 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Meta-analysis results for sympathetic inhibitors vs placebo.

TCA, tricyclic antidepressant.

https://doi.org/10.1371/journal.pmed.1003821.g003

Main findings

We have conducted a systematic review and meta-analysis of 69 placebo-controlled RCTs to investigate the extent to which specific medications are associated with OH in adults. Compared with placebo, beta-blockers and TCAs were associated with a 6- to 7-fold increased odds of OH. Alpha-blockers, second-generation antipsychotics, centrally acting antihypertensives, and SGLT-2 inhibitors were associated with up to a 2-fold increased risk of OH, compared to placebo. There was no difference in odds of OH with CCBs, ACE inhibitors/ARBs, and SSRIs compared to placebo. These findings are based on varied certainty of evidence, which ranged from low to high. Our study has characterised a range of commonly prescribed drug classes according to OH risk to guide selective prescribing and monitoring of postural BP in practice.

Reports on antihypertensives and risk of OH in the literature are conflicted. There have been reports of OH with ACE inhibitors [12] and diuretics [14,9] that were not replicated in our study. It is possible that risk of OH was attenuated in our study due to the selective younger, fitter RCT population. Antihypertensives also induce OH through their therapeutic effect; such drugs often require much higher doses to observe adverse effects [23]. However, many studies suggest a protective postural effect associated with ACE inhibitors [9,24]. Juraschek and colleagues recently found that intensive BP-lowering treatment does not increase risk of OH [12]. It is known that hypertension itself increases risk of OH [13], so it is likely the complex interaction between baseline risk of OH in well-controlled hypertensive patients and antihypertensive drug effects complicates the true picture. Current consensus suggests that optimal control of BP (even among older adults) should be prioritised over potential risk of OH [12,25].

Beta-blockers induce OH through sympathetic inhibition decreasing heart rate and contractility, alongside combined independent vasodilatory effects [10,26]. Previous observational studies with strict measurement of postural BP have reported that beta-blockers are strongly associated with OH, independent of comorbidities, consistent with our results [13,27]. Current consensus states that beta-blockers should not be prescribed in preference for hypertension, due to both its potential to cause harm and lack of efficacy relative to other antihypertensives [27].

Among the drugs identified, alpha-blockers were associated with least risk. This is likely related to the majority of alpha-blockers in our study being uroselective (such as tamsulosin and alfuzosin), which have fewer cardiac effects [26]. Nevertheless, alpha-blockers almost doubled the odds of OH compared to placebo and are widely prescribed for prostatic hypertrophy to treat LUTS [28]. Dizziness and OH associated with alpha-blockers is a particular problem among older patients, and patients frequently discontinue alpha-blocker treatment in clinical practice [28]. The use of 5-alpha reductase inhibitors (5-ARIs) could reduce the need of alpha-blockers for LUTS [29]. Discontinuation of alpha-blocker treatment after 6 months in patients receiving combination therapy has been recommended and has been shown to have no significant effect on LUTS [28].

TCAs similarly exert their effects on postural BP through combined sympathetic inhibition and reduced vascular resistance [26]. Clinical guidelines support prescription of SSRIs in preference to TCAs for depression due to fewer adverse effects [10,30]. However, TCAs are still the second most prescribed antidepressant in older people (likely related to low-dose off-label use for pain and insomnia) [31]. While our study focused on TCAs at higher doses for depression and identified a 6-fold increase in odds of OH compared to placebo, it is likely that TCAs at lower doses also cause harm, especially in older adults at higher risk of OH due to a decrease in baroreflex sensitivity [32].

Interestingly, SGLT-2 inhibitors were the only drugs among vasodilators associated with significantly higher odds of OH. Their cardiac effects are exerted through diuresis and independent cardio-inhibitory effects (that are less well understood), alongside vasodilation [33]. SGLT-2 inhibitors are now considered preferential as second-line treatment for T2DM [34]. T2DM itself can increase the risk of OH, as a manifestation of autonomic neuropathy. However, improved glucose control and other positive effects on body weight related to SGLT-2 therapy can reduce the risk of neuropathic complications [34]. Therefore, the relationship between SGLT-2 use and OH in patients with T2DM is complex, similar to antihypertensives.

The drugs associated with highest odds of OH in our study (alpha-blockers, alpha-agonists, antipsychotics, beta-blockers, and TCAs) all share a common key mechanism of sympathetic inhibition causing cardioinhibitory effects [10,26]. Among these, beta-blockers and TCAs demonstrate the strongest association, with 6 to 7 times increased odds of OH compared to placebo. Both these groups induce OH through combined mechanisms of sympathetic inhibition and vasodilatory effects suggesting that OH risk rises with cumulative drug targets. This suggests that coprescription of drugs with the potential to cause OH may also result in cumulative harm, in keeping with a growing body of evidence [3,35].

Study strengths and limitations

To our knowledge, this is the first systematic review providing an overview of which drugs are associated with OH. Strengths of this review included (1) a comprehensive literature search, enabling an overview of all prescription drugs associated with OH, the first of its kind; (2) analysis by drug mechanism relevant to clinical practice; (3) predominantly moderate to high-quality RCT evidence; and (4) detailed characterisation of RCTs and subgroup analyses that addressed differences in reporting of OH, risk of bias, dose, and patient characteristics.

This review has several limitations. The primary aim of this study was to provide an overview of all drug groups and their relative association with OH. Due to the extensive nature of this question, we limited studies to drug versus placebo RCTs reporting acceptable OH outcomes. We therefore excluded head-to-head drug comparisons, which are likely to be newer RCTs comparing drugs to existing treatment. We have reported the prevalence of adverse OH during medication or placebo use in trials; however, due to limited data in the included studies, we cannot be sure if all participants were established to have a normal postural BP prior to treatment. While we included a large number of RCTs overall in this study, for some drug groups, the meta-analysis results have been driven by only a few trials that carried greatest weight. We also grouped drugs by classes and acknowledge that not all individual drugs within a class will be similarly associated with OH. Future research should examine head-to-head drug comparisons and differences between individual drugs in those at high risk of OH.