Types of studies

We will include RCTs. RCTs are the gold standard for research to inform medication efficacy and for systematic reviews as they use the most systematic and rigorous design possible.

Types of participants

Older adults prescribed one or more antihypertensive drug for hypertension or primary prevention of cardiovascular disease living in the community, residential aged care facilities or in hospital settings. Older adults are defined as adults aged 50 years and over. We will include a study with some adults under 50 years if the majority of participants are over the age of 50 years or if data on adults aged over 50 years are reported separately (Leipzig 1999).

Types of interventions

Withdrawal of antihypertensive medications in older adults prescribed for hypertension or primary prevention of cardiovascular disease. The intervention can be directed to the person, healthcare professional, organisation or higher level, as long as individual participant use of antihypertensive drugs is captured. We will only include studies where an antihypertensive withdrawal intervention strategy is present. Withdrawal of medications may be through a deprescribing strategy including complete withdrawal or dose reduction.

The control intervention will be no withdrawal of antihypertensive medications.

Antihypertensive agents will include:

• diuretics which act primarily by blocking reabsorption of sodium at four major sites in the nephron. Different classes of diuretics act at different sites. Loop diuretics (e.g. furosemide, torsemide) act in the thick ascending limb of the loop of Henle. Thiazide‐type diuretics (e.g. hydrochlorothiazide, chlorthalidone, indapamide) act in the distal tubule and connecting segment. Potassium‐sparing diuretics (e.g. amiloride, triamterene) increase diuresis, but without causing potassium to be lost from the body. Aldosterone receptor antagonists (e.g. spironolactone, eplerenone) stop the entry of aldosterone into the principle cells of the collecting duct and late distal tubule of the nephron, which prevents sodium and water retention;

• beta‐blockers (e.g. atenolol, carvedilol) block the effects of catecholamines at receptor sites in the heart, peripheral vasculature, bronchi, pancreas, uterus, kidney, brain and liver. Beta‐blockers reduce BP by blocking the effects of catecholamine on the heart and blood vessels;

• ACE inhibitors (e.g. captopril, enalapril) act by blocking the renin‐angiotensin system, specifically they block conversion of angiotensin I to angiotensin II and bradykinin. ACE inhibitors reduce the effects of angiotensin II‐induced vasoconstriction, sodium retention and aldosterone release.

• calcium channel blockers (e.g. amlodipine, felodipine) act by blocking inward current of calcium via L‐type calcium channels. Calcium channel blockers lower BP by blocking the effects of calcium on blood vessels;

• angiotensin II receptor antagonists (e.g. candesartan, irbesartan) act by blocking binding of angiotensin II to type 1 angiotensin (AT₁) receptors. This leads to reduction in angiotensin II‐induced vasoconstriction, sodium reabsorption and aldosterone release, blood vessels dilate leading to reduction in BP.

• renin inhibitors (e.g. aliskiren) prevent the conversion of angiotensinogen to angiotensin I by binding to the S3^bp binding site of renin.

Types of outcome measures

Electronic searches

The Cochrane Hypertension Information Specialist will search the following databases from date of inception for published, unpublished and ongoing studies:

• the Cochrane Hypertension Specialised Register via the Cochrane Register of Studies (CRS Web);

• the Cochrane Central Register of Controlled Trials (CENTRAL) via the Cochrane Register of Studies (CRS Web);

• MEDLINE Ovid (from 1946), MEDLINE Ovid Epub Ahead of Print, and MEDLINE Ovid In‐Process & Other Non‐Indexed Citations;

• Embase Ovid (from 1974);

• ClinicalTrials.gov (www.clinicaltrials.gov);

• WHO International Clinical Trials Registry Platform (www.who.it.trialsearch).

We will model the subject strategies for databases on the search strategy designed for MEDLINE (see Appendix 1). Where appropriate, we will combine these with subject strategy adaptations of the sensitivity and precision‐maximising search strategy designed by Cochrane for identifying RCTs (as described in Box 6.4.d of the Cochrane Handbook for Systematic Reviews of Interventions; Higgins 2011).

Searching other resources

The Cochrane Hypertension Information Specialist will search the Hypertension Specialised Register segment (which includes searches of MEDLINE and Embase) for systematic reviews and Epistemonikos to retrieve published systematic reviews related to this review title, so that we can scan their reference lists to identify additional relevant trials.

We will check the bibliographies of included studies and any relevant systematic reviews identified for further references to relevant trials.

We will contact experts/organisations in the field to obtain additional information on relevant trials.

We may contact trial authors for clarification and further data if trial reports are unclear.

The Cochrane Hypertension Information Specialist will search the Hypertension Specialised Register segment for information of adverse effects relevant to this review.

Selection of studies

Two review authors (TG and ZB) will independently search and screen identified studies for relevance and adherence to inclusion criteria. If studies do not meet the inclusion criteria, we will exclude them and record reasons for exclusion in a 'Characteristics of excluded studies' table. We will resolve disagreements through consultation with a third review author. We will present the selection process using a Prisma diagram (Moher 2009). We will utilise GRADE to determine the strength of studies and thus additional reasoning for inclusion or exclusion of potential studies.

Data extraction and management

Two review authors (TG and ZB) will independently extract data. We will utilise a data extraction form and that will be piloted on two studies. We will resolve disagreements by consultation with an external expert (Ms Joanna Stewart, biostatistician).

The summary statistics required for each trial and outcome for continuous data will be the mean change from baseline (or control group), the standard deviation and the number of participants for each group (treatment and control). The baseline assessment will be defined as the latest available assessment between the treatment and control group, or from baseline.
For binary outcomes (i.e. success rate), we will require the number/percentage in (each) treatment group compared to baseline or control groups at each time point.

We will extract the following data from the studies and present them in a table:

• author, year of publication, type of intervention, design of study, number of participants, country;

• inclusion, exclusion and matching criteria;

• sex of participant;

• age of participant by age groups (50 to 65 years, over 65 to 80 years, over 80 years);

• withdrawal method (immediate or tapered);

• mortality status (from baseline);

• treatment effect on outcomes;

• fatal and non‐fatal cardiovascular events (i.e. myocardial infarction, stroke);

• adverse drug reactions;

• BP;

• QoL score (at baseline, from baseline);

For each outcome measure, we will seek data on each assessed participant using an intention‐to‐treat method. We will include per‐protocol analyses if available.

Assessment of risk of bias in included studies

Two review authors (TG, ZB) will independently assess the risk of bias in the included studies utilising the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011; Sterne 2014). The risk of bias will be assessed in terms of internal validity criteria for RCTs including random sequence generation (randomisation), allocation concealment, participant and study personnel blinding, blinding of outcome assessors, level of incomplete outcome data, selective reporting and other risks of bias that may be relevant (Higgins 2011; Sterne 2014). We will resolve disagreements through consultation with two additional review authors (ER, DG).

Measures of treatment effect

Studies may or may not utilise similar rating scales in outcome assessment. For this reason, for continuous outcomes, we will use the mean difference when the collective studies utilise identical scales of rating or tests. However, when dissimilar scales of rates or tests are utilised, we will use the standardised mean difference. In the case of binary outcomes such as mortality, we will use a Peto odds ratios to measure the treatment effect. We will calculate 95% confidence intervals (Cl).

Unit of analysis issues

If there are any cluster RCTs, we will determine if the risk of unit of analysis error was dealt with appropriately. Where the analysis was carried out correctly taking into account the clustering design, we will consider the studies for meta‐analysis and use the reported effect sizes and standard errors. Where the analysis was incorrect (i.e. not taking the clustering design into account), we will apply an interclass correlation (Higgins 2011).

Dealing with missing data

We will contact the corresponding author of included studies in the event of missing data that will compromise the ability of the review authors to examine the data and eligibility for study exclusion/inclusion in the final analysis.

Assessment of heterogeneity

The review authors will analyse and present each study separately. We will only perform meta‐analysis where studies are satisfactorily homogenous in terms of interventions, outcomes and participants. In the evaluation of heterogeneity, we will determine clinical heterogeneity by review author opinion and use an I² test to determine statistical heterogeneity. An I² value higher than 50% is considered as evidence for the presence of heterogeneity of the studies. If heterogeneity is present, we will not add these studies to any meta‐analyses undertaken and present them separately, with an investigation into the reasons for the presence of heterogeneity. Due to the nature of the intervention of interest, it is anticipated that a level of heterogeneity between studies will be present. Therefore, we will utilise a random‐effects model to incorporate and address this issue.

Assessment of reporting biases

Two review authors (TG, ZB) will independently assess the risk of reporting bias in the included studies utilising the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). We will resolve disagreements through consultation with two additional review authors (ER, DG).

Data synthesis

We will synthesise data for each study separately using Review Manager 5 (RevMan 2014), and undertake meta‐analyses if we find data from multiple studies.

We will compare outcome measures for binary data utilising risk ratios and 95% CI. If there is heterogeneity, we will use a random‐effects model. For continuous variables or data, we will utilise mean differences and 95% CI.

Subgroup analysis and investigation of heterogeneity

Where there are sufficient data from two or more studies, we will perform subgroup analysis of medications within antihypertensive drug classes (e.g. ACE inhibitors and diuretics) if sufficient and meaningful data are available. Where studies are conducted for a period of 12 months or less, we will analyse these studies separately from longer‐term studies that are conducted for more than 12 months of follow‐up. This is due to the possible large disparity in outcomes due to the time period. Additionally, we will include subgroups for age group (50 to 65 years, over 65 to 80 years, over 80 years) and gender (men and women).

Sensitivity analysis

We will undertake a sensitivity analysis and consider how the results of any meta‐analyses undertaken change under different assumptions, related to the reasons for these effects. More specifically, we will conduct a sensitivity analysis on the choice of utilising a random‐effects model. Since many issues pertaining to sensitivity analysis are only identified in the review process as a result of the characteristics of the studies, we will undertake suitable sensitivity analyses in the relevant context (Higgins 2011).