Types of studies

For inclusion, studies must conform to either of the following designs:

• Randomised controlled trial comparing the blood pressure (BP)‐lowering effect of a standard dose* of a beta‐blocker of interest (see Types of interventions), against a parallel control group (placebo or no intervention)

• Baseline controlled trial measuring BP before and after treatment with a beta‐blocker of interest, in a single group of subjects

We included baseline controlled trials because there is negligible, or no placebo effect, with 24‐hour ambulatory BP measurement (Gould 1981; Dupont 1987).

The following criteria must also be met for a study to be included:

• Minimum of three weeks of follow‐up

• Baseline measurements must be taken after an appropriate washout period

• Cross‐over trials reporting combined data from before and after cross over can be included as baseline controlled trial, if BP is also measured prior to commencing treatment. Cross‐over trials can be included as randomised controlled trials (placebo or no intervention) only if data before the cross over are available.

• Study participants must be ambulatory during BP measurementǂ

Blinding of intervention to investigators, participants, or both is not required

*Standard doses are those recommended for current clinical practice.

ǂWe excluded studies where BP was measured by an invasive technique, which requires people to be on strict bed rest. As the effects of beta‐blockers are particularly pronounced during exercise, its efficacy in blood pressure reduction in an ambulatory person would be more informative of its clinical usage.

Types of participants

Individuals over 18 years of age, with essential hypertension, with a baseline systolic blood pressure (SBP) of at least 140 mmHg, or diastolic blood pressure (DBP) of at least 90 mmHg, or both, were included.

Types of interventions

We included monotherapy with any beta‐blockers with partial agonist activity (BBPAA), including acebutolol, celiprolol, oxprenolol, pindolol, alprenolol, and bopindolol. Medication should be given at a standard dose, and may be administered at any frequency (for example, once daily or twice daily).

Types of outcome measures

Electronic searches

The Cochrane Hypertension Information Specialist searched the following databases, without language, publication year, or publication status restrictions:

• the Cochrane Hypertension Specialised Register via the Cochrane Register of Studies (CRS‐Web; searched 10 June 2020);

• the Cochrane Central Register of Controlled Trials (CENTRAL; 2020, issue 5) via the Cochrane Register of Studies (CRS‐Web; searched 10 June 2020);

• MEDLINE Ovid, MEDLINE Ovid Epub Ahead of Print, and MEDLINE Ovid In‐Process & Other Non‐Indexed Citations (searched 10 June 2020);

• Embase Ovid (searched 10 June 2020);

• ClinicalTrials.gov (www.clinicaltrials.gov; searched 11 June 2020);

• World Health Organization International Clinical Trials Registry Platform (www.who.it.trialsearch; searched 11 June 2020).

The Information Specialist modelled subject strategies for databases on the search strategy designed for MEDLINE. Where appropriate, they were combined with subject strategy adaptations of the highly sensitive search strategy designed by Cochrane to identify randomised controlled, as described in the Cochrane Handbook for Systematic Reviews of Interventions Version 6 (Lefebvre 2019). We present search strategies for major databases in Appendix 1.

Searching other resources

• The Cochrane Hypertension Information Specialist searched the Hypertension Specialised Register segment of CRS‐Web (which includes searches of MEDLINE, Embase, and Epistemonikos for systematic reviews) to retrieve existing reviews relevant to this systematic review, so that we could scan their reference lists for additional trials. The Specialised Register also includes searches of CAB Abstracts & Global Health, CINAHL (Cumulative Index to Nursing and Allied Health Literature), ProQuest Dissertations & Theses, and Web of Science for controlled trials.

• We checked the bibliographies of included studies, and any relevant systematic reviews identified for further references to relevant trials.

• Where necessary, we contacted authors of key papers and abstracts to request additional information about their trials.

Selection of studies

Two review authors (XYZ and SS) independently screened studies identified through database searches for relevance, based on titles and abstracts. Studies that failed to meet the inclusion criteria, or that fulfilled the exclusion criteria, were rejected. The same two review authors independently reviewed and assessed full texts of selected studies from the initial screening, to establish whether they were suitable to include in this review. A third review author (VM) resolved any discrepancies.

Data extraction and management

Two reviewers (XYZ and SS) independently extracted and cross‐checked all data from included studies. Hourly BP and HR data were often presented in graphical format. We extracted the data either manually, or by using a suitable data extraction software. After resolving any significant discrepancies (defined as greater than 1 mmHg or 1 beat/min), we calculated an average of the two values obtained by the review authors, and used this in the meta‐analysis. A third review author (VM) cross‐checked data extraction and imputation of standard deviation from all included studies.

In studies where the same cohort of people was treated with either different doses of the same drug, or with the same drug but for different lengths of time, we used data from the highest dose and longest treatment period in the meta‐analysis. The logic behind this is that a higher dosage and longer treatment are likely to produce the greatest change in BP and HR.

Where ambulatory BP and HR were measured and reported at less than a one‐hourly interval (for example, every 30 or 15 minutes), we calculated the hourly average, and used this in the meta‐analysis.

Beta‐blockers are commonly prescribed to be taken once or twice daily, with the first or only dose taken in the morning. In analysing ambulatory BP data, the time at which the morning dose was taken was considered 'hour 0'. Where the time the drug was taken was not specified in the study, 9 a.m. was assigned 'hour 0'. When taken as twice daily dosing, the time at which the afternoon dose was taken was assigned 'hour x', where x was the number of hours since the morning dose, not 'hour 0'. The reason behind this is that while we anticipated beta‐blockers would have differential effect on BP, depending both on the time of day and the time since the drug was taken, the emphasis of this review was on their circadian effect, not their duration of effect after a dose.

We contacted the authors of Neutel 1990; Stephan 1993; Schmieder 1985; and Schmieder 1989 for missing data. Neutel informed us that the original data were no longer available. We did not receive any reply from the other three authors. We were unable to contact other authors for missing data due to lack of contact details.

Assessment of risk of bias in included studies

We assessed the risk of bias in included studies by following Cochrane methodology, detailed in the Cochrane Handbook for Systematic Reviews of Interventions, Chapter 8 (Higgins 2011). Features of included studies assessed included sequence generation (selection bias), allocation sequence concealment (selection bias), blinding of participants, investigators, and outcome assessors, incomplete outcome data (attrition bias), selective outcome reporting (reporting bias), and other potential sources of bias.

Measures of treatment effect

The treatment effect was measured as the mean difference with 95% confidence interval (CI) in systolic and diastolic BP (in mmHg) between baseline and after treatment, or between placebo and treatment group, at each hour over a 24‐hour period. However, only baseline controlled trials met the inclusion criteria in this review.

Unit of analysis issues

Where studies measured ambulatory BP at multiple points during treatment, we extracted data for one treatment period only, to avoid collecting multiple observations from each participant. Where a randomised cross‐over design was employed, we extracted data pre‐cross‐over, when available. Where only combined data for each intervention (pre‐ and post‐cross‐over) was available, we extracted data from a single intervention group only, to avoid unit of analysis error.

See Characteristics of included studies for further details.

Dealing with missing data

Standard deviation (SD) of change in BP at each hour was often not included in the published reports. When we could not obtain this information from trial authors, we imputed standard deviations according to the hierarchy below, which is a modified version of the strategy described by Sekhon 2008.

1. Standard deviation of the change in average daily BP from the same trial

2. Standard deviation of the average BP at each hour at the end of treatment from the same trial

3. Standard deviation of the average daily BP at the end of treatment from the same trial

4. Weighted mean standard deviation of the change in average daily BP from at least three other trials using the same drug and dose regimen

5. Weighted mean standard deviation of average daily BP at the end of treatment from at least three other trials using the same drug and dose regimen

6. Weighted mean standard deviation of change in average daily BP from other trials using the same drug

7. Weighted mean standard deviation of average daily BP at the end of treatment from other trials using the same drug

8. Weighted mean standard deviation of change in average daily BP from all included trials (any drug and dose)

9. Weighted mean standard deviation of average daily BP at the end of treatment from all included trials (any drug and dose)

We imputed the standard deviation according to steps one, and three to nine of the hierarchy at each hour.

For steps four to nine, weighting was by the number of participants in the included trials.

Assessment of heterogeneity

We used Review Manager 5 software's built‐in test for heterogeneity of treatment effects to determine the degree of heterogeneity at each hourly time point (Review Manager 2014).

Assessment of reporting biases

If ten or more studies met the inclusion criteria, we wanted to assess reporting bias using funnel plots, as outlined in the Handbook, Chapter 10 (Sterne 2011). However, since only seven studies met the criteria, we did not create funnel plots.

Data synthesis

All eligible studies identified for inclusion in this review were non‐randomised baseline controlled trials. We entered the mean difference from baseline measurement in BP or HR plus its standard error, and pooled them as generic inverse variance data to obtain a mean difference with 95% CI. We could not use RevMan 5.4 for the analyses of total effects across time points because of correlated errors introduced by repeated observations on the same people. We used the following process instead.

We defined each hourly time point as a categorical variable, with 24 categories corresponding to a 24‐hour day, and numbered 0 through 23. We retrieved the standard deviation for each hourly mean difference (MD) in each study by multiplying the standard error of the MD for that hour by the square root of the study sample size.

In order to simulate random variability of individual subjects around each hourly MD, we randomly drew a sample of individual differences, where the sample size was equal to the study sample size, from a normal distribution centred on the MD, and corresponding standard deviation for that study and hour. Within‐patient correlation across hours in these simulated patients was ignored and assumed to be independent. After simulating a sample of individuals across hours for each study, we calculated an overall MD for each hour, by averaging the differences across all simulated subjects and studies.

We repeated this process 1001 times, and ranked the overall MDs in ascending order. After ranking, we took the 501st observation from the ranked set of overall MDs to be the final estimated MD for that hour. We took the 26th and 976th observations to be the 95% confidence limits. Finally, we completed the process for each hour and in 8‐hour time categories, to summarize the day, evening, and night time differences in BP or HR.

We used SAS and SAS/STAT software, Version 9.4 of the SAS System for Windows to generate the data analysis for this paper (SAS 2012). In particular, we used the RAND function to generate individual subjects to simulate within‐study variation between participants.

Subgroup analysis and investigation of heterogeneity

Due to the limited number of eligible studies identified, we did not carry out subgroup analysis.

Sensitivity analysis

Due to the limited number of eligible studies identified, we did not carry out sensitivity analysis.

Summary of findings and assessment of the certainty of the evidence

We used the GRADE approach to assess the certainty of the evidence for each outcome (Schünemann 2019a; Schünemann 2019b). We presented key findings of the review, including a summary of the data, the magnitude of the effect size, and the overall certainty of the evidence, in summary of findings Table 1. We determined GRADE assessments of certainty by considering five domains: study limitations (risk of bias), inconsistency, indirectness, imprecision, and publication bias. We categorised the certainty in a body of evidence for each outcome as high, moderate, low, or very low.

We presented evidence for the following key outcomes: variation in the decrease in 24‐hour ambulatory hourly SBP, DBP, and HR.