Patients (≥ 18 years) were recruited from the Hypertension Clinic at Aarhus University Hospital, Denmark and from two local private cardiology clinics. Suspected treatment resistant hypertension was the main reason for attending these clinics. Exclusion criteria were pregnancy, change in antihypertensive medication within 14 days prior to inclusion, treating physician preference of using an ABPM-device other than the devices described below, arm circumference above or below available cuff sizes (20–42 cm), inability to give informed consent, or not expecting to sleep during the night on either day of ABPM (nightshift work or other reasons). Patients were not invited to participate if permanent- or persistent atrial fibrillation were noted in their electronic patient records three years prior to inclusion. Information about patient characteristics, comorbidities, and medication were obtained from electronic patient records.

Included patients were assigned a hypertension category based on their first ABPM. Normotension/controlled hypertension was defined as 24 h SBP/diastolic BP (DBP) < 130/80 mmHg, mild hypertension as 24 h SBP 130–149 mmHg and moderate/severe hypertension as 24 h SBP ≥ 150 mmHg. Chronic kidney disease was defined as estimated glomerular filtration rate below 60 ml/min/1.73 m2 on two consecutive blood samples more than three months apart or when urine albumin-creatinine ratio was above 30 mg/g in two out of three urine samples. Diabetes was defined as a hemoglobin A1c concentration above 48 mmol/mol or at least one prescribed antidiabetic drug. Ischemic heart disease was considered present in patients with previous coronary interventions, if significant coronary calcification had been documented on cardiac computerized tomography scan or coronary angiography or in the case of clear angina symptoms.

Prior to the first ABPM, a Microlife watchBP Office BP device (Microlife, AG Widnau, Switzerland) was used for bilateral BP measurements. If the inter-arm difference was larger than 10/5 mmHg (SBP/DBP) the ABPM device was fitted to the arm with the highest BP. If not, patients were given the option to choose which arm the cuff was fitted on. The same arm was used for both ABPMs.

Spacelabs Ontrak and Spacelabs 90217A (Spacelabs Healthcare, Snoqualmie, Washington, USA) were used for the ABPMs. Both devices are validated according to internationally recognized standards [20]. Arm circumference was measured to select the correct cuff size. Sentinel software (v11.5.2.13260, Spacelabs Healthcare, Snoqualmie, Washington, USA) was used to retrieve data from the ABPM device. A recording was considered acceptable if 70% or more of the measurements were successful [21]. The software was preset to discard any SBP measurements outside of the range 70–240 mmHg, and any DBP outside the range 40–150 mmHg.

During high frequency ABPM (HF-ABPM), daytime was predefined as 07:00–23:00 and BP was measured every 20 min in this time interval and every 30 min for the remaining eight hours. Low frequency ABPM (LF-ABPM) measured BP once every hour during all 24 h. Patients reported individual bed- and rising times after each ABPM, and these were used to define awake and sleep measurements post hoc. If a planned measurement failed, the device was programmed to perform one extra measurement within two minutes. Initially, four measurements were made during fitting to ensure that the device worked correctly. These initial measurements and measurements performed after 24 h of recording, were removed from our analysis.

All patients scheduled for a planned ABPM were screened and eligible patients were asked to participate. Patients who accepted to enter the study were randomized 1:1, using an online random number generator [22], to either HF-ABPM or LF-ABPM as their first recording.

Patients were prepared for the ABPM in accordance with current guidelines [1], and instructed to keep the two measuring days similar in terms of physical activity, bed/rising times, medication, and time of medication ingestion. We did not apply restrictions concerning caffeine or alcohol intake. The second ABPM was conducted using the alternative measurement frequency no earlier than 48 h after finishing the first ABPM. At the time of the second ABPM, neither the person fitting the cuff nor the participant, were aware of the results of the first measurement. To minimize inter-device variation, the exact same ABPM device and cuff was used in both ABPM for almost all patients. However, three patients were unable to return to the hospital for the second fitting and were therefore handed two devices, each configured to one of the frequencies. They were fitted with the first device and instructed how and when to fit the second device themselves at home.

Immediately after each ABPM, the patients filled in a questionnaire concerning antihypertensive medication taken for the last 24 h, self-perceived sleep interruptions attributed to the measurements, symptoms from the arm after removal of the cuff, and pain perceived during day and night measurements. Pain associated with cuff inflations was reported on a visual analogue scale (VAS) from 0 to 10. Only patients who completed both questionnaires were included in the questionnaire analysis.

The primary outcome was the difference in 24 h mean SBP between HF-ABPM and LF-ABPM. Secondary outcomes were mean differences in SBP for day- and nighttime, DBP values (24 h, daytime, and nighttime means), 95% limits of agreement (95% LoA) interval, coefficient of variation (CV) for both SBP and DBP values calculated as standard deviation (SD) / mean × 100, and questionnaire data as reported above. The mean difference and the LOA between HF- and LF-ABPM were calculated using Bland-Alman plots and the relationships between differences and means were tested with simple linear regression and plotted appropriately, 24 h means are also presented in scatter plots with a line of equality as reference [23]. Using a 2-sided significance level of 0.05, a power of 0.9, and a SD of 13.5 mmHg [19] we calculated a needed sample size of about 120 patients to detect a 4 mmHg SBP difference between LF-ABPM and HF-ABPM.

Data were analyzed using STATA (Version 17, StataCorp, College Station, TX, USA). Continuous variables were evaluated for normal distribution using QQ-plots and histograms. BP results are presented as means with 95% confidence intervals (CI). All other continuous, normally distributed variables are presented as means ± SD, and data with a skewed distribution are presented as median with interquartile range (IQR). Dichotomous variables are presented as number of patients with % of total population. For independent data unpaired t-test, Mann-Whitney-U test and chi-squared-test were used for normally distributed, skewed, and dichotomous data, respectively. For paired data a paired t-test, Wilcoxon matched-pairs signed-rank test, and McNemar’s test were used for normally distributed, skewed, and dichotomous, data respectively. A P-value less than 0.05 was considered statistically significant. The main author has full access to all the data in the study and takes responsibility for its integrity and the data analysis.