Predictive value of ambulatory blood pressure shortly after withdrawal of antihypertensive drugs in primary care patients

Introduction

In hypertensive patients, antihypertensive drugs are usually given for an undetermined period of time. In patients with well controlled blood pressure, withdrawal of medication can be considered, particularly in those patients who comply with non-pharmacological recommendations or who suffer from side effects of the drugs that affect their quality of life.1 2 Fletcher et al reviewed case reports and randomised trials and found that success of withdrawal, defined as the proportion of patients remaining normotensive after withdrawal of medication, varied from 15% to over 50%.1 However, it takes at least six to 12 months of follow up before a steady state has been reached because blood pressure may increase slowly after withdrawal of medication.

An indicator of likely return to raised blood pressure in an early phase after withdrawal of medication would be desirable. Ambulatory blood pressure monitoring might be helpful in this regard. The reproducibility of ambulatory blood pressure is superior to that of seated blood pressure.3 To determine whether ambulatory blood pressure measured soon after stopping antihypertensive drugs is a more sensitive measure than seated blood pressure to predict blood pressure in the long term, we followed primary care patients who fulfilled the criteria for withdrawal of antihypertensive drugs for one year after withdrawal of medication.

Methods

A total of 145 patients (59% men) with untreated diastolic hypertension (four measurements >/=95 mm Hg on three occasions) aged between 25 and 75 years were included in an one year trial of treatment. Patients were randomised to treatment with amlodipine 5 mg or lisinopril 10 mg. Seated blood pressure was measured before and 4, 6, 12, 26, 39, and 52 weeks after the start of antihypertensive drugs. The therapeutic response was defined as a reduction in the average diastolic blood pressure (measured while the patient was seated) to a value of </= 90 mm Hg or as a fall from baseline of at least 10 mm Hg to a value of </= 100 mm Hg. For patients who did not meet the therapeutic response criterion after four weeks, the dosage was doubled. After one year of treatment, 122 patients receiving drug treatment completed the study. Reasons for not completing the study were: no therapeutic response after 12 weeks of treatment (two patients), withdrawal of informed consent (nine), hypotension (one), adverse events (nine), and angina pectoris (two). Ambulatory blood pressure was measured before and one year after the start of antihypertensive drugs (weeks 0 and 52).

Results

Thirty four of the 122 patients who completed the first year of treatment fulfilled the criteria for withdrawal of antihypertensive drugs. Twenty nine of these 34 patients agreed to participate in the follow up study. Their mean age was 51 (SE 2) years; 18 (63%) were male; mean body mass index (kg/m²) was 26.3 (0.6). After one year of follow up, data from 25 patients could be analysed for seated measurements and from 23 for ambulatory measurements. Reasons for not completing the study were: restart of antihypertensive drugs because of side effects of stopping (one patient) or increase in blood pressure (one); withdrawal of informed consent (two); refusal of remeasurement of ambulatory blood pressure after one year (two).

After eight weeks without antihypertensive drugs, mean diastolic blood pressure increased to the pretreatment level for ambulatory measurements whereas mean diastolic blood pressure for seated measurements did not change (table 1). After 52 weeks without antihypertensive drugs, mean diastolic blood pressure returned to the pretreatment level for both seated and ambulatory measurements. No differences in mean systolic blood pressure were found for seated measurements. The mean systolic blood pressure for ambulatory measurements was significantly lower after 52 weeks of treatment than at weeks 0, 60, and 104.

Six of the nine patients with raised ambulatory blood pressure at week 104 were already hypertensive at week 60 (table 2); sensitivity was thus 66%. All 14 patients with normal ambulatory pressure at week 104 had a normal pressure at week 60 (specificity 100%). The positive and negative predictive values were 100% and 82%, respectively. Table 2 shows that none of the five patients with raised seated blood pressure at week 104 were hypertensive at week 60 (sensitivity 0%). Nineteen of the 20 patients with normal seated blood pressure at week 104 had normal pressure at week 60 (specificity 95%). The positive and negative predictive values were 0% and 79%, respectively.

Receiver operating characteristic curves for seated and ambulatory blood pressure show that the better prediction with ambulatory blood pressure than with seated blood pressure was not dependent on the cut off value that was used (fig 1). The power of ambulatory measurements after eight weeks of withdrawal to predict ambulatory blood pressure after one year of withdrawal was large: the area under the curve was 0.92. The S shape of the curve for seated measurements was caused by a false positive case having the highest seated blood pressure at week 60 (see table 3). The area under the curve was 0.72.

Fig 1

Receiver operating characteristic curves for seated and ambulatory blood pressure

• Download figure
• Open in new tab
• Download powerpoint

Discussion

This study shows that ambulatory blood pressure measured eight weeks after withdrawal of antihypertensive drugs was more sensitive than seated blood pressure for predicting blood pressure in the long term. The sensitivity and the positive predictive value were superior for ambulatory measurements, and the specificity and negative predictive value were comparable for both types of measurements. Most patients with raised ambulatory blood pressure in the long term could be identified eight weeks after withdrawal of medication. Therefore, restarting of antihypertensive drugs in this early phase would have been justified on the basis of early ambulatory blood pressure monitoring.

The incidence of raised seated blood pressure eight weeks after withdrawal of medication was low, occurring in only two patients. One year after withdrawal of medication, one of these patients seemed normotensive and the other was excluded from the analysis because antihypertensive drugs were restarted. This low incidence of raised seated blood pressure, together with the relatively small number of patients, explains the very low sensitivity and positive predictive value of seated blood pressure found in this study. It is unlikely, however, that the large differences in area under receiver operating characteristic curves can be explained by the results in these patients. It is also unlikely that the superior predictive value of the ambulatory blood pressure after withdrawal of medication is dependent on regression to the mean. Since blood pressure was normal at all five visits during a period of 46 weeks, it is reasonable to assume that a steady state was reached.

It is well known that ambulatory blood pressure monitoring is a useful tool in a research setting. The method of recording of the blood pressure is more objective than for measurements in clinics or surgeries. Since mean blood pressure returned to the pretreatment level for ambulatory measurements, the results of this study suggest that seated measurements after withdrawal of medication are subject to a kind of placebo effect. Moreover, the relatively rapid return to pretreatment levels suggests that treatment in these patients with mild hypertension did not result in structural changes which influence the blood pressure.

No other studies have evaluated ambulatory blood pressure after withdrawal of antihypertensive medication. The present study shows that ambulatory blood pressure measured eight weeks after withdrawal of antihypertensive medication is more sensitive for predicting blood pressure in the long term than is seated blood pressure. These results should be confirmed by a double blind, placebo controlled study.

We thank K H Groenier and AVerhoeven for their support.