No relevant systematic reviews fulfilling our inclusion criteria were identified, but 34 articles relating to 16 unique individual studies published between 1974 and 2012 were included [
24‐
57]. Thirty-two references were identified from search 3A and two references from the handsearch. The characteristics of the included studies and the corresponding articles are shown in Additional file
2: Table S1. Eleven studies were randomized controlled trials, one study was a non-randomized prospective clinical trial and four studies were observational studies (two retrospective cohort studies and two case control studies). The mean age of participants in the studies ranged from 59 to 83.8 years, but three studies were included based on availability of a subgroup analysis for patients 65 years or older [
24,
36,
40].
Different types of thiazides were studied. Four trials used chlorthalidone as the study drug, three hydrochlorothiazide, and two indapamide. Bendroflumethiazide, methylclothiazide and trichlormethiazide were used in one trial each. Four randomized trials used a fixed combination therapy including a thiazide as a starting drug. Three of the four observational studies did not focus on a specific type of thiazide and included any thiazides. One observational study compared hydrochlorothiazide and chlorthalidone. Seven trials compared thiazides to placebo, one to no treatment. In three trials thiazides were compared to a calcium channel blocker, one of these was a subgroup analysis. In addition one trial compared thiazides to β-blocker in a subgroup analysis.
Clinically relevant outcomes reported in the included studies were cardiovascular morbidity and
mortality including stroke as single outcome measure, all-cause mortality, fractures and adverse
events.
Effects of the use of thiazides in the management of hypertension in older people
For each study and outcome Additional file
4: Table S3 summarises the results for the thiazide and comparison groups, provides estimated risk ratios with 95% confidence intervals, and reports any statistical comparisons of the study itself. To help interpretation, Additional file
4: Table S3 organises the results by type of outcome. The results for Dhalla et al. [
29] are not represented in the table as this observational study compares two different thiazides. We did not identify any study reportin on other outcomes that were covered by our search strategy such as falls or hospitalisation.
Cardiovascular outcomes
Most trials reported on cardiovascular outcomes. Nine trials reported on cardiovascular outcomes as the primary outcome.
Stroke
Four trials evaluated fatal and non-fatal stroke as the primary outcome [
36,
37,
44,
49] with placebo as the comparison and two trials stroke as a secondary outcome against other active treatments [
24,
43]. Study drugs included different types of thiazides. Two large-scale trials showed a significant advantage for thiazides in the reduction of stroke in comparison to placebo [
44,
49]. In addition, a significant reduction of stroke was seen in the SHEP pilot trial which compared chlorthalidone to placebo [
47]. The Hypertension in the Very Elderly Trial (HYVET) reported a 30% reduction for indapamide compared to placebo in the rate of stroke which did not quite reach statistical significance [
37]. The HSCS trial found a non-significant 25% reduction of stroke events in the 65 and older subgroup for deserpidine plus methylchlothiazide compared to placebo but the sample size was small [
36]. Across the full study sample of this trial (stroke survivors with a mean age of 59 years) there was no difference in the occurrence of stroke (including transient ischemic attacks). In comparison to other active treatments (lisinopril, amlodipine, lacidipine, atenolol) in three RCTs [
24,
43,
44], the only significant advantage regarding the risk of stroke was for chlorthalidone compared to lisinopril, reported for the subgroup analysis for patients aged ≥65 years from the ALLHAT trial [
24].
CVD/CHD
Two trials reported on CVD and CHD [
24,
49]. Chlorthalidone was associated with lower risk for CVD and CHD compared to placebo [
49] and compared to lisinopril and doxazosin [
24].
Heart failure
Two trials reported on heart failure [
24,
49]. In the SHEP trial [
49], non-fatal heart failure occurred less often in the chlorthalidone than in the placebo group. In the ALLHAT trial [
49], heart failure was less frequent with chlorthalidone compared to other active treatments (amlodipine, lisinopril, doxazosin).
Combined cardiovascular/cerebrovascular endpoints
Ten studies reported on composite endpoints of cardiovascular morbidity/mortality events, four trials [
24,
40,
41,
43] and one observational study [
29] as primary outcome and five RCTs as secondary outcome [
31,
36,
44,
48,
49]. The specific events making up the composite endpoint varied markedly across studies.
Four trials showed a significant advantage for thiazides compared to placebo: The SHEP trial showed benefits for chlorthalidone compared to placebo regarding nonfatal MI or coronary heart disease (CHD)(see Additional file
4: Table S3) [
49]. The SHEP pilot study showed significantly less hypertensive events associated with chlorthalidone than with placebo but no significant difference for all atherosclerotic events [
46]. The EWPHE trial showed significantly less non-fatal cardiovascular study terminating events for HCT/triamterene compared to placebo [
31]. In the MRC-O trial, coronary events occurred less often in the diuretic group compared to the placebo group [
44]. Three studies compared thiazides to other active treatments and used a composite of cardiovascular outcomes as primary outcome.
In the ACCOMPLISH trial, the combination of an ACE-inhibitor with a calcium channel blocker was more effective for a composite of cardiovascular endpoints than the combination of an ACE-inhibitor with hydrochlorothiazide, in the subgroups for patients ≥65 years and ≥70 years [
40]. In the ALLHAT trial, in participants aged ≥65 years chlorthalidone showed the same advantage regarding a composite of fatal CHD and non-fatal MI compared to both amlodipine and lisinopril. Results on the primary outcome for the older subgroup were not presented for the groups using doxazosin and ACE-inhibitors, but for all participants chlorthalidone showed less cardiovascular events than doxazosin, which led to a premature closing of the doxazosin arm [
24,
26].
The SHELL study compared chlorthalidone to the calcium channel blocker lacidipine and found no significant difference regarding cardiovascular morbidity and mortality [
43].
In addition, the MRC-O trial showed a significantly lower rate of coronary events (as a secondary endpoint) amongst patients on hydrochlorothiazide compared to those taking β-blockers [
44].
Two different types of thiazide, chlorthalidone and HCT, were compared against each other by Dhalla et al. [
29] in a large observational study of 30,000 patients. No significant difference was found on a composite endpoint of death or hospitalization for heart failure, stroke or myocardial infarction.
Mortality
Ten studies reported on all-cause mortality and one trial, the EWPHE trial, analyzed all-cause mortality as the primary outcome [
24,
29,
31,
37,
39,
41,
43,
44,
48,
49]. Comparison groups included no treatment, placebo and other antihypertensive drugs. The HYVET trial showed a significant 21% reduction of all-cause mortality for the indapamide group vs. placebo. In all other trials there was no significant benefit or risk regarding all-cause mortality for the thiazide-treatment groups.
Cardiovascular mortality did not appear as a primary outcome in any study, but was analyzed as a secondary outcome in four trials that compared thiazides to placebo [
31,
37,
39,
44]. The EWPHE and the MRC-O trials showed a significant reduction of cardiovascular mortality in the diuretic treatment group [
31,
44]. In the EWPHE trial, cardiovascular mortality was significantly reduced in the HCT group compared to placebo. In a secondary analysis, the effect of treatment was negatively associated with age, and in the subgroup aged >80 years no effect was observed [
30]. The HYVET trial and the HYVET pilot demonstrated no significant benefit regarding cardiovascular mortality for indapamide and bendroflumethiazide compared to placebo/no treatment [
37,
39].
The MRC-O trial additionally compared HCT to atenolol and found significantly fewer cardiovascular deaths occurred in the hydrochlorothiazide group compared to the β-blocker group. The two treatment groups combined showed no reduction in cardiovascular mortality compared to placebo [
44].
Adverse events
Two trials reported on serious adverse events (SAE) using thiazides compared to placebo [
28,
37]. In the HYVET trial, serious adverse events occurred significantly less often in the indapamide treatment group than in the placebo group, and only five SAEs were judged to be related to study medication (3 in the placebo group, 2 in the indapamide group) [
37]. A RCT that compared the combination of perindopril/indapamide to placebo reported that two SAEs possibly related to study medication occurred in each group [
28].
The SHEP trial reported that the prevalence of intolerable symptoms was higher in the chlorthalidone group compared to placebo, and the EWPHE trial found three symptoms significantly more common in patients on hydrochlorothiazide compared to placebo (dry mouth, nasal stuffiness and diarrhea) but none to the opposite. The EWPHE trial also reported that significantly more treated patients stopped the study medication because of side effects or concomitant disease compared to those taking placebo [
32,
57]. Likewise, in the MRC-O trial withdrawals due to major side effects were considerably higher in the diuretic group than in the placebo group, partly due to increased gout, but also to significantly greater incidence of impaired glucose tolerance, skin disorders, muscle cramp, nausea and dizziness.
Less evidence was available for comparing rates of adverse events on thiazides to other antihypertensive drugs. In the SHELL study, chlorthalidone was compared to lacidipine: fatigue occurred significantly more often in the chlorthalidone group, but edema (mainly pretibial), headaches and skin rashes significantly more often in the calcium channel blocker group [
43]. Other side effects were similar in both groups. In the MRC-O trial, withdrawals due to major side effects were less frequent in the thiazide group than in the β-blocker group, although gout and muscle cramp were more common. In the SHEP trial cases of new-onset diabetes did not differ between chlorthalidone and placebo groups at years one and three [
54]. Finally, the ACCOMPLISH trial found no increased risk of SAE associated with benazepril/HCT treatment compared to benazepril/amlodipine across the main study population with a mean age of 68 years [
40].
Gout
A retrospective cohort study investigated the risk of initiation of anti-gout therapy in hypertensive patients and found a higher risk for patients exposed to thiazides increasing with higher doses. No increased risk could be seen at doses <25 mg/dl [
35], but the wide confidence interval did not exclude a possible association. This study agrees with the findings in the EWPHE trial in which reports of gout were significantly more frequent in the HCT treated group [
31] and the MRC-O trial with high numbers of study medication withdrawal due to gout [
44].
Fractures
Two observational studies investigated a potential protective effect of antihypertensive treatment with thiazides on the incidence of hip fracture [
42,
56]. A case control study with female participants found no difference in the risk of hip fracture between thiazide users and non-users for either current or former thiazide users [
56]. In contrast, a prospective case control study with 9518 patients reported a considerably lower incidence of hip fracture among thiazide users compared to non-users [
42].
Other outcomes
No association between thiazide use and the occurrence of new-onset diabetes [
54], dementia [
38,
45] or depression [
45] was found in the included studies.
Development of recommendations
Based on the evidence identified and the additional references of interest we developed three recommendations which are presented in Additional file
5: Table S4. All recommendations were considered to have a weak strength and low quality of evidence. Additional file
5: Table S4 displays the reasons for the strength and quality of evidence, and the main articles which constitute the evidence base for each recommendation, although all included studies were taken into account for the risk/benefit balance during the team meetings.
We found that thiazides reduced cardiovascular endpoints in comparison to placebo in older people, particularly regarding the risk of stroke, and that benefit was clinically relevant compared to risk.
For comparing thiazides to other antihypertensive drugs, the available evidence is more limited for our age group of interest. Data comparing thiazides with calcium channel blockers are available from three trials, with conflicting results. In one randomized controlled trial (ACCOMPLISH) the combination of an ACE inhibitor with hydrochlorothiazide was less effective than the combination of an ACE inhibitor with the calcium channel blocker amlodipine [
40]. This is in line with the recommendations of the NICE guideline to choose a calcium channel blocker as first line therapy in adults aged over 55 years [
6]. In contrast, two trials showed no significant difference between thiazides and calcium channel blockers [
24,
43]. Based on this evidence, a general recommendation that thiazides are less effective than calcium channel blockers in older people could not be developed, but a specific recommendation for the combined treatment with benazepril and hydrochlorothiazide based on the evidence from ACCOMPLISH was made. This recommendation is currently discussed by the editorial team.
Data comparing thiazides with β-blockers were available from the MRC-O trial [
44]. In an analysis comparing the two treatment groups, patients with hydrochlorothiazide had fewer coronary events. Studies comparing thiazides with ACE inhibitors in the treatment of hypertension for older patients could not be identified.
The included studies reported on different types of thiazides and we did not find any evidence that certain types of thiazides may be safer for older people. Beneficial effects were not restricted to any particular type of thiazide.
In the studies included in our review results were mixed regarding any association with adverse events, except for the increased risk of gout [
32,
35,
44]. In the absence of studies showing clinically relevant outcomes of adverse effects in the age group ≥65 years, and taking into consideration the evidence on potential benefits, we did not find any justification for the development of a recommendation to discontinue thiazides in general for the management of hypertension in older people.
Aside from the risks of thiazide treatment, a potential protective effect of thiazides on the risk of fractures, due to the reduction of urinary calcium excretion, has been hypothesised. Two observational studies in our SR addressed this topic but showed conflicting results [
42,
56]. In two other observational studies including older adults, thiazide users did not have a lower risk of hip fracture [
60,
61], while another observational study showed a protective effect for long-term use, and particularly for high dose thiazide use [
62]. Thus, evidence is not convincing to recommend thiazides for fracture prevention in older adults with hypertension.
Use of thiazides was associated with increased risk of gout in one observational study, and in two randomized controlled trials the incidence of gout was higher in the diuretic group [
32,
35,
44]. Thiazides are listed on the STOPP list for patients with a history of gout and should be carefully considered in patients with gout according to the American College of Rheumatology Guidelines for Management of Gout [
9,
59]. The observational study found no significant relationship with gout at lower doses of thiazides, although the confidence interval was wide. Furthermore, in a SR of interest on morbidity and mortality in the management of hypertension with different drugs, low doses of thiazides (25 mg/d HCT equivalent) were superior in reducing death and coronary artery disease than high doses [
58], although this systematic review did not focus on older people. Considering this evidence as a whole, we developed a recommendation to adopt low dose thiazide treatment in general. This recommendation was discussed and approved by the editorial team.
A third recommendation was developed based on the evidence from the EWPHE trial where hydrochlorothiazide/triamterene demonstrated no benefit in the age group of 80 years and over. Evidence on the effectiveness of other types of thiazides in very old people was lacking, so this recommendation was restricted to this specific treatment. This recommendation is currently discussed by the editorial team and a final decision has not been reached.