Open questions — hypertension¶
Last curated: 2026-09-01 · replaces the seed set written 2026-09-01
The test applied. With roughly three-quarters of a million PubMed records, almost any loosely phrased question has already been studied. A question earns a place here only if a specific decision is left unresolved by specific retrievable evidence — usually two results that point different ways, or a recommendation that rests on a study design that cannot support it. Questions that failed that test were dropped rather than dressed up. Each entry states what is known, what is not, and what would settle it.
Tier 1 — would change practice, and could be designed today. Tier 2 — important, but either not yet designable, or the answer would refine rather than change practice.
Tier 1¶
OQ-1 — Should everyone with hypertension be screened for primary aldosteronism?¶
What is known. Systematic confirmatory testing across the blood-pressure spectrum found biochemically overt primary aldosteronism in 11.3% (95% CI 5.9–16.8) of normotensive people, 15.7% of stage 1, 21.6% of stage 2 and 22.0% of resistant hypertension, with the aldosterone-renin ratio showing poor sensitivity and negative predictive value (Brown 2020, PMID 32449886). Compared with essential hypertension, primary aldosteronism carries roughly 2.6-fold odds of stroke and 3.5-fold of atrial fibrillation (Monticone 2018, PMID 29129575). The 2025 Endocrine Society guideline now suggests screening all people with hypertension (Adler 2025, PMID 40658480).
What is not known. No general hypertension guideline has adopted universal screening, and no health system has published the false-positive burden, downstream testing volume or cost of doing so in a population where the underlying trait is continuous rather than dichotomous — which is the authors' own caveat (Brown 2020, PMID 32449886).
What would settle it. A modelled or pragmatic implementation study reporting, per 1,000 people screened: intermediate results, confirmatory tests triggered, adrenalectomies performed, and treatment changes made. This is designable now.
OQ-2 — Does treating subclinical renin-independent aldosteronism change outcomes?¶
What is known. In 1,284 population-based participants aged 40–69, a higher aldosterone-to-renin ratio below the diagnostic threshold was independently associated with arterial stiffness, adverse cardiac remodelling on MRI, left ventricular hypertrophy (OR 1.32, 95% CI 1.002–1.73) and incident hypertension (OR 1.29, 1.03–1.62), including in the normotensive subgroup alone (Hundemer 2024, PMID 38031887).
What is not known. Whether mineralocorticoid receptor antagonism in this group prevents anything.
What would settle it. NCT07727252 — a phase 2/3 trial of eplerenone versus placebo in subclinical primary aldosteronism (n=880) — is not yet recruiting. It is the single most directly targeted trial in this list.
OQ-3 — What office reading corresponds to SPRINT's <120 mm Hg?¶
What is known. SPRINT achieved 121.4 mm Hg by unattended automated office measurement (SPRINT 2015, PMID 26551272). Pooled attended-versus-unattended differences are 3.66/1.67 mm Hg but with I² of 97.1% and 89.0%, so the pooled estimate is unusable as a conversion (Andreadis 2019, PMID 31290085). Attended and unattended readings correlate near-identically with left ventricular mass and carotid intima-media thickness (Salvetti 2019, PMID 30686088). KDIGO responded by mandating standardised office measurement (KDIGO 2021, PMID 33637192); NICE by mandating out-of-office confirmation (Jones 2020, PMID 32001477); most bodies leave it implicit.
What is not known. Whether a clinically usable mapping exists at all, or whether the heterogeneity is irreducible because it reflects setting, operator and rest period rather than a fixed offset.
What would settle it. A prospective multicentre study measuring both methods under standardised conditions across settings, reporting whether the between-setting variance component exceeds the offset. Designable now, and cheap.
OQ-4 — Should asymptomatic elevated blood pressure be treated in hospital?¶
What is known. Three large observational studies agree it is harmful: intensive inpatient treatment carried a composite adverse-outcome OR of 1.28 (95% CI 1.18–1.39), rising to 1.90 (1.65–2.19) with intravenous agents (Anderson 2023, PMID 37252732); treated inpatients had more acute kidney injury (10.3% vs 7.9%) and myocardial injury (1.2% vs 0.6%) with no blood-pressure interval in which treatment looked better (Rastogi 2021, PMID 33369614); as-needed medication carried an acute kidney injury HR of 1.23 (1.18–1.29) and a composite MI/stroke/death HR of 1.69 (1.49–1.92) (Canales 2025, PMID 39585709).
What is not known. Whether the association is causal or reflects confounding by illness severity. No guideline sets an inpatient target.
What would settle it. A randomised trial of a treat-versus-observe policy for asymptomatic inpatient elevation. The Canales authors state explicitly that equipoise exists. This is the largest completely unaddressed evidence gap in the condition.
OQ-5 — Where is the frailty threshold beyond which antihypertensive treatment is net-harmful?¶
What is known. Relative benefit persists to age 75–84 (HR 0.91 per 5 mm Hg, 95% CI 0.87–0.96) with larger absolute reductions in older groups, and only the ≥85 stratum is null (0.99, 0.87–1.12) (BPLTTC 2021, PMID 34461040). Trial-level meta-analysis shows no association with falls (RR 1.05, 0.89–1.24) (Albasri 2021, PMID 33568342), and intensive treatment reduces orthostatic hypotension (OR 0.93, 0.86–0.99) (Juraschek 2021, PMID 32909814). But observational data show serious adverse events rising with frailty (Sheppard 2023, PMID 37075078), fracture risk after initiation in nursing-home residents (Dave 2024, PMID 38648065) and harm in people with dementia (Fujiwara 2025, PMID 40961130). Cochrane names frail and ≥80-year-old populations as the priority evidence gap (Falk 2024, PMID 39688187).
What is not known. Whether the trial reassurance transfers to people who were never in the trials.
What would settle it. A randomised trial in a frailty-defined population — feasible in care-home networks, and OPTIMISE2 (protocol, PMID 41689048) is a partial step toward it.
OQ-6 — Should there be a diastolic floor when treating systolic pressure?¶
What is known. In 11,565 ARIC participants over 21 years, diastolic pressure <60 mm Hg carried 2.2-fold adjusted odds of high-sensitivity troponin-T ≥14 ng/L versus 80–89 mm Hg, predicted progressive troponin rise, and predicted incident coronary heart disease and mortality but not stroke — most pronounced when systolic pressure was ≥120 mm Hg (McEvoy 2016, PMID 27590090). No target trial used a diastolic floor, and SPRINT, ESPRIT and BPROAD were all positive without one (SPRINT 2021, PMID 34010531; Liu 2024, PMID 38945140; Bi 2025, PMID 39555827).
What is not known. Whether applying a floor prospectively prevents harm or simply undertreats systolic pressure.
What would settle it. A randomised comparison of an intensive systolic target with and without a diastolic stopping rule, in people with wide pulse pressure — the group in whom the two conflict.
OQ-7 — Is salt substitution safe and effective where sodium is embedded in processed food?¶
What is known. SSaSS cut stroke (rate ratio 0.86, 95% CI 0.77–0.96), major cardiovascular events (0.87, 0.80–0.94) and death (0.88, 0.82–0.95) in rural China, without excess hyperkalaemia events (Neal 2021, PMID 34459569), and pressure effects are consistent across geographies with a dose–response by substitution fraction (Yin 2022, PMID 35945000).
What is not known. In high-income settings most sodium is added during food manufacture, not at the table, so the intervention has far less to act on; and people with CKD or on renin-angiotensin blockade plus a mineralocorticoid receptor antagonist were largely excluded, so the safety result does not transfer (Brand 2022, PMID 35944931).
What would settle it. Two recruiting trials address the safety half — NCT07460882 (serum potassium in 607 unselected users) and NCT07178964 (kidney transplant recipients). The efficacy half in a processed-food environment remains untested.
OQ-8 — Should biochemical adherence testing precede treatment escalation?¶
What is known. Pooled non-adherence in apparent treatment-resistant hypertension is 31.2% (95% CI 20.2–44.7), with the assessment method the dominant source of variance — self-report and pharmacy data understate it relative to mass spectrometry or directly observed therapy (Durand 2017, PMID 28777133). Chemical testing gives pooled complete non-adherence of 15.0% and any non-adherence 33.0% (Highton 2025, PMID 40371625). Feeding results back was associated with pressure reduction and improved adherence (Gupta 2017, PMID 28847892).
What is not known. Whether routine testing before escalation reduces unnecessary drug additions, procedures and cost — and whether it damages the therapeutic relationship.
What would settle it. A randomised trial of test-before-escalate versus usual care in apparent resistance, with drug additions, denervation referrals and blood pressure as endpoints. Designable now.
OQ-9 — Do renal denervation's pressure effects translate into prevented events?¶
What is known. Sham-controlled trials give 24-hour ambulatory systolic reductions of about 4–6 mm Hg — SPYRAL HTN-OFF MED Pivotal −3.9 mm Hg (Bayesian 95% CrI −6.2 to −1.6) (Böhm 2020, PMID 32234534); pooled ultrasound denervation −5.9 mm Hg (95% CI −8.1 to −3.8) (Kirtane 2023, PMID 36853627); meta-analysis of 10 trials −4.4 mm Hg (−6.1 to −2.7) (Azizi 2026, PMID 41870448).
What is not known. Every trial has a blood-pressure endpoint. No device trial has ever reported cardiovascular outcomes, and none in the current registry is designed to.
What would settle it. An outcome trial. Whether one is commercially viable given a ~5 mm Hg effect is the real question, and it is not a scientific one.
OQ-10 — What is the active ingredient in the implementation trials?¶
What is known. Three trials produced among the largest effects in the hypertension literature: village-doctor-led care raised control to <130/80 from 19.9% to 57.0% (difference 37.0 percentage points, 95% CI 34.9–39.1) (Sun 2022, PMID 35500594); a barbershop pharmacist programme lowered systolic pressure 21.6 mm Hg more than active control (14.7–28.4) (Victor 2018, PMID 29527973); a non-physician health-worker model lowered systolic pressure 11.45 mm Hg more (Schwalm 2019, PMID 31488369).
What is not known. Each bundled protocolised titration, non-physician prescribing authority, relocation of care, free or subsidised medication and social support. None has been factorially decomposed, and none has been compared with another.
What would settle it. A factorial or head-to-head implementation trial. Expensive, unglamorous, and the highest-value unfunded study in this condition.
OQ-11 — Does communicating absolute rather than relative benefit change adherence?¶
What is known. Across qualitative studies the recurring request is for decision-relevant information: what the number means and what treatment buys (Malkon 2023, PMID 38106370; van Bussel 2019, PMID 31427342), and self-monitoring users report not knowing what target they are aiming at (Natale 2023, PMID 36840919). BPLTTC's own clinical corollary is that clinicians should emphasise risk reduction rather than the pressure value (BPLTTC 2021, PMID 33933205).
What is not known. No trial has tested absolute-risk communication as an intervention on adherence or control.
What would settle it. A pragmatic randomised trial embedded in primary care. Cheap and designable now.
OQ-12 — Should renin guide treatment selection in ordinary hypertension?¶
What is known. Spironolactone's superiority as a fourth agent was greatest at low plasma renin, though present throughout the renin distribution (Williams 2015, PMID 26414968), and its mechanism substudies tie the advantage to sodium retention (Williams 2018, PMID 29655877). In medically treated primary aldosteronism, excess cardiovascular risk was confined to patients whose renin remained suppressed (HR 2.83, 95% CI 2.11–3.80), with no excess in those titrated to unsuppressed renin (Hundemer 2018, PMID 29129576). The ALLHAT race interaction — worse stroke and composite outcomes with lisinopril versus chlorthalidone in Black participants, with no interaction for amlodipine (Wright 2005, PMID 15811979) — is consistent with a renin-phenotype explanation.
What is not known. No trial has randomised treatment selection or titration on renin.
What would settle it. A renin-guided versus guideline-guided treatment strategy trial. This would also test whether renin is the biological variable that race is currently used as a proxy for.
Tier 2¶
OQ-13 — Is the observational sodium J-curve real or a measurement artefact?¶
Randomised trials show a monotonic dose–response for pressure down to low intake, with short trials underestimating the effect by more than half (Huang 2020, PMID 32094151). PURE found higher event risk below 3 g/day estimated sodium excretion (OR 1.27, 95% CI 1.12–1.44) (O'Donnell 2014, PMID 25119607), with high sodium associated with events mainly in hypertensive individuals (Mente 2016, PMID 27216139). The dispute is about spot-urine estimating equations and reverse causation; umbrella review does not resolve it (Kong 2025, PMID 41243115). A randomised trial with hard endpoints at low intake would settle it and will almost certainly never be done.
OQ-14 — Is visit-to-visit variability a target or a marker?¶
Top-decile visit-to-visit systolic standard deviation carried a stroke HR of 6.22 (95% CI 4.16–9.29), and maximum systolic pressure 15.01 (6.56–34.38), independent of mean pressure; residual on-treatment variability predicted events independently of clinic and ambulatory means (Rothwell 2010, PMID 20226988). Drug classes differ in the variability they produce (Webb 2011, PMID 21817143; Mehlum 2020, PMID 32336236). No trial has randomised to a variability-reducing strategy with a hard endpoint.
OQ-15 — Can any cuffless device be validated for diagnosis?¶
Pooled biases of 3.42 and 1.16 mm Hg with I² >87%, and precision often unassessed (Islam 2022, PMID 36713001). ESH has issued a dedicated validation protocol precisely because existing protocols do not test calibration drift and behaviour between calibrations (Stergiou 2023, PMID 37303198). The 2024 Chinese guidelines already admit validated wearables for screening (Liu 2025, PMID 39762483), creating a direct international conflict. At least two consumer-device validation reports in this literature have been retracted.
OQ-16 — Should nocturnal pressure be a treatment target?¶
Night-time systolic pressure is the most informative single index — 591% as informative as clinic pressure for all-cause death (Staplin 2023, PMID 37156250) — and an estimated 13.3% of US adults have isolated masked asleep hypertension (Li 2021, PMID 33112362). No outcome trial has randomised on a nocturnal target. Chronotherapy, the obvious intervention, was neutral in TIME (Mackenzie 2022, PMID 36240838) after a contested single-programme claim of large benefit (Hermida 2020, PMID 31641769; Expression of Concern PMID 32318736).
OQ-17 — Does white-coat hypertension carry excess risk?¶
Cohort meta-analysis of 27 studies found untreated white-coat hypertension associated with cardiovascular events (HR 1.36, 95% CI 1.03–2.00) and all-cause mortality (1.33, 1.07–1.67), with no excess for treated white-coat effect (Cohen 2019, PMID 31181575). The largest single registry, 59,124 patients, found no excess all-cause or cardiovascular mortality for white-coat hypertension at all (Staplin 2023, PMID 37156250). Both are shown in this knowledge base; neither has been reconciled.
OQ-18 — Why did US blood-pressure control fall after 2014?¶
Control rose from 31.8% (1999–2000) to 53.8% (2013–2014) and fell to 43.7% (2017–2018) (Muntner 2020, PMID 32902588), during a period of expanding generic availability and intensified guideline attention. Comorbidity burden rose over the same interval and control of hypertension plus diabetes plus hyperlipidaemia has been stuck at about a quarter for a decade (Lee 2025, PMID 41295934). Whether NHANES is representative has itself been argued (Egan 2023, PMID 37967159). No account explains the reversal.
OQ-19 — Should treatment thresholds be sex-specific?¶
Sex-specific longitudinal modelling of 144,599 observations across 43 years found systolic, diastolic, mean arterial and pulse pressure rising faster in women from the third decade onwards, persisting after risk-factor adjustment (Ji 2020, PMID 31940010). Every guideline threshold is sex-neutral. No trial has randomised on a sex-specific threshold, and no risk equation in general use treats blood pressure trajectory as sex-specific.
OQ-20 — Is there an intervention window in adolescence?¶
Blood pressure tracks from childhood with correlations around 0.38 systolic (Chen 2008, PMID 18559702), and over 38 years the probability of reverting from stage 2 hypertension to normal pressure by midadulthood was 0.23 (95% CI 0.19–0.26) in males versus 0.58 (0.52–0.62) in females, with adolescence rather than childhood identified as the period after which reversion becomes unlikely (Meng 2025, PMID 39495520). Nothing has been trialled in that window.
OQ-21 — Would an immunomodulatory intervention lower blood pressure in humans?¶
Adaptive immune activation is required for full expression of experimental hypertension (Caillon 2019, PMID 29952002), and the SH2B3/LNK GWAS locus has a demonstrated immune mechanism in a mouse knock-in (Alexander 2022, PMID 36169218). No immunomodulatory therapy for hypertension has reached late-phase trials, and the safety bar for a lifelong preventive therapy is high.
OQ-22 — Does tissue sodium storage invalidate 24-hour urinary sodium as an exposure measure?¶
²³Na MRI shows sodium accumulating in skin and muscle beyond what measured water accounts for, higher with age and in hypertension (Kopp 2013, PMID 23339169), mobilisable by dialysis (Dahlmann 2015, PMID 25100048), and altered after renal denervation (Ott 2018, PMID 28845508). If sodium balance is not purely renal (Titze 2014, PMID 24401786), both the interpretation of urinary sodium in cohort studies and the classical pressure-natriuresis argument need modification. Nobody has quantified how much.
OQ-23 — Is the pressure-natriuresis account of chronic hypertension correct?¶
The Guytonian claim that sustained hypertension requires a rightward shift of the renal pressure-natriuresis relationship (Guyton 1980, PMID 6994602; Hall 1986, PMID 3536587) underpins the field's framing. It has been challenged on the grounds that the infinite-gain property is an artefact of how the equations are posed and that the nervous system's long-term role is under-weighted (Kurtz 2016, PMID 28637271; Osborn 2009, PMID 19286640; Beard 2013, PMID 24555102), with a "neo-Guytonian" middle position proposed (Evans 2016, PMID 26582636). This is a live argument about the foundation of the discipline that most clinical writing does not acknowledge.
OQ-24 — Can polygenic scores select drugs?¶
A systolic PRS predicted chlorthalidone response in Black ALLHAT participants — lowest versus median quintile ΔSBP −10.01 (95% CI −11.11 to −8.90) versus −6.57 (−7.67 to −5.48) mm Hg over six months, drug-specific and independently validated, with 67% higher odds of apparent treatment resistance in the top quintile (Armstrong 2024, PMID 39441603). The effect is smaller than the between-drug differences already free of charge, and was derived in one ancestry group in one trial; portability across ancestries is unsolved (Krieger 2026, PMID 42395874).
OQ-25 — Which guideline strategy for the measurement problem is right?¶
Three approaches exist: mandate the trial's measurement method (KDIGO 2021, PMID 33637192), mandate out-of-office confirmation before diagnosis (NICE, Jones 2020, PMID 32001477; USPSTF 2021, PMID 33904861), or leave it implicit (most bodies). No comparative evaluation exists of which produces better population control or fewer misclassifications.
OQ-26 — Can a global guideline serve settings whose optimal thresholds differ by 20 mm Hg?¶
Economic modelling across 24 low- and middle-income countries found net benefit maximised by treating from systolic ≥160 mm Hg in the lowest-income settings, with 68% of total benefit coming from that group, and positive net benefit at some cut-point in only 3 of the 12 lowest-income countries (Hutchinson 2024, PMID 38626959). ISH's two-tier essential/optimal standards are the only guideline structure that acknowledges this (Unger 2020, PMID 32370572). No clinical guideline states a resource-conditional threshold.
OQ-27 — Is there a hypertension-specific patient-reported outcome measure worth building?¶
Patients describe treatment burden — pill count, appointment frequency, monitoring anxiety, cost — as the principal cost of care (Natale 2023, PMID 36840919; Malkon 2023, PMID 38106370), and trials measure none of it. No validated hypertension-specific treatment-burden instrument is in general use.
Dots not yet connected¶
Cross-domain junctions where two bodies of evidence in this knowledge base both exist, and no study has joined them.
| # | Dot A | Dot B | The missing junction | Powers |
|---|---|---|---|---|
| 1 | Primary aldosteronism present in 11–22% across the pressure spectrum (PMID 32449886) | Resistant-hypertension treatment algorithms that proceed empirically to spironolactone (PMID 26414968) | Nobody has asked whether empirical mineralocorticoid receptor antagonism is a cheaper substitute for diagnosis, or whether diagnosis changes management enough to justify its cost | OQ-1, OQ-12 |
| 2 | Excess cardiovascular risk in treated primary aldosteronism confined to persistently suppressed renin (PMID 29129576) | Low-renin hypertension is common in ordinary practice and predicts spironolactone response (PMID 26414968) | No trial titrates ordinary hypertension to raise renin rather than to lower pressure | OQ-12 |
| 3 | Measurement-method offsets are heterogeneous and unconvertible (PMID 31290085) | Guidelines set numerical targets derived from one specific method (PMID 26551272, PMID 33637192) | The threshold debate and the measurement debate are conducted in separate literatures although one determines the other | OQ-3, OQ-25 |
| 4 | Implementation trials that moved population control by 20–37 percentage points (PMIDs 35500594, 29527973, 31488369) | Target-setting trials that moved achieved pressure by 10–15 mm Hg (PMIDs 34010531, 39555827) | Lowering a recommended number has no population effect if half of those affected are undiagnosed; no study has modelled the two levers against each other | OQ-10, OQ-18 |
| 5 | Systolic pressure rises faster in women from the third decade (PMID 31940010) | Every guideline threshold and risk equation is sex-neutral (PMID 40811516) | No sex-specific threshold has been derived from the trajectory data, let alone tested | OQ-19 |
| 6 | Blood-pressure variability predicts stroke independently of mean (PMID 20226988) | Drug classes differ systematically in the variability they produce (PMID 21817143) | No trial has selected a drug class for its variability profile and measured events | OQ-14 |
| 7 | Adherence testing changes management in a third of apparently resistant patients (PMIDs 28777133, 28847892) | Renal denervation is offered to apparently resistant patients for a ~5 mm Hg gain (PMID 41870448) | No care pathway mandates chemical adherence verification before a device procedure | OQ-8, OQ-9 |
| 8 | Salt substitution reduces stroke and death (PMID 34459569) | Most sodium in high-income diets is added during manufacture, and population reformulation is a separate literature (PMID 27633834) | The two sodium-reduction strategies have never been compared, or combined, in the same population | OQ-7 |
| 9 | Tissue sodium storage detectable by ²³Na MRI (PMID 23339169) | Cohort studies using 24-hour or spot urinary sodium as the exposure (PMID 25119607) | Nobody has quantified how much of the observational sodium J-curve is explained by storage compartments the exposure measure ignores | OQ-13, OQ-22 |
| 10 | Hypertensive pregnancy roughly doubles later maternal stroke risk (PMID 36990309) | Postpartum blood-pressure self-management lowers pressure by 6.5/5.8 mm Hg at nine months (PMID 37950919) | No trial follows postpartum optimisation through to the stroke risk it is meant to modify | — |
| 11 | Aldosterone synthase inhibitors are entering phase 3 outcome trials paired with an SGLT2 inhibitor (NCT06742723) | Primary aldosteronism is present in a fifth of resistant hypertension (PMID 32449886) | The outcome trials do not stratify by aldosterone status, so they cannot say whether the benefit is mechanism-specific | OQ-1 |
| 12 | Treating asymptomatic inpatient hypertension is associated with harm (PMIDs 37252732, 33369614, 39585709) | Discharge intensification is common and also associated with harm (PMID 30209052) | The inpatient and the discharge decisions are studied separately although they are the same clinical moment | OQ-4 |
| 13 | Frailty concentrates antihypertensive harm (PMID 37075078) | Deprescribing is feasible over 12 weeks (PMID 32453368) and patients want to discuss it (PMID 31427342) | No trial randomises deprescribing in a frailty-defined population with outcomes beyond short-term pressure | OQ-5 |
| 14 | Blood pressure tracks strongly from adolescence and reversion becomes unlikely after it (PMID 39495520) | Young adults have the worst control rates of any age group — 5.6% at 18–39 (PMID 40156902) | The life-course evidence identifies a window; the care-delivery evidence identifies the failure; no intervention joins them | OQ-20 |
| 15 | GWAS explains >60% of SNP heritability and separates PRS deciles by 16.9 mm Hg (PMID 38689001) | Adding the PRS to a clinical model raises AUROC by 0.035 (PMID 38689001) | Nobody has explained why so much explained variance yields so little discrimination, or what a genetics-informed hypertension pathway would actually do differently | OQ-24 |
| 16 | Patients describe treatment burden as the principal cost of care (PMID 36840919) | No trial in this condition measures it (PMIDs 34010531, 39555827) | The endpoint that patients name is absent from every pivotal trial | OQ-27 |