REF-1668 · Lin et al., 2026 · Journal of Clinical Hypertension · Paper · PubMed

Isometric handgrip training was followed by resting blood pressure about 5.4/2.7 mmHg lower than before training (SBP −5.38, 95% CI −6.91 to −3.85; DBP −2.71, −3.75 to −1.66; I² 86% and 90%) across 31 trials with 905 normotensive and hypertensive adults

A pre-post change in the training groups, not a comparison with controls; across studies, protocols under 30% of maximal voluntary contraction and those run on more than 3 days a week went with smaller falls

Main caveat: The authors': no adherence data; high heterogeneity; the systolic model explained little of the variance, so factors outside dose (medication, adherence…
moderate evidenceMeasured on non-athletes and recreational peopleRead in full

Lin and colleagues pooled randomised studies that tested isometric handgrip training alone for at least 2 weeks in adults with normal, high-normal or raised blood pressure, and asked which dose features go with larger falls in resting pressure. They searched three databases to June 2025 and analysed 31 studies (38 handgrip groups, 905 participants, 51% male where reported).

Unusually for a review of RCTs, the comparator was each group's own pre-training value, not the control arm, so the pooled numbers are before-after changes within the trained groups. Systolic pressure was 5.38 mmHg lower and diastolic 2.71 mmHg lower after training, with considerable heterogeneity; normotensive groups fell 4.81/2.45 and hypertensive groups 5.73/2.95 mmHg.

In meta-regression across studies, higher intensity (% of maximal voluntary isometric contraction, MVIC) went with larger falls in both pressures, more training days a week with a smaller systolic fall, more sets with smaller falls in both, and longer contractions with a larger diastolic fall.

The authors propose four 2-min unilateral contractions at 30% MVIC or more, no more than 3 times a week, for at least 8 weeks, and say this should be read as a pattern across heterogeneous studies, not a clinical guideline.

Participants

31 studies, 905 participants in handgrip groups; 51% male (two studies did not report sex); reported age limits from 20.2 to 69.3 years; mostly middle-aged and overweight by BMI.
13 normotensive groups (235 people, 18.1 ± 17.0 per group) and 25 hypertensive or prehypertensive groups (670 people, 26.8 ± 36.8 per group). Of the hypertensive groups, 14 were on antihypertensive medication, 4 unmedicated and 2 did not report it; one prehypertensive trial alone had 200 participants.
Excluded: pregnancy, stroke, heart failure, chronic kidney disease; studies combining handgrip with other exercise or therapy; non-English papers.

Protocol

Search: EMBASE, MEDLINE/PubMed, Cochrane CENTRAL, inception to June 2025, plus citation searching; 34 studies in the review, 31 with usable values.
Intervention: handgrip only, at least 2 weeks. Typical dose across groups: 4.1 ± 0.8 sets of 2 ± 0.5 min at 27.4 ± 8.7% MVIC, 2.3 ± 1.9 min rest, 3.3 ± 1.2 days a week for 9 ± 3.2 weeks. Intensities in the study table ranged from 5% to 50% MVIC; most groups trained at 30%, and only one group above 40% (4 × 45 s at 50%).
Comparison: pre-intervention values of the same group (controls not used).
Quality: PEDro (mean 5.9 ± 1.0); RoB-2 shown only as a figure, with no counts in the text; GRADE rules stated in the methods, the result given only as "moderate evidence".
Analysis: RevMan random-effects mean differences; sensitivity analyses restricted to low or some-concerns risk of bias (reported as consistent, figures only); meta-regression (R, REML) of SMD on sets, minutes per set, %MVIC, rest interval, laterality, days per week and weeks.

Key Results

Outcome (after vs before handgrip training)Mean change, mmHg (95% CI)Heterogeneity
Systolic, all−5.38 (−6.91 to −3.85), p < 0.00001I² 86%
Systolic, normotensive groups−4.81I² 86%
Systolic, hypertensive groups−5.73I² 85%
Diastolic, all (30 studies)−2.71 (−3.75 to −1.66), p < 0.00001I² 90%
Diastolic, normotensive groups−2.45I² 77%
Diastolic, hypertensive groups−2.95I² 83%
Meta-regression (dependent: SMD of the change)Systolic β (95% CI), pDiastolic β (95% CI), p
Sets (log)1.515 (0.118 to 2.911), 0.034 — more sets, smaller fall3.278 (1.695 to 4.862), < 0.001 — more sets, smaller fall
Minutes per set−0.520 (−1.119 to 0.078), 0.088 — not significant−0.863 (−1.565 to −0.171), 0.016 — longer, larger fall
%MVIC−0.029 (−0.056 to −0.008), 0.043 — higher, larger fall−0.034 (−0.065 to −0.003), 0.032 — higher, larger fall
Rest interval (U-shaped transform)−0.026 (−0.057 to 0.005), 0.099 — not significant−0.017 (−0.051 to 0.017), 0.322 — not significant
Days per week (log)0.727 (0.062 to 1.392), 0.032 — more days, smaller fall0.016 (−0.722 to 0.753), 0.967 — not significant
Model R²17.65% (table); 10.17% in the results text39.91% (table); 39.66% in the results text
No subgroup test of normotensive against hypertensive groups is reported; "greater effects in hypertensive groups" rests on the two point estimates (−5.73 vs −4.81 systolic).
Only five included studies used a number of sets other than four (the authors), so the sets finding is thin.

Practical Application

Handgrip holds at about 30% of maximum, four 2-min holds up to three times a week for 8 weeks or more, went with resting pressure about 5/3 mmHg lower than before training in mostly middle-aged, overweight adults; that figure is a before-after change in the trained groups, not a difference from people who did not train, so the true training effect may be smaller.
This agrees with the handgrip dose [SYNTH-88] already uses (about 30% MVIC, 4 × 2 min, 3 times a week, kept up for weeks): the review adds that protocols below 30% went with smaller falls and that more than 3 days a week went with a smaller systolic fall (no association for diastolic). Intensities well above 30% were tested in one group only, so "at least 30%" here means about 30%, not "harder is better".
Do not quote −5 mmHg to a normotensive athlete: normotensive groups were mostly young adults who were not athletes and fell 4.81/2.45 mmHg from their own baseline; no group of trained athletes was studied.
Inference, not tested here: for a masters athlete with high-normal or raised pressure, handgrip holds are a low-time add-on alongside aerobic training; whether they add anything to the pressure of an already well-trained person is unknown.
This is not treatment advice: a person with raised blood pressure or on blood-pressure medication sets any exercise programme for it with their physician, and keeps the aerobic training and medical care it would add to.

Limitations

The authors': no adherence data; high heterogeneity; the systolic model explained little of the variance, so factors outside dose (medication, adherence, baseline pressure, diet, genetics) likely drive much of the between-study difference; medication effects not fully accounted for.
Design: the pooled effect is each trained group against its own baseline, although the included RCTs had control arms; regression to the mean, repeated-measurement familiarity and seasonal change are not subtracted, so these numbers are not comparable with the control-referenced −6.97/−3.86 (REF-1151) or −7.10/−3.46 (REF-1150).
Quality reporting: RoB-2 results appear only in a figure and GRADE results are summarised as "moderate evidence" without a table in the text; no study blinded participants or therapists (PEDro table).
Things that do not add up: the results text gives model R² of 10.17% and 39.66%, the table and discussion 17.65% and 39.91%; the results text calls longer contractions a predictor of a larger systolic fall, but its coefficient is not significant (p = 0.088); the table footnote marks p < 0.005 for values of 0.032-0.043; laterality counts are given out of 38 and out of 39 groups; the proposed protocol is "unilateral" while half the groups alternated hands.
Dose findings are between-study associations from a continuous regression; there is no tested threshold at 30% and no direct comparison of frequencies beyond one 2 vs 4 days a week trial in normotensive adults.
No conflicts of interest declared; no funding statement in the text.
How this entry was checked and changed
Why this level

risk of bias read (PEDro 5.9 ± 1.0, range 3-8, fair to good; no study blinded participants or therapists), but the pooled effect compares the trained groups with their own baseline rather than with the control arms the RCTs had, so it carries the weaknesses of an uncontrolled pre-post design, and heterogeneity is considerable (I² 86-90%)