REF-1318 · Vieira et al., 2022 · Reviews in Cardiovascular Medicine · Paper · PubMed · Free full text

Is "can I still talk comfortably?" a real intensity check?

In 10 small studies of 426 cardiovascular patients, comfortable speech marked work at or below the first ventilatory threshold. Repeatable, but not tested in athletes.

Main caveat: Population. Cardiac patients (more than 99%), mostly coronary, many on medication and in rehabilitation; two with COPD. Not healthy adults and not athletes.
moderate evidenceMeasured on non-athletes and recreational people

Vieira and colleagues (Santa Catarina State University, Brazil) reviewed how the talk test is applied, and how valid and reliable it is, in adults with chronic cardiovascular or pulmonary disease.

In the test, the person exercises through progressive stages and near the end of each recites a standard passage (about 30 words, such as the US Pledge of Allegiance or a Danish text) and answers whether they can speak comfortably: "yes" is a positive stage, "more or less" equivocal, "no" negative, which ends the test.

From five databases (1995-2022) and hand searches they included 10 studies: nine observational, one validation study, with 426 participants in total, more than 99% with cardiovascular disease and predominantly coronary artery disease. Most protocols copied the maximal exercise test: one-minute stages on a cycle ergometer, a standard paragraph and the comfort question.

Validity was studied in four studies: workload, oxygen uptake and heart rate at the last positive stage were not different from those at the first ventilatory threshold (VT1); at the last positive or first equivocal stage patients were at or below VT1, and at the negative stage above it.

A gas-exchange study of 20 patients found correlations between talk-test stages and VT1 of only 0.37 to 0.60. Reliability in four studies gave ICCs above 0.8, with a standard error of measurement of 6.6-11 W and minimal detectable changes of 18.3-32 W; agreement between two physiotherapists was poor in one study.

In one study the last comfortable-speech stage came before ECG signs of ischaemia in 84.2% of 19 patients. The authors conclude that the talk test is suitable, safe and feasible for assessing and prescribing exercise in cardiac patients, with the last positive and equivocal stages for training near VT1, but that protocols are not standardised.

Our reading: it widens the base's evidence for "speak in full sentences = easy" from one study in 16 healthy volunteers (REF-929) to ten small studies in heart patients, which supports the direction but not the precision, and not for athletes.

Protocol

Design: systematic review registered on PROSPERO and reported to PRISMA; searches of CINAHL, EMBASE, LILACS, PubMed and Scopus (1995-2022), plus reference lists, ResearchGate and Google Scholar; three blinded researchers screened, a senior researcher resolved disagreements. No meta-analysis (not planned; findings too dissimilar to pool).
Eligibility (PECO): adults ≥18 with chronic cardiovascular and/or pulmonary disease; talk test or similar speech provocation; comparison with a cardiopulmonary exercise test or another exercise test; randomised trials, cross-sectional studies or case series; no language restriction.
Included: 10 studies (USA 4, Denmark 5, Italy 1), 2002-2020; n = 426 participants across studies (Table 1 and Results; the abstract says 414); 10 to 85 patients per study (mean 42.6 ± 25.18).
Talk-test protocols: cycle ergometer 60%, treadmill 20%, choice of ergometer 10%, treadmill and indoor track 10%; stages of 1 min (5 studies), 2 min (3) or 3 min (1); speech provocation in the last 10-30 s of each stage, reciting a standard passage of about 30 words; question "Can you speak comfortably?" (yes / uncertain / no) in most.
Quality tools: adapted Newcastle-Ottawa Scale (low ≤5 stars, moderate 6-7, high ≥8); STROBE (22 items); COSMIN risk-of-bias boxes 3 and 8 (validity) and 6 (reliability).
Funding and interests: funded by CAPES (Higher Education Personnel Improvement Coordination, grant number 001) and by an institutional scientific-initiation scholarship of Santa Catarina State University (PROBIC/UDESC, PIC2018030000146); the authors declare no conflict of interest.

Key Results

PropertyStudies (patients)Result
Validity vs VT14 counted by the authorsWorkload, oxygen uptake and heart rate at the last positive stage not different from VT1; last positive / first equivocal stage at or below VT1; negative stage above VT1; VT2 not studied
Correlation with gas-exchange VT1Voelker 2002 (10)VO2 at VT1 vs talk-test stages r = 0.71 (last positive), 0.75 (equivocal), 0.83 (negative) (Table 2); VO2 and HR differed between VT1 and the negative stage, not the positive or equivocal
Correlation with gas-exchange VT1Sorensen 2020 (20)r = 0.37 to 0.60 across stages and variables (VO2 0.54-0.60, workload 0.45-0.51, HR 0.37-0.39)
Test-retest reliabilityKrawcyk 2017 (60, lacunar stroke)ICC 0.97; SEM 10.6 W (last positive) and 6.6 W (negative); MDC 29.4 and 18.3 W
Test-retest reliabilityNielsen 2014 (64)ICC 0.90 / 0.91 / 0.90 (positive / equivocal / negative); SEM 8.8-9.3 W; MDC 24.4-25.9 W
Inter-rater reliabilityPetersen 2014 (64)ICC 0.85, SEM 11 W, MDC 32 W; reliability between two or more therapists judged poor
Patient vs physiotherapist ratingZanettini 2012 (50)R = 0.81-0.85
ResponsivenessNielsen 2016 (85)30 W proposed as the minimal clinically important difference; 36% improved significantly in rehabilitation (surpassed two test stages)
IschaemiaCannon 2004 (19)last positive stage preceded the ischaemic threshold in 84.2%
Study quality10Newcastle-Ottawa high 3, moderate 4, low 3; COSMIN reliability adequate 1, doubtful 2, inadequate 1
Our arithmetic: the sample sizes in Table 1 (24, 19, 60, 30, 64, 85, 64, 20, 10, 50) sum to 426, matching the Results, not the abstract's 414.

Mechanisms

Measured: the stage at which speech stops being comfortable sat close to VT1 in these patients; once speech was not comfortable, intensity was above VT1.
Authors' interpretation: during exercise, rising gas exchange and a shorter expiratory time conflict with speech; around VT1, where ventilation starts to rise out of proportion to oxygen uptake, the conflict becomes insurmountable. Longer passages and longer stages may lower the intensity at which speech fails but reduce prediction error.
Inference: the test reads the ventilatory signal behind the moderate-heavy boundary, so it should transfer across populations in direction; the size of its error in trained athletes, whose ventilation at a given power differs, is not shown here.

Practical Application

Is "can I still talk comfortably?" a real intensity check? In 10 small studies of 426 cardiovascular patients, comfortable speech marked work at or below the first ventilatory threshold. Repeatable, but not tested in athletes.
Keep the instruction simple (B+, cardiac patients; B in healthy volunteers, [REF-929]): full, comfortable sentences = at or below the first threshold; speech that is "more or less" possible = at or just below it; cannot speak comfortably = above it. This supports the base's easy-day rule in SYNTH-75 and SYNTH-51 in direction.
Treat it as a boundary, not a zone grid: it separates below from above one threshold; it grades nothing within the easy range or above the threshold (REF-930: may not be practical for high-intensity intervals).
Same passage, same question (inference from the protocol findings): in patients, reliability was good when they judged their own speech (and for a single physiotherapist) and poor between different therapists; for an athlete a self-applied test with the same short passage is the consistent form (inference; not tested in athletes).
Long sessions are untested: these were incremental tests lasting minutes. Not from this paper; the base's own practice: late in a long ride or in heat, use breathing and talk as the cap rather than fresh watts (SYNTH-40, reasoned from REF-180).

Limitations

Population. Cardiac patients (more than 99%), mostly coronary, many on medication and in rehabilitation; two with COPD. Not healthy adults and not athletes; the athlete studies cited in its introduction were outside its scope.
Small studies, no pooling. 10 to 85 patients per study and no meta-analysis; the tabulated validity correlations come from two studies of 10 and 20 patients, one with weak correlations (0.37-0.60); "not different" in small samples is not agreement, and no limits of agreement are given.
Incremental tests only. No study tested the talk test during long steady exercise or at the second threshold.
Heterogeneous protocols. Ergometers, stage lengths (1-3 min), passages and wording differed; the authors stress there is no standard protocol.
Reporting. The abstract gives 414 participants, the Results and Table 1 426; the abstract's PROSPERO number (CRD420181068930) has one digit more than the Methods' (CRD42018106893); one reliability row reports ICC 0.97 with a 95% CI of 0.87 to 0.95, which does not contain the estimate; the authors count four validity studies, including one with no maximal test (VT1 inferred) and one (Zanettini, 50 patients) that the text lists for concurrent validity against VT1 but whose Table 2 row gives only patient-versus-physiotherapist correlations, while the 20-patient gas-analysis study is discussed separately.Item-level quality ratings are in a supplement not read here.
Funding. Brazilian public research funding (CAPES; UDESC scholarship); no conflicts declared.
How this entry was checked and changed
Why this level

read from the full text (Europe PMC, CC BY 4.0). A systematic review, but of 10 small observational or validation studies (10 to 85 patients each) with no pooled estimate, rated high quality in three, moderate in four and low in three; validity correlations against the first ventilatory threshold are tabulated for two studies of 10 and 20 patients, and reliability design was adequate (COSMIN) in one study of four

Full text (checked 2026-10-27)
Ten studies were included; four hundred and twenty-six individuals underwent the TT, more than 99% CVD patients; sample size varied from 10 to 85 patients (mean 42.6 ± 25.18) (full text)
Workload, VO2 or heart rate in the last positive stage was not different from the same parameters at VT1; at TT+ or TT± participants were on or below VT1, and at TT- surely above it (full text)
Sorensen et al. found correlations between the TT stages and VT1 from 0.37 to 0.60 (full text)
Reliability was evaluated in four studies with ICC >0.8; Petersen 2014: ICC = 0.85 (0.78 to 0.91), SEM = 11W, MDC = 32 W (full text)
Nielsen 2014: ICC 0.9, 0.91 and 0.90 for TT+, TT± and TT-; MDC 24.7, 24.4 and 25.9 W (full text)
TT+ preceded the ischemic threshold in 84.2% of subjects (Cannon 2004, n of 19) (full text)
A 30 W change in the test protocol is suggested as MCID (Nielsen 2016) (full text)
Funding: CAPES, grant number 001, and PROBIC/UDESC, PIC2018030000146; the authors declare no conflict of interest (full text)