The 6-minute walk test: How important is thelearning effect?Grace Wu,a Bonnie Sanderson, PhD, RN,b,c and Vera Bittner, MD, MSPHc Birmingham, Ala

Background The 6-minute walk test is a submaximal exercise test that is widely used as an outcomes measure inclinical trials and cardiopulmonary rehabilitation. An initial learning effect with repeated administration is well docu-mented, but it is unknown whether this learning effect persists with time.

Methods This study was designed to determine whether the learning effect persists after 2 months. Fifty healthyadults (mean age, 30.6 years; 48% men, 82% white) unfamiliar with the 6-minute walk test completed 3 walks at base-line (walks 1–3) and 3 walks at follow-up (walks 4–6). Height, weight, and self-reported physical activity were assessedat both points. Distances walked during the 6 walks were compared with a general linear model for repeated measureswith post-hoc pairwise comparisons corrected by the Bonferroni method. A P value �.05 was considered to be signifi-cant.

Results The distance walked increased significantly between walks 1 and 3 (2046 � 228 ft to 2194 � 266 ft, P�.05). There was no difference in distance walked between walks 3 and 4, which were conducted 2 months apart. Thedistance walked increased significantly between walks 4 and 6 (2201 � 233 ft to 2285 � 257 ft, P �.05). The overalllearning effect was inversely correlated with distance walked at walk 1, but was unrelated to age, sex, height, weight, orphysical activity level.

Conclusion The initial learning effect is maintained during a 2-month period. A more modest additional learningeffect occurs during the follow-up walks. Both learning effects should be taken into account when using the 6-minute walktest as an outcomes measure. (Am Heart J 2003;146:129-33.)

Exercise capacity has been used to measure the dis-ability resulting from and assess the efficacy of treat-ments for chronic cardiovascular and respiratory dis-eases. The gold standard of evaluation for exercisecapacity is maximal exercise testing on a treadmill orbicycle, but such testing is expensive, is often not wellaccepted by patients, and may not be a good measureof an individual’s ability to perform day-to-day activi-ties.1 A widely used submaximal exercise test is the6-minute walk test. Reference equations are availablefor different age and sex groups.2,3 The 6-minute walkdistance correlates with both maximal exercise capac-ity and subjective functional status questionnaires.1,4

The test is safe, easy to administer, well accepted bypatients, and does not require any expensive equip-ment. It is more reproducible than functional statusquestionnaires and has been found to be responsive inits ability to detect small, but clinically significantchanges in a patient’s exercise capacity.1,4 It predictsprognosis in patients with congestive heart failure andhas been used as an outcomes measure in a variety ofsettings, including research trials and cardiopulmonaryrehabilitation.4–9

Several studies have shown that walking distancetends to increase with repeated test administration (ie,a learning effect occurs because of test familiariza-tion).3,8,10,11 The magnitude of the reported learningeffect is quite variable from study to study and rangesfrom around 4.5% to 33% of the initial distancewalked.3,8,10,11 Because the distance walked tends toplateau after 3 walks, 1 to 2 practice walks have beensuggested before determining an individual’s exercisecapacity.1,10 It is unknown whether this initial learningeffect persists with time. Persistence of the learningeffect or lack thereof has important implications forresearch trial design and documentation of treatmentoutcomes in clinical care. Thus, the purpose of thisstudy is to determine whether the learning effect asso-

From the aUniversity of Alabama School of Medicine, bUniversity Hospital, and cDe-partment of Medicine, Division of Cardiovascular Disease, University of Alabama atBirmingham, Birmingham, Ala.Supported by a Summer Student Research Award (G.W.) funded through a grant fromthe University of Alabama Health Services Foundation to the University of Alabama atBirmingham General Clinical Research Center (RR00032).Submitted June 21, 2002; accepted Oct 30, 2002.Reprint requests: Vera Bittner, MD, MSPH, Professor of Medicine, University of Ala-bama at Birmingham, LHRB 310, 1530 3rd Ave S, Birmingham, AL 35294-0007.E-mail: vbittner@uab.edu© 2003, Mosby, Inc. All rights reserved.0002-8703/2003/$30.00 � 0doi:10.1016/S0002-8703(03)00119-4

ciated with repeated performance of the 6-minutewalk test persists during 2 months of follow-up.

MethodsThe study was approved by the University’s Institutional

Review Board. Informed consent was obtained from all re-search volunteers before participating in the study. Fiftyasymptomatic healthy volunteers who were �20 years oldwere recruited from the campus and community throughfliers and newspaper advertisements. Criteria for exclusionfrom the study included a history of diabetes mellitus requir-ing oral hypoglycemic medications or insulin, asthma orother pulmonary disorders, ischemic heart disease, valvularheart disease other than mitral valve prolapse, angina or exer-tional dyspnea, uncontrolled hypertension (�160/100 mmHg) or the use of antihypertensive medications, or orthope-dic or neurologic problems impairing mobility.

Sex, age, race, height, weight, and waist circumferencewere recorded for each participant. Smoking and medicalhistories and medication use were reviewed, and a brief car-diopulmonary examination was conducted to exclude previ-ously undiagnosed cardiovascular illness. Self-reported physi-cal activity at work and during leisure activities was assessedwith a questionnaire modeled after the Seven-Day PhysicalActivity Recall and expressed in MetHours.12 Participants re-turned 7 to 9 weeks after the baseline assessment at approxi-mately the same time of day. The same demographic andclinical data were obtained, and the physical activity ques-tionnaire was completed again.

6-Minute walk testEach participant performed six 6-minute walk tests, 3 at

baseline and 3 at follow-up. Consecutive walks were sepa-rated by 30-minute rest periods. The walks were completedon the 140-ft walking track in the CardioPulmonary Rehabili-tation facility. The track is marked at 10-ft intervals through-out its length so that the distance walked can be accuratelydetermined. Resting vital signs were recorded at baseline andbefore each walk to ensure that heart rate and blood pres-sure had returned to baseline levels. Participants were giventhese instructions: “The purpose of this test is to find outhow far you can walk in 6 minutes. You will start from thispoint (indicate marker at 1 end of track) and follow the pathuntil the 6-minute period is complete. If you need to stopduring this time, please do so, but remain where you are un-til you can go again. I will tell you the time every 2 minutes,and I will tell you to stop when the 6 minutes are up. WhenI say stop, please stand right where you are.” Because en-couragement has been shown to significantly impact the dis-tance walked, standardized encouragement was provided ev-ery 2 minutes, with phrases like “Keep up the good work” or“You’re doing great.”13 The time elapsed was measured witha stopwatch, and the distance walked was measured to thenearest foot. All walks were conducted by the same investiga-tor (G.W.) to minimize variability in test administration.

Data analysesBaseline and follow-up data were summarized by calculat-

ing means, SDs, and percentages as appropriate. Correlations

between walks were computed with Pearson correlation co-efficients. Coefficients of variation were calculated by com-puting the within-subject SD and expressing it as a percent-age of the mean distance walked. Determinants of walkperformance were assessed with univariate and multivariatelinear regression techniques using walked distance as the de-pendent variable. The learning effect on consecutive walkswas assessed with the general linear models procedure forrepeated measures. Post-hoc comparisons between individualwalks were corrected for multiple comparisons with the Bon-ferroni method. Determinants of the learning effect were as-sessed with linear regression analyses. All analyses were per-formed with the SPSS Statistical Package software version10.0 (SPSS, Chicago, Ill).

ResultsCharacteristics of the study population (n � 50, 26

women, 24 men) at baseline and at follow-up areshown in Table I. The men were younger, taller, andheavier than the women and had a higher blood pres-sure. Activity levels were comparable in men andwomen and did not change between baseline and fol-low-up.

There were no complications during the walks, andnone of the participants experienced angina, dyspnea,or other cardiovascular symptoms. The correlationsbetween distances walked for walks 1 to 6 are shownin Table II. As expected, consecutive walks were morehighly correlated with each other (correlation coeffi-cients between 0.93 and 0.98, P �.001) than walksperformed further apart in time, but even walks 1 and6 were still significantly correlated (r � 0.7, P �.001).The overall coefficient of variation was 5%. It washigher for men than for women (6% vs 4%; P � .009)

Determinants of walk performance during the initialwalk were assessed with linear regression. The dis-tance walked tended to increase with increasingheight, tended to decrease with increasing weight,body mass index (BMI), and age, and was shorter inwomen than in men. In a multiple regression modelthat included age, sex, height, BMI and self-reportedphysical activity, only BMI was a statistically significantpredictor of distance walked (�25 feet for every kg/m2; 95% CI, �44–�7; P �.05). The model explained19% of the variability in the distance walked.

Figure 1 shows the means and SDs of the distancewalked for all walks. The walking distance increasedsignificantly (148 ft, 7.25%) between walks 1 and 3,which confirmed the previously reported learning ef-fect. The third walk was the longest for 84% of partici-pants; 12% had their longest walk at walk 2, and 4%walked their longest distance at the initial walk. Thedistance walked was comparable for walks 3 and 4,which were conducted approximately 2 months apart(mean change, 7 ft [0.32%]; P � not significant), sug-gesting that the initial learning effect was maintained

American Heart JournalJuly 2003

130 Wu, Sanderson, and Bittner

during this period. The walking distance increased sig-nificantly between walks 4 and 6, with an average in-crease of 84 ft (3.82%; ie, there was an additional, al-beit more modest, learning effect when the 6-minutewalk test was re-administered repeatedly after 2months). Among the 3 follow-up walks, 76% of partici-pants reached their maximum distance at walk 6, 14%of participants reached their maximum distance atwalk 5, and 10% of participants reached their maxi-mum distance at walk 4. Overall, the mean distancewalked increased 239 ft (11.68%) between walks 1and 6. With walk 1 as the reference, walking distancesincreased 4.07%, 3.06%, 0.32%, 1.87%, and 1.92% witheach consecutive walk, respectively. There was a sig-nificant inverse relationship between the change indistance walked between walk 1 and walk 6 and the

distance walked at walk 1, which remained significantafter adjusting for age, sex, height, weight, activitylevel at baseline, and change in activity level betweenbaseline and follow-up (�0.27 ft for every foot in base-line distance; 95% CI, �0.51–�0.018; P �.05). Noneof the other variables in the model were significant.

DiscussionOur study makes 2 important contributions to the

6-minute walk literature: it is the first to show that thepreviously reported initial learning effect is maintainedby healthy subjects during a 2-month period, and itdemonstrates that an additional, yet more modest,learning effect occurs when healthy subjects perform

Table II. Correlations between walks

Walk 1 Walk 2 Walk 3 Walk 4 Walk 5 Walk 6

Walk 1 1.000 0.934* 0.879* 0.788* 0.739* 0.700*Walk 2 0.934* 1.000 0.951* 0.843* 0.824* 0.798*Walk 3 0.879* 0.951* 1.000 0.861* 0.875* 0.857*Walk 4 0.788* 0.843* 0.864* 1.000 0.959* 0.919*Walk 5 0.739* 0.824* 0.875* 0.959* 1.000 0.980*Walk 6 0.700* 0.798* 0.857* 0.919* 0.980* 1.000

Pearson correlation coefficients for walks 1–6.*All correlations are significant at P � .001.

Table I. Characteristics of study population

VariableWomen(n � 26)

Men(n � 24)

Total(n � 50)

P-valuewomen vs men

Baseline assessmentAge (y) 34.9 � 12.8 26.1 � 7.4 30.6 � 11.3 .005Height (in) 65.2 � 2.1 70.5 � 2.4 67.7 � 3.5 �.001Weight (lbs) 143.0 � 24.4 177.1 � 26.1 159.4 � 30.3 �.001Waist circumference (in) 31.4 � 3.6 35.2 � 3.3 33.3 � 3.9 �.001BMI (kg/m2) 23.8 � 3.9 25.1 � 3.6 24.4 � 3.8 .215Leisure METhours 12.4 � 9.3 18.2 � 13.1 15.2 � 11.5 .072Work METhours 12.1 � 29.8 5.7 � 9.3 9.0 � 22.4 .322Total METhours 24.4 � 30.6 23.9 � 15.3 24.2 � 24.3 .994Heart rate 67.5 � 7.4 62.3 � 9.0 65.0 � 8.6 .030Systolic blood pressure (mm Hg) 108.5 � 14.4 119.6 � 12.8 113.8 � 14.6 .006Diastolic blood pressure (mm Hg) 67.0 � 9.3 71.8 � 9.3 69.3 � 9.5 .076

Follow-up assessmentTime between baseline and

follow-up (days)50.5 � 6.6 49.5 � 7.1 50.0 � 6.8 .582

Weight (lbs) 143.1 � 23.9 177.5 � 26.1 159.6 � 30.2 �.001BMI (kg/m2) 23.8 � 3.8 25.1 � 3.5 24.4 � 3.7 .192Leisure METhours 12.4 � 10.9 12.5 � 10.8 12.5 � 10.7 .956Work METhours 6.5 � 9.2 2.5 � 7.4 4.6 � 8.6 .100Total METhours 18.8 � 14.9 15.0 � 12.2 17.0 � 13.4 .328Heart rate 68.5 � 7.1 66.8 � 10.2 67.6 � 8.7 .491Systolic blood pressure (mm Hg) 110.8 � 13.4 120.0 � 10.1 115.2 � 12.7 .009Diastolic blood pressure (mm Hg) 68.2 � 9.0 73.8 � 7.6 70.9 � 8.8 .022

BMI, Body mass index; in, inches; lbs, pounds.

American Heart JournalVolume 146, Number 1

Wu, Sanderson, and Bittner 131

multiple 6-minute walks at the 2-month follow-up as-sessment.

The distances walked by our participants were onaverage somewhat lower than those predicted by theequations of Enright et al, but somewhat higher thanthose predicted by Gibbons et al, who studied a groupof individuals who were similar in age to our popula-tion and modeled the “best distance walked.”2,3 Otherresearchers have observed strong relationships be-tween the distance walked and age, sex, height, andweight.2,3,8 In our population, only BMI was an inde-pendent predictor of walked distance in multivariateanalysis. The overall learning effect observed in ourstudy was of similar magnitude as that reported byother researchers.3,8,10 The only significant predictorof the learning effect in our study was the distancewalked during the first 6-minute walk. The inverse re-lationship between baseline distance and change indistance with successive test administration suggeststhat individuals with the longest baseline distance en-counter a “ceiling effect,” beyond which the distancewalked cannot be increased without changing into ajog/run. An alternate explanation is that the learningeffect is modified by regression to the mean.

We believe that our results have significant implica-tions for the design of research trials that use pharma-ceutical interventions and submaximal exercise capac-ity as a primary or secondary outcome. Our datasuggest that 3 walks should be performed at the begin-ning of the research study, as previously suggested byother researchers, and that 3 additional walks shouldbe performed at the end of the intervention period.The treatment effect should be computed by compar-

ing the longest distance of each set of walks (whichtends to be the third walk in each set of walks in mostindividuals). The variation of walk distance observed inour study will also help to determine sample size esti-mates when using the 6-minute walk as an efficacyvariable in clinical trials. The applicability of our re-sults to the cardiac rehabilitation setting or to formalexercise training studies in which the interventionconsists of walking either on the treadmill or on awalking track is less clear. In these settings, it may besufficient to perform 1 baseline walk and interpret anyimprovement with exercise training in light of theoverall modest learning effect observed in most stud-ies, including ours.

Study limitationsIt is not known whether the findings obtained in this

small group of young, healthy volunteers can be di-rectly extrapolated to patients commonly enrolled inclinical trials or cardiac rehabilitation programs, whotend to be older and to have multiple comorbiditiesand lower exercise capacities. Our volunteers per-formed the walk tests at baseline and after 2 monthsof follow-up. It is unknown whether the learning ef-fect would persist during longer periods and howmuch of an additional learning effect would be ob-served at more remote points.

ConclusionWe conclude that there is a learning effect when the

6-minute walk is administered repeatedly to individualswho are “walk-naive,” that the initial learning effect ismaintained during a 2-month period, and that an addi-tional, albeit more modest, learning effect occurs withrepeated test administration at follow-up. Both learningeffects should be taken into account when using the6-minute walk as an outcomes measure.

References1. Steele B. Timed walking tests of exercise capacity in chronic car-

diopulmonary illness. J Cardiopulm Rehabil 1996;16:25–33.2. Enright PL, Sherrell DL. Reference equations for the 6-minute walk

in healthy adults. Am J Respir Crit Care Med 1998;158:1384–7.3. Gibbons WJ, Fruchter N, Sloan S, et al. Reference values for a

multiple repetition 6-minute walk test in healthy adults older than20 years. J Cardiopulm Rehabil 2001;21:87–93.

4. Bittner V. Six-minute walk test in patients with cardiac dysfunction.Cardiologia 1997;42:897–902.

5. Bittner V, Weiner DH, Yusuf S, et al. for the SOLVD Investigators.Prediction of mortality and morbidity with a 6-minute walk test inpatients with left ventricular dysfunction. JAMA 1993;270:1702–7.

6. Zugck C, Krueger C, Duerr S, et al. Is the six minute walk test areliable substitute for peak oxygen uptake in patients with dilatedcardiomyopathy? Eur Heart J 2000;21:540–9.

7. Guyatt GH. Use of the 6-minute walk as an outcome measure inclinical trials in chronic heart failure. Heart Failure 1987:October-November:211–7.

Figure 1

Means and SDs of distances walked are shown for walks 1 to 3(baseline) and walks 4 to 6 (after 2 months of follow-up). The dis-tance walked increases significantly from walk 1 to walk 3, doesnot change between walk 3 and walk 4, which were performed 2months apart, and increases significantly between walks 4 and 6.

American Heart JournalJuly 2003

132 Wu, Sanderson, and Bittner

8. Hamilton DM, Haennel RG. Validity and reliability of the 6-minutewalk test in a cardiac rehabilitation population. J Cardiopulm Re-habil 2000;20:156–64.

9. Bittner V, Sanderson B, Breland J, et al. Assessing functional ca-pacity as an outcome in cardiac rehabilitation: role of the 6-minutewalk test. Clin Exerc Physiol 2000;2:19–26.

10. Guyatt GH, Sullivan MJ, Thompson PJ, et al. The 6-minute walk: anew measure of exercise capacity in patients with chronic heartfailure. Can Med Assoc J 1985;132:919–22.

11. Knox AJ, Morrison JFJ, Myers MF. Reproducibility of walking testresults in chronic obstructive airways disease. Thorax 1988;43:388–92.

12. Sallis JF, Haskell W, Wood PD, et al. Physical activity assessmentmethodology in the Five-City Project. Am J Epidemiol 1995;121:91–106.

13. Guyatt GH, Pugsley SO, Sullivan MJ, et al. Effect of encour-agement on walking test performance. Thorax 1984;39: 818–22.

The following article is an AHJ Online Exclusive.Full text of this article is available at no charge at our website:

www.mosby.com/ahj.

Beneficial effects of angiotensin-converting enzyme inhibitor andnitrate association on left ventricular remodeling in patients withlarge acute myocardial infarction: The Delapril Remodeling afterAcute Myocardial Infarction (DRAMI) trialRoberto Latini, MD,a Lidia Staszewsky, MD,a Aldo P. Maggioni, MD,a,b Paolo Marino, MD,c Francisco Hernandez-Bernal, MD,a,b

Gianni Tognoni, MD,a Violeta Labarta, MSc,a Silvana Gramenzi, MD,d Federico Bianchi, MD,e Giuseppe Sarcina, MD,f

Giovanni Cremonesi, MD,g Gian Luigi Nicolosi, MD,h and Enrico Geraci, MD,i on the behalf of the Delapril Remodeling afterAcute Myocardial Infarction (DRAMI) Collaborative Group Milano, Italy

Background In the large-scale trial, Gruppo Italiano per lo Studiodella Sopravvivenza nell’Infarto Miocardico-3 (GISSI-3), patients receivingthe combination of lisinopril and glyceryl trinitrate benefited most from ex-perimental therapy. Therefore, a multicenter, randomized, double-blindstudy, Delapril Remodeling After Acute Myocardial Infarction (DRAMI),was designed to assess (1) the possible additive beneficial effect on leftventricular remodeling of nitrates when combined with an angiotensin-converting enzyme inhibitor (ACEI), and (2) the tolerability of a new ACEI,delapril, in respect to lisinopril in patients with large myocardial infarction(MI).

Methods A total of 177 patients were randomized to receivedelapril plus isosorbide-5-mononitrate (IS5MN) placebo, delapril plusIS5MN, lisinopril plus IS5MN placebo, or lisinopril plus IS5MN start-ing within the first 36 hours after the onset of symptoms and continuingfor 3 months.

Results More than 80% of the patients showed extensive ST-segmentchanges and 36.7% had signs or symptoms of heart failure during the first36 hours. Over 3 months, IS5MN reduced, by 76%, the increase inLVEDV (17.4 � 5.0 mL placebo vs 4.2 � 4.4 mL IS5MN, P � .0439),reversed the increase in LVESV (7.5 � 3.9 mL placebo vs �5.5 � 2.9 mLIS5MN, P � .0052), and increased the recovery of LVEF (1.9% � 1.3%placebo vs 6.7% � 1.2% IS5MN, P � .0119). Overall, 3-month mortalitywas 10.2%; the most frequent clinical events were new episodes of severeheart failure (18.1%), persistent hypotension (10.7%), and post-MI angina(18.1%), with no differences between treatment groups.

Conclusions Administration for 3 months of IS5MN combined withan ACEI, both started within 36 hours from the onset of symptoms, wassafe and effective in reducing LV dilation and dysfunction after MI. The 2ACEIs, delapril and lisinopril, appeared to be equally well tolerated. (AmHeart J 2003;146:e2.)

American Heart JournalVolume 146, Number 1

Wu, Sanderson, and Bittner 133