Neuropharmacology 39 (2000) 835–841

www.elsevier.com/locate/neuropharm

Defining post-stroke recovery: implications for design and

interpretation of drug trials
Pamela W. Duncan *, Sue Min Lai, John Keighley
Center on Aging, Department of Health Policy and Management, Department of Preventive Medicine, University of Kansas Medical Center,
Kansas City, Kansas and the Department of Veterans Affairs Medical Center, Kansas City, MO, USA

Accepted 4 January 2000

Abstract

Measurement of stroke recovery is complex because definition of successful recovery is highly variable across measures and cut-
off points for defining successful outcomes vary. The purpose of this paper is to describe patterns of recovery in stroke patients
of varying severity when different measures are used and when different cut-off points are selected. 459 individuals enrolled in a
prospective cohort study were assessed within 14 days post stroke and re-evaluated at 1, 3, and 6 months. Recovery was assessed
using the NIH Stroke Scale, the Fugl-Meyer Assessment of Motor Recovery, the Barthel Index of Activities of Daily Living, the
Physical Function Index of the SF-36, and the Modified Rankin Outcome Scale. Subjects also defined their preference (utility) for
their current health state with a time-trade off question. We compared patterns of recovery using the different measures and varying
the cut-off points for defining successful recovery. The percentage of patients who are believed to have recovered depends on how
recovery is defined. If recovery is defined at the disability level (Barthel⬎90), the majority 57.3% of stroke survivors experience
a full recovery. Fewer individuals are considered to be fully recovered if impairments are measured (NIHⱕ1, 44.9% and Fugl-
Meyer⬎90, 36.8%. Less than 25% of stroke survivors are considered recovered if recovery is defined relative to reported prior
function in higher levels of physical activity. Shifting the definition of recovery on the modified Rankin scale from ⱕ1 to ⱕ2 shifts
the percentage of those deemed recovered from ⱕ25% to 53.8%. In designing drug trials the methods for defining stroke recovery
should be carefully considered. If recovery is defined in terms of disability, a higher proportion of the placebo group will achieve
the outcome than if impairments are used to define recovery. The benchmarks for recovery in minor strokes must include measures
of higher functioning (e.g. the SF-36 physical functioning index or a Rankin 0 (no symptoms).  2000 Published by Elsevier
Science Ltd. All rights reserved.

Keywords: Stroke; Outcomes; Recovery

1. Introduction symptoms of stroke probably reflect some resolution of

Most patients who survive stroke experience some
degree of recovery. Yet, the best methods for charac-
terizing recovery following stroke are elusive. There is
lack of precise definitions of recovery. Recovery may be
measured by assessing impairments, disabilities or qual-
ity of life. Impairments are the neurological signs and
symptoms of stroke, disabilities are the functional conse-
quences of the neurological deficits, and handicap or
quality of life reflects limitations in role performance
(WHO, 1980). Recovery of the neurological signs and
* Corresponding author.
E-mail address: pduncan@kumc.edu (P.W. Duncan).
the pathological sequela of stroke and neuroplasticity.
Changes in the functional consequences (physical and
psychosocial) may be associated with improvements in
neurological function, but they may also reflect
behavioral compensation. Finally, reintegration into
prior physical and social roles is often modified by fac-
tors other than the degrees of neurological recovery
(Duncan, 1999).
In a review of 174 stroke trials, recovery was most
often assessed by improvements in neurological impair-
ments and assessment of quality of life was rare (Roberts
and Counsell, 1998). The methods for measuring recov-
ery at any level were highly variable and most measures
did not have established reliability or validity. Even if
there was some degree of consistency across trials in the

0028-3908/00/$ - see front matter  2000 Published by Elsevier Science Ltd. All rights reserved.
PII: S 0 0 2 8 - 3 9 0 8 ( 0 0 ) 0 0 0 0 3 - 4
836 P.W. Duncan et al. / Neuropharmacology 39 (2000) 835–841
methods of measuring stroke recovery, there is disagree-
ment on the extent to which deficits must be reduced in
order to claim recovery. In a recent review of phase three
drug trials, Sulter and colleagues reported that the Bar-
thel Activities of Daily Living Index and the Rankin
Stroke Outcome scale were the most commonly used
measures of recovery (Sulter et al., 1999). Yet, the cut-
off scores of these measures selected for defining recov-
ery were highly variable. Changing cut-off scores for
definitions of recovery may change the results of clinical
studies of therapeutic interventions. For example in
analysis of the ECASS II trial of r-TPA for acute stroke,
the cut-off scores for defining recovery on the Rankin
was shifted from Rankin score of ⱕ1 to ⱕ2. In the first
analysis the results of r-TPA trial were not statistically
significant but they were statistically significant in the
second analysis (Hacke et al., 1998).
In summary, measurement of stroke recovery is com-
plex because definition of successful recovery is highly
variable across measures and cut-off points for the meas-
ures are varied. Until we understand patterns of recov-
ery, the influence that our selection of levels of measure-
ment has, and the cut-off scores selected to deem
someone recovered, it will continue to be difficult to
understand theories of neuroplasticity or evaluate if vari-
ous therapeutic interventions work. The purpose of this
paper is to describe patterns of recovery in stroke
patients of varying severity when different measures are
used and when different cut-off scores for the measures
are selected.
2. Methods
2.1. Subjects
Subjects for this study include 459 individuals
enrolled in the Kansas City Stroke Study. The Kansas
City Stroke Study is a prospective cohort study designed
to characterize the patterns of recovery of stroke
patients. As described by Lai et al., individuals with
stroke were assessed with a battery of outcome measures
within 14 days after stroke and a follow-up was perfor-
med at 1, 3, and 6 months (Lai et al., 1998). To be
included in this study subjects had to have a confirmed
eligible stroke as defined by the World Health Organiza-
tion criteria as “rapid onset and of vascular origin
reflecting a focal disturbance of cerebral function,
excluding isolated impairments of higher function and
persisting longer than 24 h” (WHO, 1993). The stroke
was confirmed by clinical assessment and/or CT or MRI
scan. Subjects were included if they were ⱖ18 years of
age, had stroke onset within 14 days of baseline assess-
ment, did not have a stroke due to subarachnoid
hemorrhage; did not have renal, hepatic, or heart failure,
did not have a condition in which they were expected
to die within 6 months, were living the in the community
prior to stroke and were independent in activities of daily
living, and were not lethargic, obtunded or comatose
post stroke, and lived within 70 miles from the partici-
pating hospitals.
2.2. Measures of recovery
For the purpose of this study we selected the follow-
ing measures of stroke recovery: The NIH Stroke Scale
(Brott et al., 1989), the Fugl-Meyer Assessment of
Motor Recovery (Fugl-Meyer et al., 1975), the Barthel
Index of Activities of Daily Living (Mahoney and Bar-
thel, 1965), the Physical Function Index (PFI) of the SF-
36 (Ware and Sherbourne, 1992), and the Modified Ran-
kin Stroke Outcome Scale (Van Swieten et al., 1988).
Each of the measures has established reliability and val-
idity. The NIH, a measure of stroke related impairments,
consists of 14 items that assess the severity of impair-
ments (loss of consciousness, ability to respond to ques-
tions and to obey simple commands, deviation of gaze,
hemianopsia, facial palsy, limb movements of paretic
limb, limb ataxia, sensory loss, neglect, dysarthria, and
aphasia). Each item is rated on a 3–4 point ordinal scale,
with lower scores reflecting less impairment. The Fugl-
Meyer is a measure of motor impairments. Upper and
lower extremity movements and reflexes are rated on a
3 point ordinal scale. The maximum score for the entire
motor scale is 100 but may be divided for upper
extremity motor function (66) and lower extremity motor
function (34). The Barthel Index is a reliable disability
scale. It includes 10 items that can be divided into self
care (feeding grooming, bathing, dressing, bowel and
bladder care, and toileting) and mobility (ambulation,
transfers and stair climbing). The max score is 100 and
lowest score is 0, patient is bedridden. The SF-36 physi-
cal function index includes 10 question about limitations
in physical activities (vigorous activities such as running,
lifting heavy objects, moderate activities such as moving
a table, pushing a vacuum, lifting or carrying groceries,
climbing several flights of stairs, climbing one flight of
stairs, bending, kneeling or stooping, walking more than
a mile, walking several blocks, walking one block, bath-
ing or dressing yourself). The maximum score is 100.
Scores on the physical function index of the SF-36 have
been normalized for age and gender. The average score
for 70 year old women on the SF-36 PFI is 66 and for
70 year old men the average score is 75 (Ware, 1993).
The modified Rankin Scale is a global outcomes rating
scale for stroke patients. The scale includes 6 grades,
from 0 to 5, with 0 corresponding to no symptoms and
5 to severe disability, bedridden (Table 1).
Scores to define stroke recovery on the various meas-
ures are highly variable (Sulter et al., 1999; Duncan et
al., manuscript in preparation). For the purposes of this
paper we examined patterns of recovery using the fol-
 P.W. Duncan et al. / Neuropharmacology 39 (2000) 835–841
Table 1
Modified Rankin scale
Classification Description
0 No symptoms
1 No significant disability, despite symptoms; able
to perform all usual duties and activities
2 Slight disability; unable to perform all previous
activities but able to look after own affairs
without assistance
3 Moderate disability; requires some help but able
to walk without assistance
4 Moderately severe disability; unable to walk
without assistance and unable to attend to own
bodily needs without assistance
5 Severe disability; bedridden, incontinent, and
requires constant nursing care and attention
lowing scores: NIHⱕ1, Fugl-Meyer⬎90, Barthel ADL
Recovery⬎90, SF-36⬎66 for women and ⬎75 for men,
and Rankinⱕ1 and Rankinⱕ2.
In this study, patients were asked to define their pref-
erence (utility) for their current health state with a “time-
trade off” question (Wolfson et al., 1982). Time trade
off is a method that directly assesses how much time in
perfect health is equivalent to a period in their current
state of health. Subjects were asked if they preferred 10
years of life in their current state or “X” years in excel-
lent health. If the patients preferred the shorter period in
perfect health, the question was repeated with shorter
periods until the subjects became indifferent between the
alternative, i.e. no longer willing to trade more time for
excellent health. If we assigned a utility of 9 that meant
the patient would be willing to trade off 1 year out of
10 in their current state of health for perfect health.
2.3. Classification of stroke severity
Patients with different levels of stroke severity have
different probabilities of recovery (Jorgensen et al.,
1995). In order to describe patterns of recovery we strati-
fied patients by severity. Stroke severity was determined
by administration within 3–14 days after stroke of the
Orpington Prognostic Scale, a weighted measure that
screens for motor deficits, sensory loss, balance, and
cognition (Kalra and Crome, 1993). The Orpington
Prognostic Scale ranges for stroke severity are as fol-
lows: <3.2, minor stroke; 3.2–5.2, moderate stroke; and
⬎5.2 major stroke.
2.4. Statistical analysis
Mixed model software (SAS—Mixed Procedure) was
used to analyze differences in recovery profile over time
by stroke severity. This procedure took into account
repeated measures over time in each subject and missing
values in the data.
837
3. Results
459 patients with mean age 70±11.4 years, 53.4%
females, 80% white were enrolled in the Kansas City
Stroke Study. Of the 459 strokes, 93.7% were ischemic
strokes and 6.3% were hemorrhagic strokes. As meas-
ured by the Orpington Prognostic Scale, 39% of the indi-
viduals had a minor stroke, 50% experienced a moderate
stroke and 11% experienced a major stroke. Prior to the
stroke all the subjects were independent in activities of
daily living (Barthel⬎90) and their self-reported assess-
ments of their higher levels of physical function 1 week
prior to their stroke as measured by the SF-36 physical
function questions were mean=63.6±29.4 for the females
and mean=76.1±24 for the males. These functional states
are similar to population based norms from men and
women of this age group (Ware, 1993).
All of the stroke patients were evaluated within 14
days post stroke, Mean time from stroke onset to base-
line assessment was 8.5±3.6 days. Across the 6 month
assessments 82 patients were loss to follow-up (32
expired, 12 moved out of the region, 37 refused to con-
tinue participation, and 1 was unable to be reached).
Almost all individuals with stroke experience recov-
ery of neurological function as measured by the NIH
stroke scale (Fig. 1) and recovery of motor function as
measured by the Fugl-Meyer Assessment (Figs. 2 and
3). The most dramatic recovery from neurological
impairments occurs in the first 30 days but in moderate
and severe strokes recovery may continue up to 3 and
6 months respectively. Recovery of functional abilities
as measured by the Barthel ADL index is similar to
recovery from impairments (Fig. 4). Yet, many individ-
uals who achieve independence in Barthel ADL (BI⬎90)
still have residual neurological deficits (Table 2).
Assessment of recovery of higher levels of physical
function as measured by question 3 of the SF-36 demon-
strates similar patterns of recovery as do the profiles of
changes in neurological function and activities of daily
living (Fig. 5). However, even the mildest stroke patients
do not achieve their prior levels of physical function.
The recovery profiles (Figs. 1–5) depended on time and
stroke severity (i.e. time and severity interaction). The
F statistics associated with the time and severity interac-
tions and corresponding p values for the NIH, Fugl-
Meyer upper and lower extremity motor functions, Bar-
thel Index, and SF-36 physical functioning, respectively
were 13.1 (P=0.0001), 13.9 (P=0.0001), 14.6
(P=0.0001), 24.5 (P=0.0001), and 3.7 (P=0.006).
The percentage of patients who are believed to have
recovered depends on how recovery is defined (Table 3).
If recovery is defined at the disability level (Barthel⬎90)
the majority (57.3%) of stroke survivors experience a
full recovery but the severity of stroke effects the pro-
portion of patients who achieve a Barthel⬎90 (Table 4).
Fewer individuals are considered to be fully recovered
838 P.W. Duncan et al. / Neuropharmacology 39 (2000) 835–841

Fig. 1. Plot of means and 1 SD of NIH.

Fig. 2. Plot of means and 1 SD of Fugl-Meyer upper extremity after stroke.

if impairments are measured (NIHⱕ1, 44.9% and Fugl-
Meyer⬎90, 36.8%) and <25% are considered recovered
if recovery is defined relative to reported prior function
in higher levels of physical activity.
Shifting the definition of recovery on the modified
Rankin scale from ⱕ1 to ⱕ2 shifts the percentage of
those deemed recovered from <25% to 53.8%. Compari-
son of scores on various measures across six month Ran-
kin Scores, demonstrates that the Rankin Scores capture
different levels of function, especially in higher physical
function and the patients assign very different prefer-
ences for health states (utilities) for Rankin categories
(Table 5). For example, those individuals in Rankin
Classification of 0 would only trade off 5.4 months of
life for excellent health, but in Rankin classification of
1 and 2 the patients would trade off 1 year, 4 months
and 2 years, 8 months respectively.
4. Discussion
Using the Kansas City Stroke Study Cohort, we dem-
onstrated that the degree and time course of recovery
 P.W. Duncan et al. / Neuropharmacology 39 (2000) 835–841 839

Fig. 3. Plot of means and 1 SD of Fugl-Meyer lower extremity after stroke.

Fig. 4. Plot of means and 1 SD of Barthel ADL after stroke.

varies by stroke severity and number of patients deemed

Table 2 recovered varied as we changed measures of recovery
Range of residual deficits in individuals with Barthel⬎90 (n=217) and cut-off scores for the measures. Previously com-
pleted drug trials have not adequately considered pat-
Mean SD Minimum Maximum terns of stroke recovery in the heterogeneous group of
stroke patients, nor have they considered the impli-
NIH 1.63 1.71 0.0 9.0
Fugl-Meyer 86.30 15.03 22.0 100.0 cations for sample size when they varied the cut-off
PFI 62.60 23.22 5.0 100.0 scores to define recovery (Duncan et al., manuscript in
preparation). For example, if recovery is defined in terms
of disability, a higher proportion of the placebo groups
will achieve the outcome than if impairments are used
to define recovery. Subsequently, studies that select dis-
840 P.W. Duncan et al. / Neuropharmacology 39 (2000) 835–841

Fig. 5. Plot of means and 1 SD of physical functioning after stroke.

Table 3
Percent of cohort that achieved recovery

Women SF36- Men SF36-

Time NIHⱕ1 Fugl-Meyer⬎90 Barthel⬎90 Rankinⱕ1 Rankinⱕ2
PFI⬎66 PFI⬎75

Baseline 10.46 13.07 8.06 0.00 0.00 1.74 12.20

1 Month 16.78 12.85 26.14 9.44 12.08 5.45 21.57
3 Month 11.11 7.84 13.94 10.73 11.59 9.80 12.85
6 Month 6.54 3.05 9.15 3.86 4.83 7.41 7.19
Never 55.10 63.20 42.70 76.00 71.50 75.60 46.20

ability measures of recovery will require larger sample

Table 4 sizes. Investigators must also consider the severity of the
Percentage of patients achieving Barthel⬎90
stroke population they are enrolling in drug trials as they
Lost to select measures to define recovery. If the study enrolls
No Yes Died
follow-up minor stroke patients, within 3 months most all of the
patients will become independent in basic activities of
Major 49 16 27 8
daily living as a course of natural recovery while they
Moderate 35 52 4 9
Minor 6 89 1 4 continue to have residual neurological impairments and
limitations in higher levels of physical function. For
example, the benchmarks for recovery in minor and

Table 5
Six month outcomes by modified Rankin classifications

Rankin 0 1 2 3 4 5

NIH 0.07±0.27 0.78±1.12 2.06±1.63 3.93±2.55 7.94±4.80 18.71±11.74

Fugl-Meyer 99.31±1.49 91.76±9.51 82.94±17.18 62.58±24.04 39.15±25.42 16.69±23.89
Barthel 99.23±1.88 98.97±2.83 96.52±5.08 83.40±13.50 42.83±18.25 6.43±8.86
SF36-PFI 85.38±9.46 70.12±21.06 54.23±22.98 29.38±21.04 8.68±11.87 1.67±4.08
Time trade-off 9.27±1.47 8.77±2.11 7.30±3.12 6.86±3.15 5.43±3.41 3.36±4.13
 P.W. Duncan et al. / Neuropharmacology 39 (2000) 835–841
moderate strokes must include measures of higher physi-
cal functioning, e.g. the SF-36 physical functioning
index or a Rankin 0 (no symptoms). If severe stroke
patients are enrolled, assessment of recovery of basic
activities of daily living as measured by the Barthel ADL
Index may be an appropriate outcome measure that
should be assessed at 6 months post stroke.
Most previous studies have dichotomized outcomes as
favorable or unfavorable using arbitrary cut-off scores
on the selected measures. Dichotomizing scales to define
recovery or no recovery reduces outcome information
and may limit our ability to detect shifts in disability
states. As reflected in Table 5, the levels of impairments
and limitations in functions vary for each Rankin dis-
ability level and patients value the most severe states
less. If a pharmaceutical intervention shifts disability
from a Rankin 5 to a Rankin 3 (a clinically meaningful
improvement), the outcome is not considered favorable.
However, most patients perceived this is a better state
and would not be willing to disregard a treatment benefit.
The ultimate goal of every trial of pharmacological
intervention is to prove that the experimental therapy
improves stroke recovery. To date the results from pre-
vious drug trials have been unconvincing. The lack of
convincing results may be due to ineffective drugs but
they may also be due poor design that have not stratified
by stroke severity, adequately considered post-stroke
recovery and optimal definitions of recovery for the
given cohort.
Acknowledgements
This study was funded by the Department of Veterans
Affairs Rehabilitation Research and Development and
the University of Kansas Claude D Pepper Center,
funded by the National Institute of Medicine (P60 AG
14635-02). Participating facilities in the greater Kansas
City area included: Baptist Hospital, Department of Vet-
erans Affairs Medical Centers at Kansas City and Leav-
enworth, Liberty Hospital, Medical Center of Indepen-
dence, Mid-American Rehabilitation Hospital,
Rehabilitation Institute, Research Medical Center, St
841
Luke’s Hospital, St Joseph Health Center, Trinity
Lutheran Hospital, and University of Kansas Medical
Center.
References
Brott, T., Adams, H.P., Olinger, C.P. et al., 1989. Measurements of
acute cerebral infarction: a clinical examination scale. Stroke 20,
864–870.
Fugl-Meyer, A.R., Jaasko, L., Leyman, I. et al., 1975. The post-stroke
hemiplegic patient: a method for evaluation of physical perform-
ance. Scand. J. Rehab. Med. 7, 13–31.
Duncan, P.W., 1999. Measuring recovery of function after stroke. In:
Goldstein, L.B. (Ed.) Restorative Neurology. Futura, Armonk, NY,
pp. 224–240.
Hacke, W., Kaste, M. et al., 1998. Randomized double-blind placebo-
controlled trial of thrombolytic therapy with intravenous alteplase
in acute ischemic stroke (ECASS II). Second European–Australa-
sian Acute Stroke Study Investigators. Lancet 352, 1245–1251.
Jorgensen, H.S., Nakayma, H., Raaschou, H. et al., 1995. Outcome
and time course of recovery in stroke. Part 1: the Copenhagen
Stroke. Archs Phys. Med. Rehabil. 76, 399–405.
Kalra, L., Crome, P., 1993. The role of prognostic scores in targeting
stroke rehabilitation in the elderly. J. Am. Geriat. Soc. 41, 396–400.
Lai, S.M., Duncan, P.W., Keigley, J., 1998. Prediction of functional
outcome after stroke. Stroke 29, 1838–1842.
Mahoney, F.I., Barthel, D.W., 1965. Functional evaluation: the Barthel
Index. MD State Med. J. 14, 61–65.
Roberts, L., Counsell, C., 1998. Assessment of clinical outcomes in
acute stroke. Stroke 29, 986–991.
Sulter, G., Steen, C., De Keyser, J., 1999. Use of the Barthel index ad
modified Rankin scale in acute stroke trials. Stroke 30, 1538–1541.
Van Swieten, J.C., Koudstaal, P.F., Visser, M.C., Schouten, H.J.A.,
van Gigin, J., 1988. Interobserver agreement for the assessment of
handicap in stroke patients. Stroke 19, 604–607.
Ware, J.E., 1993. SF-36 Health Survey: Manual and Interpretation
Guide. The Health Institute, New England Medical Center, Bos-
ton, MA.
Ware, J.E., Sherbourne, C.E., 1992. MOS-36 item short form health
survey (SF-36): conceptual framework and item selection. Medical
Care 30, 473–483.
WHO, 1980. International Classification of Impairments, Disabilities,
and Handicaps. World Health Organization, Geneva.
WHO, 1993. World Health Organization. Proposal for the multi-
national monitoring and determinants in cerebrovascular disease.
Monica Project WHO/MNC82.1 REV 1.
Wolfson, A.D., Sinclair, A.J., Bombardier, C. et al., 1982. Preference
measurements for functional status in stroke patients: interrater and
intertechnique comparisons. In: Kane, R.L., Kane, R.A. (Eds.)
Values and Long-Term Care. Lexington Books, Lexington MA.