The Journal of Pain, Vol 16, No 10 (October), 2015: pp 1012-1021

Available online at www.jpain.org and www.sciencedirect.com

Validation of the Hospital Anxiety and Depression Scale in Patients

With Acute Low Back Pain

Dennis C. Turk,* Robert H. Dworkin,y Jeremiah J. Trudeau,z Carmela Benson,x

David M. Biondi,x Nathaniel P. Katz,z,{ and Myoung Kim{
*Department of Anesthesiology and Pain Medicine, University of Washington, Seattle, Washington.
y
Departments of Anesthesiology and Neurology and Center for Human Experimental Therapeutics, University of
Rochester, Rochester, New York.
z
Analgesic Solutions, Natick, Massachusetts.
x
Janssen Scientific Affairs, LLC, Raritan, New Jersey.
{
Department of Anesthesiology, Tufts University, Boston, Massachusetts.

Abstract: The Hospital Anxiety and Depression Scale (HADS) is a self-report instrument used to eval-
uate depression and anxiety in clinical research. The HADS has advantages over other assessments of
anxiety and depression; it is efficient in assessing both anxiety and depression with a total of 14
items, and it was originally developed on a general medical rather than psychiatric sample. However,
the HADS has not been evaluated specifically for use in clinical trials of acute pain. Validation ana -
lyses were conducted on data from a randomized, double-blind, parallel-group study of tapentadol
immediate release vs oxycodone immediate release for acute low back pain (N = 666). Analyses of
psychometric properties, internal consistency, convergent validity, assessments of bias, and confirma -
tory factor analysis were conducted on pretreatment data. Additional analyses were performed to
test the responsiveness and predictive validity of the HADS. Both the Anxiety and Depression sub -
scales 1) showed good psychometric properties, 2) had high internal consistency, 3) displayed
good convergent validity, 4) had no unexpected biases, 5) fit the a priori factor structure, and 6)
were highly sensitive to changes as a result of analgesic treatment. We conclude that the HADS is
a valid instrument for efficient, low-burden assessment of anxiety and depression in clinical trials
with an acute low back pain population.
Perspective: Considered together with the results of other recent studies, the data suggest that the
HADS can provide a valid, responsive, and efficient assessment of anxiety and depression in acute
low back pain for clinical trials and other clinical research examining acute pain populations.
2015 by the American Pain Society
Key words: Anxiety, depression, acute pain, low back pain, Hospital Anxiety and Depression Scale.

Received February 27, 2015; Revised June 24, 2015; Accepted July 1, 2015.
o determine the benefits of treatment, investiga-

Support for the clinical trial from which the data analyzed in this article
were derived was provided by Janssen Scientific Affairs, which also pro-
vided a research grant to Analgesic Solutions to support data analyses
and publication preparation.
D.C.T. has received in the past 12 months research grants from the U.S.
National Institutes of Health and the U.S. Food and Drug Administration
and consulting fees or honoraria in the past year from Lilly, Mallincrodt,
Orexo, Ortho-McNeil Janssen, Pfizer, and Xdynia. R.H.D. has received in
the past 12 months research grants from the U.S. Food and Drug Admin-
istration and U.S. National Institutes of Health and compensation for ac-
tivities involving clinical trial research methods from Acetylon, Astellas,
AstraZeneca, Avanir, Axsome, Biogen, Centrexion, Charleston, Chromo-
cell, Daiichi Sankyo, Eli Lilly, Hydra, Johnson & Johnson, Lpath, Maxwell
Biotech, Metys, Olatec, Phosphagenics, QRx, Relmada, Salix, Sorrento,
Spinifex, Teva, and Virobay. J.J.T. was a full-time employee of Analgesic
Solutions at the time the study was conducted and the first draft of the
manuscript was prepared. D.M.B., M.K., and C.B. are full-time employees
of Johnson & Johnson. N.P.K. is the owner of Analgesic Solutions.
Address reprint requests to Dennis C. Turk, PhD, Department of Anesthe-
siology & Pain Medicine, University of Washington, Box 356540, Seattle,
WA 98195. E-mail: turkdc@uw.edu
1526-5900/$36.00
2015 by the American Pain Society
http://dx.doi.org/10.1016/j.jpain.2015.07.001
T
tors must decide the appropriate end points for
establishing both the statistical significance and
the clinical importance of the treatment. In a clinical trial
of a treatment for chronic pain, pain reduction and
safety are essential end points, but they may not be suf-
ficient outcomes for a comprehensive evaluation of the
overall benefit of treatment. The complexity of chronic
pain and its negative impact on diverse aspects of func-
tion require the assessment of multiple outcome do-
mains to evaluate treatments comprehensively. 35
The importance of emotional states in chronic pain
has been well recognized.17 Numerous studies18 suggest
that chronic pain is often associated with significant
emotional distress, particularly depression and anxiety.
Moreover, emotional health is central in peoples assess-
ment of their well-being and satisfaction with life.7,36

1012
Turk et al
As a consequence, anxiety and depression have been
suggested to be important outcome domains that
should be considered in all chronic pain clinical trials. 35
Measures of mental state can be useful both as end
points and to identify subgroups of patients for targeted
treatment.
Effective patient-reported outcomes (PROs) are a key
component in the efficient development of analgesic
treatments.1,7,36,39 Several self-report measures of
emotional functioning have been included in clinical tri-
als. Based on a consensus meeting, IMMPACT (Initiative
on Methods, Measurement, and Pain Assessment in Clin-
ical Trials13) recommended that consideration should be
given to using the Beck Depression Inventory (BDI4) and
the Profile of Mood States (POMS26) in clinical research.
Although the Hospital Anxiety and Depression Scale
(HADS41) was considered of potential value, a key limita-
tion was the availability of only 1 preliminary report
testing its sensitivity to change. This concern is balanced
against the ability of the HADS to assess both anxiety
(HADS-A) and depression (HADS-D) with a low-burden
questionnaire that was designed with a medical (not psy-
chiatric) population in mind.
Zigmond and Snaith41 developed the HADS specifically
to avoid reliance on aspects of depression and anxiety
that are also common somatic symptoms of illness and
side effects of treatment (eg, fatigue, weight change,
and insomnia or hypersomnia). Symptoms related to
serious mental disorders were also excluded, because
such symptoms are less common in patients attending
general medical clinics. This was done in an effort to
develop a tool for assessing anxiety and depression in pa-
tients with physical health problems. The HADS consists
of two 7-item subscales; HADS-A (eg, I feel tense or
wound up, I can sit at ease and feel relaxed) and indicated their current and past-24-hour average low
the HADS-D (eg, I feel as if I am slowed down; I have back pain intensity and leg pain intensity at baseline
lost interest in my appearance).
Several studies have demonstrated the psychometric
properties of the HADS for use in the general and diverse
medical population,24,28,40 including chronic pain
patients.15,38 However, to our knowledge, the HADS
has not been validated for use in samples of patients
with acute low back pain (aLBP) and radicular leg pain.
It is recommended that all PROs be evaluated in each
patient population in which they are intended to be
used.37 The primary objective of this study was to deter-
mine the psychometric properties of the HADS, including
sensitivity to change in patients with aLBP and radicular
leg pain.
Methods
We conducted secondary analysis of data from a ran-
domized, double-blind, parallel-group study of tapenta-
dol immediate release (IR) vs oxycodone IR with
treatment duration of 10 days for patients with less
than 30 days previous onset (acute) and without a history
of more than mild previous low back pain (details of the
clinical trial, including patient inclusion criteria, can be
found in Biondi et al5). Validation analyses were con-
ducted using baseline (pretreatment) data; responsive-
The Journal of Pain 1013
ness and predictive validity analyses specifically
included additional data from day 5 of treatment.
HADS subscale scores were considered invalid if they
were missing 2 or more responses, consistent with the
scoring recommendations on the test website. No impu-
tation or replacement of missing values was used for
HADS responses because of the time elapsed between
assessments; missing pain intensity ratings were imputed
via linear interpolation, consistent with the primary end
point calculation of summed pain intensity difference
(SPID).
HADS
The HADS41 requires the respondents to rate the de-
gree to which they have experienced several emotions
in the past week. All items on both subscales are rated
on a 4-point response scale (03). Scale scores are
computed by summing the scores on the individual items
of each subscale. Scale scores range from 0 to 21, with
higher scores indicating more severe symptoms. The
HADS has been shown to be a reliable and valid measure
for use in the general population,6,19,41 as well as in
patients with chronic medical conditions. 23,28,40 On
each subscale, a total score of 0 to 7 is typically
considered normal, 8 to 10 is borderline or suggestive
of possible anxiety/depression, and >10 is indicative of
mood disorder or disease.
Pain Intensity: 11-Point Numeric Rating
Scale
The Numeric Rating Scale (NRS) is commonly used for
assessment of pain and is a core recommended outcome
measure for chronic pain studies.13,16 Respondents
on a standard 0 to 10 scale with 0 being no pain and 10
being worst pain imaginable.
The Brief Pain Inventory-Short Form
The Brief Pain Inventory-Short Form (BPI-SF) is a self-
administered validated tool for the assessment of pain
severity and the impact of pain on daily functions, loca-
tion of pain, pain medications, and pain relief in the
past 24 hours.8 The BPI-SF has been used extensively in
both clinical and research settings, in patients with
pain from chronic diseases or conditions such as cancer,
osteoarthritis, and low back pain, or with pain from
acute conditions such as postoperative pain. The BPI-SF
has been recommended by IMMPACT13 because it pro-
vides a reliable and valid measure of pain and pain inter-
ference with physical and emotional functioning.
The BPI-Pain Interference scale (BPI-I) is used primarily
to measure the extent to which pain interferes with pa-
tients 1) general activity, 2) mood, 3) walking ability, 4)
work both inside and outside the home, 5) relations
with people, 6) sleep, and 7) enjoyment of life. Respon-
dents use an 11-point NRS (0 = does not interfere,
10 = completely interferes) to indicate how much pain
has interfered with these activities in the past 24 hours.
The pain interference score is calculated as the average
1014 The Journal of Pain
of the 7 interference items, and subscale scores for phys-
ical and emotional functioning are also often examined.
A higher mean score denotes higher impairment. In addi-
tion, pain intensity is rated on the 4 BPI-SF questions us-
ing an 11-point NRS ranging from 0 (no pain) to 10 (pain
as bad as you can imagine) at the time of completing the
questionnaire (right now), on average, and at its worst
and least over the past 24 hours.
Pain Qualities: Short-Form McGill Pain
Questionnaire
The Short-Form McGill Pain Questionnaire version
2 (SF-MPQ-2)15 was used to assess both neuropathic
and nonneuropathic aspects of patients pain. The SF-
MPQ-2 consists of 22 descriptor items, each of which
uses a 0 to 10 NRS (0 = none, 10 = worst possible) for re-
sponses. It has been previously validated in patients with
neuropathic and nonneuropathic pain 15 and aLBP.14 The
items are divided into 4 subscales: 6 items each on contin-
uous pain (eg, aching pain), intermittent pain (eg, shoot-
ing pain), and neuropathic pain (eg, cold-freezing pain),
and 4 items of affective descriptors (eg, sickening).
Physical Function: Roland and Morris
Disability Questionnaire
The Roland and Morris Disability Questionnaire
(RMDQ) was used to assess patients level of physical
disability. The RMDQ is a self-administered 24-item vali-
dated functional questionnaire for assessing physical
disability caused by low back pain (eg, I walk more
slowly than usual because of my back.). 30,31 This
questionnaire is suitable for observing short-term
changes in back pain or short-term changes in response
to treatment. The questionnaire has been previously
used in studies on acute20,25 and subacute27 back pain,
as well as chronic back pain. 31 Respondents completing
the RMDQ indicate the number that corresponds to
each item if it applies to how they feel today. Items
circled by the respondent indicate normal activities
that they usually do but have difficulty doing at present
because of their back pain.31 The RMDQ is scored from
0 to 24 by adding up the number of items circled by
the respondent, with higher scores indicating more
physical disability.
Measure of Sleep Quality
Because sleep quality was of particular interest but no
specific sleep outcome was included in the assessments,
the sleep item from the BPI-I scale8 was examined indi-
vidually in lieu of a more detailed assessment. The item
asked patients to Circle the one number that describes
how, during the past 24 hours, pain has interfered with
your: Sleep. Responses are 11-point NRS (0 = does not
interfere, 10 = completely interferes).
Statistical Analysis
The overall approach to evaluating the validity of the
HADS was accomplished in a series of stages. The presence
of floor and ceiling effects was assessed by examination of
HADS in Acute Back Pain
the response distributions for the HADS total (HADS-T)
and the subscale scores. Descriptive statistics and coeffi-
cients for internal consistency and reliability (ie, Cronbach
a) were calculated for the HADS-Tand subscale scores. The
association between the HADS-A and HADS-D subscales
was examined by using a Pearson correlation. Construct
validity was evaluated by examining the associations be-
tween the HADS-Tand subscale scores and other measures
of pain31and pain-related
15
constructs (ie, BPI-I scale,8
RMDQ, SF-MPQ-2 ) using Pearson correlations. The
separate NRS ratings were used to assess back and leg
pain intensity rather than the BPI pain intensity because
this allowed for discrimination between both back pain
and leg pain. Sensitivity to change was evaluated by
determining whether the HADS-T, HADS-A, and HADS-D
scores improved from baseline to day 5 after treatment
with tapentadol IR and oxycodone IR.
Given the objective of the present investigation,
confirmatory factor analysis (CFA) provides a more direct
and informative approach to evaluating measurement
models than conventional exploratory factor analysis
(EFA). In CFA, a priori constraints are necessary to esti-
mate a solution. Briefly, the investigator specifies an
implied model of how covariances between a group of
variables should have been caused by underlying factors
or latent structures. Using the investigators model, the
CFA program estimates a solution covariance matrix.
The program then compares the estimated covariance
matrix with the covariance matrix used. A model that
produces a solution that closely matches the input
covariance matrix is a good fit to the data. Each solution
is direct and unique, and therefore no rotation of the so-
lution is necessary to interpret it theoretically. Moreover,
the tests can be performed in CFA to determine if the es-
timates of sample covariances using the a priori
theoretical model are consistent with the sample covari-
ances, or in other words, whether the data confirm the
substantively generated model.
In CFA, an investigator constructs a measurement
model (ie, a factor structure) that is derived from assump-
tions of the theory of interest. When a researcher choo-
ses a confirmatory approach, the model must be
identified, unlike solutions estimated in traditional
EFA. The issue of identification is complicated, but the
necessary (but not sufficient) condition for a model to
be identified is that the number of parameters to be esti-
mated by the CFA program should be equal to or less
than the number of correlations between measured vari-
ables. EFA solutions do not meet this minimum criterion.
A CFA was performed to evaluate the constructs of
anxiety and depression as assessed by 2 HADS 7-item
subscales (HADS-A, HADS-D). Acceptability of fit of the
factor solutions for the CFA was evaluated based on
the goodness of fit (GFI) >.90, stardardized root mean
square residual (SRMR) <.08, and root mean square
average (RMSEA) <.10. Paired sample t-tests for HADS
subscales and total scores between baseline and the
end of treatment (day 5) were used to determine
responsiveness for each of the Quebec Task Force Classi-
fication (QTFC) groups34 as a result of a known treat-
ment (tapentadol IR and oxycodone IR).
Turk et al
Results
Descriptive Statistics and Psychometrics
A total of 666 patients (50.6% male; 71.5% white;
mean age = 45 years) who had aLBP with associated
radicular leg pain (baseline mean back pain = 7.5, leg
pain = 6.8) consistent with QTFC34 category 3, 4, or 6
(see Tables 1 and 2) were included in the analyses
(category 3, patients with aLBP and pain radiating
below the knee; category 4, patients meeting criteria
for category 3 and neurologic findings suggestive of
lumbosacral radiculopathy [ie, at least 2 abnormal
findings: unilateral abnormality in muscle strength,
deep tendon reflex, or sensation in a dermatomal
pattern]; category 6, patients meeting criteria for
category 4 and having evidence of nerve root
compression on an imaging study. Patients in different
QTFC categories did not show any differences, and
results reported used all patients unless otherwise
noted in the responsiveness and predictions of efficacy
results).
Descriptive statistics confirm that patients used the full
range of responses with reasonable distributions and no
excessive floor (absolute bottom) or ceiling (absolute
top) effects for the HADS-T and both HADS-A and
HADS-D subscales (Fig 1). Two patients with multiple
missing HADS items were excluded. There were no statis-
tically significant correlations between the HADS-T,
HADS-A, and HADS-D with any of the demographic vari-
ables (age, sex, height, weight, or race/ethnicity)
(Table 3). Patients in this study had high normal levels
of anxiety (mean = 7.7, standard deviation [SD] = 3.9)
and depression (mean = 6.7, SD = 3.9) with the total score
mean of 14.4 (SD = 6.9); the internal consistency for the
HADS-T, HADS-A, and HADS-D was good (Cronbach
a = .86, .78, .80, respectively) (Table 4), supporting the
reliability of the HADS-T and both subscales.2
For each of the HADS subscales, scores were calculated
by taking the sum of the item ratings included in the
scale (ie, 14 for the HADS-T and 7 each for the HADS-A
The Journal of Pain 1015
and HADS-D). Patients scores were on the high end of
normal for depression and slightly increased for anxiety.
These findings are consistent with observations for other
medical populations in the literature23,28,40 (Table 4).
There was a moderate correlation between the
HADS-A and HADS-D scales (r = .56, P < .01). An associa-
tion between anxiety and depression has been
frequently reported in the literature.21,32
Correlations With Pain and Disability
Pearson correlations were performed between the
HADS-T and subscales and the BPI-I,8 the BPI sleep quality
item,8 the BPI mood item,8 the RMDQ,31 and the SF-MPQ-
215 scores of pain quality. The correlation coefficients for
each of the measures were statistically significant and
ranged from .296 to .430 (all P values <.001) (Table 5).
These results support the convergent validity of the
HADS.
Bias Analysis
To determine any unexpected biases or inappropriate
discrimination for any demographic factors, Pearson cor-
relations were performed between the HADS scales and
patient age, sex, height, weight, and race (Table 3).
Neither the HADS-T score nor the HADS-A or HADS-D
showed any significant correlation with demographic
factors that might be considered biasing or discrimina-
tory.
CFA
CFA was used to test the a priori factor structure with
items specified to load on each of the HADS subscales
(HADS-A, HADS-D). Single group CFA models including
the entire sample were examined to serve as an overall
test of the constructs and to facilitate comparisons with
previous HADS studies. CFA was performed with Amos
version 19 for Windows (IBM Corp, Armonk, NY). Invari-
ance testing was restricted to measurement models of
the constructs: Anxiety and Depression.

Table 1. Distribution by Sex and Ethnicity of the Study Sample

QTF CATEGORY 3 (N = 326), N (%) QTF CATEGORY 4 or 6 (N = 340), N (%) ALL (N = 666), N (%)
Sex
Male 166 (50.9) 171 (50.3) 337 (50.6)
Female 160 (49.1) 169 (49.7) 329 (49.4)
Race
White 242 (74.2) 234 (68.8) 476 (71.5)
Black or African American 73 (22.4) 82 (24.1) 155 (23.3)
Asian 3 (.9) 14 (4.1) 17 (2.6)
American Indian or Alaska Native 2 (.6) 3 (.9) 5 (.8)
Hawaiian or Pacific Islander 1 (.3) 0 (.0) 1 (.2)
Other 5 (1.5) 7 (2.1) 12 (1.8)
Ethnicity
Hispanic or Latino 47 (14.4) 38 (11.2) 85 (12.8)
Not Hispanic or Latino 278 (85.3) 301 (88.5) 579 (86.9)
Unknown 0 (.0) 0 (.0) 0 (.0)
Not reported 1 (.3) 1 (.3) 2 (.3)

Abbreviation: QTF, Quebec Task Force on Spinal Disorders.

1016 The Journal of Pain HADS in Acute Back Pain
Table 2. Height, Weight, Pain, and HADS Characteristics at Baseline of Study Sample
QTF CATEGORY 3 (N = 326)* QTF CATEGORY 4 OR 6 (N = 340) ALL (N = 666)

MEAN 6 SD MEDIAN MIN, MAX MEAN 6 SD MEDIAN MIN, MAX MEAN 6 SD MEDIAN MIN, MAX
Height (cm) 169.9 6 11.2 170.2 106.0, 198.1 169.6 6 13.3 170.1 103.0, 201.2 169.8 6 12.3 170.1 103.0, 201.2
Weight (kg) 87.1 6 22.7 85.2 47.7, 193.0 87.9 6 25.6 84.5 44.5, 201.0 87.5 6 24.2 84.8 44.5, 201.0
Average low back pain 7.3 6 1.3 7.0 5.0, 10.0 7.6 6 1.3 8.0 5.0, 10.0 7.5 6 1.3 7.0 5.0, 10.0
(010)y
Average index leg pain 6.6 6 1.8 7.0 .0, 10.0 6.9 6 1.9 7.0 .0, 10.0 6.8 6 1.9 7.0 .0, 10.0
(010)z
Current low back pain 7.0 6 1.3 7.0 5.0, 10.0 7.2 6 1.4 7.0 5.0, 10.0 7.1 6 1.4 7.0 5.0, 10.0
(010)x
Current index leg pain 6.2 6 1.9 6.0 .0, 10.0 6.5 6 2.1 7.0 .0, 10.0 6.3 6 2.0 6.0 .0, 10.0
(010){

Abbreviations: QTF, Quebec Task Force on Spinal Disorders; Min, minimum; Max, maximum.
*One subject was missing pain data and is excluded from those rows.
yPlease rate your back pain by checking the one number that best describes your low back pain on average in the last 24 hour s related to this current episode.
zPlease rate your pain by checking the one number that best describes your pain in your affected leg on average in the last 24 hours.
xPlease rate your low back pain by checking the one number that best describes your current pain level.
{Please rate your leg pain by checking the one number that best describes your current pain level.

The adequacy of the specified model was evaluated on hypothesized model fit each of these criteria extremely
3 criteria: GFI, SRMR, and RMSEA with acceptability well for both subscales on all 3 evaluated measures of
thresholds of >.90, <.08 and <.10, respectively. The model fit listed above (see Fig 2 for models and item

Figure 1. HADS total and subscale distributions.

Turk et al The Journal of Pain 1017
Table 3. Correlations of HADS-T, HADS-A, and HADS-D With Demographic Variables
HADS SUBSCALE BASELINE BASELINE BLACK OR AFRICAN
(N = 664) AGE DERIVED SEX WEIGHT (KG) HEIGHT (CM) WHITE AMERICAN ASIAN RACE OTHER
Anxiety
r .050 .009 .006 .045 .045 .042 .016 .028
P .202 .809 .887 .248 .242 .274 .676 .468
Depression
r .017 .024 .052 .043 .031 .024 .006 .066
P .671 .543 .177 .271 .429 .545 .877 .091
Total
r .037 .019 .033 .050 .043 .037 .006 .021
P .337 .630 .397 .202 .268 .337 .883 .582

coefficients). The HADS-A and HADS-D had GFI scores of
.97 and .98, SRMR of .05 and .04, and RMSEA of .08 and
.07, respectively.
Responsiveness (Ability to Detect
Change)
The t-tests for the HADS-T, HADS-A, and HADS-D
showed statistically significant differences between
baseline and day 5 scores for low back pain (Fig 3).
When the QTFC groups are examined independently,
the results are the same: the 2 HADS subscales and total
score show statistically significant differences from base-
line to day 5 scores (results not shown, all P values <.01).
These data suggest that the HADS-T and subscale scores
are responsive to change (in the whole sample and in
the QTFC subgroups), which is an important criterion in
the development and identification of suitable outcome
measures for clinical trials of pain treatments.
Predictions of Efficacy
Predictive validity analysis involved examining the
ability of the HADS to predict individuals response to
treatment. Patient response to treatment was based on
SPIDs (SPID120) at day 5, which was the primary end
point for the clinical trial. SPID120 was computed for
both low back pain and leg pain. The pain intensity dif-
ference (PID) was calculated at each assessment time
point, with PID equal to patients baseline pain intensity
minus patients current pain intensity. The SPID was
calculated to include all ratings over the 120 hours. It
was possible that at 120 hours the current pain level in-
formation was not available. In this case, the PID at hour
120 was calculated by linear interpolation from the ob-
servations immediately before and after 120 hours.
Data were imputed using the last-observation-carried-
Table 4.
Basic Psychometric Properties of the
HADS-T, HADS-A, and HADS-D
MEAN MIN/ % FLOOR/
(SD) MEDIAN MAX % CEILING SKEW CRONBACH a
HADS-A 7.7 (3.9) 7.0 0/21 .8/.2 .42
HADS-D 6.7 (3.9) 6.0 0/21 2.9/.2 .38
HADS-T 14.4 (6.9) 14.0 0/36 .2/0 .31
forward method if the pain intensity was not recorded
for up to hour 120. Missing intermediate observations
were imputed by linear interpolation of valid data adja-
cent to the missing data points. Missing data were
imputed to ensure that all patients had at least 1 morn-
ing and 1 evening observation.
Predictive validity was determined by Pearson correla-
tions between the HADS-T, HADS-A, and HADS-D at base-
line with SPID120 for aLBP and leg pain. A significant
correlation indicated a predictive relationship between
baseline score and treatment efficacy. HADS scores at
baseline were not significantly correlated with efficacy
of treatment (pain relief), as measured by the SPID120
for either back pain or leg pain (Table 6).
Because the HADS and SF-MPQ-2 affective scale were
significantly correlated, we examined the predictive
value of the SF-MPQ-2 affective scale by examining the
correlations between this measure and aLBP and leg
pain. As was true for the HADS, the SF-MPQ-2 was not
associated with either back pain (SF-MPQ-2 affective:
r = .016, P = .0696; HADS-T: r = .003, P = .937) or leg
pain (SF-MPQ-2 affective: r = .078, P = .061, HADS-T:
r = .048, P = .243). Thus, neither the HADS nor the SF-
MPQ-2 affective scale appears to predict the outcomes
of the trial.
Discussion
The HADS has been shown to perform well in assess-
ing the presence and symptom severity of anxiety disor-
ders and depression in somatic, psychiatric, and primary
care patients, as well as in the general population. 6,19
However, there has been some debate regarding the
relative independence of the 2 separate factors of
anxiety and depression; the wording of some of the
items, along with conceptual and theoretical issues,
has led some10 to call for abandoning the measure,
whereas others9,29 suggest that this would be
premature.
We analyzed the reliability, construct validity, and
responsiveness of the HADS in a sample of patients
with aLBP. To our knowledge, this is the first evaluation
of the HADS in this population. Both the Anxiety and
.70
Depression subscales 1) showed good psychometric
.80
.86 properties, 2) had high internal consistency, 3) displayed
good convergent validity, 4) had no unexpected biases,
1018 The Journal of Pain HADS in Acute Back Pain
Table 5. Correlations Between HADS-T, HADS-A, and HADS-D With Interference, Function, and Pain
Scales
HADS SUBSCALE BPI RMDQ BPI SLEEP INTERFERENCE SF-MPQ-2- SF-MPQ- SF-MPQ-2- SF-MPQ-2- SF-MPQ-2-
(N = 664)y BPI-I MOOD TOTAL ITEM CONTINUOUS 2-INTERMITTENT NEUROPATHIC AFFECTIVE TOTAL

Anxiety
r .370* .358* .332* .293* .302* .283* .407* .443* .406*
Depression
r .389* .333* .358* .246* .205* .223* .282* .285* .283*
P <.001 <.001 <.001 <.001 <.001 <.001 <.001 <.001 <.001
Total
r .430* .391* .391* .305* .287* .286* .389* .411* .389*

*P < .001.
y2 subjects were excluded from this and subsequent HADS analyses because of missing HADS data.

5) fit the a priori factor structure, and 6) were highly
sensitive to changes as a result of analgesic treatment.
The moderate correlation between the Anxiety and
Depression subscales is to be expected given the exten-
sive comorbidity observed between anxiety and depres-
sion in large community surveys21,32; thus, it is difficult
to distinguish these constructs empirically. The
correlations of HADS-T, HADS-A, and HADS-D subscales
with pain and disability measures were as expected, and
there does not appear to be any bias with regard to de-
mographic variables. This suggests good convergent
validity and applicability across the demographic
groups in the sample, with the caveat that it was a pre-
dominantly white sample. Consistent with earlier
research,12,28 the CFA models showed extremely good
fit to the a priori Anxiety and Depression subscale
structure in the aLBP sample. In light of the concerns
noted in the IMMPACT recommendations, 13 it was grat-
ifying that both subscales showed strong responsive-
ness to change and may be appropriate for use in
analgesic clinical trials. Of course, responsiveness needs
to be demonstrated in chronic pain trials as well, and
the current study focused on aLBP. Based on the results
of our analyses, we conclude that the HADS is a valid

Figure 2. CFA model fit. Bottom circles represent the subscale construct, labeled rectangles are individual items, and e circles are in-
dependent sources of error. Coefficients are shown next to the link between constructs.
Turk et al
Figure 3. HADS-A, HADS-D, and HADS-T, from baseline to day
5. **P < .01.
instrument for efficient, low-burden assessment of anx-
iety and depression in clinical trials with an aLBP popu-
lation.
The overall responses of the sample to the HADS re-
vealed that the patients in the sample of aLBP included
in this study had levels of anxiety and depression on
the upper edge of normal but not, on average, patholog-
ical. These levels are substantially lower than that
observed in psychiatric patients 11,23 and comparable
with observations in patients with chronic pain.38
Although the HADS41 was originally designed for hos-
pital outpatients, it has been extensively used in primary
care6 and in clinical research. The HADS was found to
perform well in assessing the presence and symptom
severity of anxiety disorders and depression both in
somatic, psychiatric, and primary care patients and
in the general population. 6,19 The results of the
current psychometric analyses and CFAs provide a
comprehensive test of the internal consistency,
convergent validity, and factor structure of the HADS in
many patients (N = 666) with aLBP and radicular leg pain.
Although many self-report measures of emotional
states, including anxiety and depression, are available,
they are often limited by their length and consequent
respondent burden. Moreover, many focus on either
anxiety or depression but not both in a single measure
(eg, the Generalized Anxiety Disorder-7 scale has the
same number of items as the HADS-A but no correspond-
ing depression score 33). There is a growing body of
Table 6.Correlations of HADS-T, HADS-A,
HADS-D, With Back and Leg Pain (n = 585 each)
SPID120 LOW BACK PAIN SPID120 LEG PAIN
HADS-A r = .003 r = .070
P = .951 P = .090
HADS-D r = .003 r = .015
P = .938 P = .715
HADS-T r = .003 r = .048
P = .937 P = .243
The Journal of Pain 1019
research published in the pain literature supporting the
importance of both of these symptom constellations.
Although they may be related, measures of anxiety and
depression provide unique information about patients
experiences and should be considered in clinical trials.35
The brevity of the HADS (7 questions for anxiety and
depression each) compared with other instruments,
such as the POMS26 (65 adjective descriptors), is an impor-
tant advantage for its use in clinical studies because of
low respondent burden. Moreover, unlike the BDI 4 and
Beck Anxiety Inventory,3 the HADS was developed for
use with nonpsychiatric medical patients. In addition,
the HADS does not include items related to physical
symptoms that can be associated with medical conditions
and treatments (eg, appetite, as in the Patient Health
Questionnaire-922) and that may not reflect the psychiat-
ric symptoms observed in patients with mood or anxiety
disorders.
The present study has several limitations as well as
strengths. The results are based on secondary analyses
of a treatment study. The patients were selected if they
met 1 of 3 of the 6 QTFC categories34 that specifically
group patients with aLBP and associated radicular leg
pain. Therefore, the results need to be confirmed in
the other QTFC categories of patients without radicular
leg pain or signs of lumbosacral radiculopathy, particu-
larly those with reported symptoms but few objective
signs. Consequently, caution must be exercised in gener-
alizing the study results to patients with aLBP with
nonradicular symptoms. Moreover, the study sample
comprised patients with aLBP, limiting the generaliz-
ability to other acute pain conditions. Correlation with
other measures of depression or anxiety would provide
stronger support for convergent validity, but no such
measures were available in the current data set. Rela-
tionships with sleep interference were also limited to
assessment by the single BPI item in the absence of a
more comprehensive sleep-specific scale. Similarly, the
HADS was not administered at 2 time points while pa-
tients were in the same state so we were unable to eval-
uate test-retest reliability. Study patients were
predominantly white; hence, the psychometric proper-
ties of the HADS may need to be further confirmed
with other racial and ethnic groups.
The strengths of the study include the relatively large
sample size (N = 666) with specified inclusion and exclu-
sion criteria for eligibility.5 We assessed responsiveness of
the HADS in a clinical trial of treatments with known ef-
fects, thereby increasing our confidence in the sensitivity
of the HADS to change and potential usefulness in future
clinical trials of acute pain conditions. A sophisticated set
of analyses was performed, including the use of CFA,
which allows for direct testing of the underlying struc-
ture proposed (ie, anxiety and depression) rather than
using exploratory procedures.
The results of these analyses are consistent with the re-
sults of previous studies of the HADS and suggest that
this measure may be an acceptable alternative in anal-
gesic clinical trials to the BDI4 and POMS,26 as recommen-
ded by IMMPACT.13 These results provide a basis for use
of the HADS to measure anxiety and depressed mood
1020 The Journal of Pain
in future clinical research, including clinical trials of
treatments for acute pain conditions. Future research
should be conducted to demonstrate the stability of re-
References
1. Acquadro C, Benson R, Dubois D, Leidy NK, Marquis P,
Revicki D, Rothman M, for the PRO Harmonization Group:
Incorporating the patients perspective into drug develop-
ment and communication: an ad hoc task force report of
the patient-reported outcomes (PRO) harmonization group
meeting at the Food and Drug Administration, February 16,
2001. Value Health 6:522-525, 2003
2. Barker C, Pistrang N, Elliott R: Research Methods in Clin-
ical Psychology, 2nd ed. Chichester, UK, John Wiley, 2002
3. Beck AT, Epstein N, Brown G, Steer R: An inventory for
measuring clinical anxiety: psychometric properties. J Con-
sult Clin Psychol 56:893-897, 1988
4. Beck AT, Ward CH, Mendelson M, Mock J, Erbaugh J: An
inventory for measuring depression. Arch Gen Psychiatry 4:
561-571, 1961
5. Biondi D, Xiang J, Benson C, Etropoloski M, Moskovitz B,
Rauschkolb C: Tapentadol versus oxycodone IR for treat-
ment of acute low back pain. Pain Physician 16:E237-E246,
2013
6. Bjelland I, Dahl AA, Haug TT, Neckelmann D: The val-
idity of the Hospital Anxiety and Depression Scale. An
updated literature review. J Psychosom Res 52:69-77,
2002
7. Casarett D, Karlawish J, Sankar P, Hirschman K, Asch D:
Designing pain research from the patients perspective:
what trial endpoints are important to patients with chronic
pain? Pain Med 2:309-316, 2001
8. Cleeland CS, Ryan KM: Pain assessment: global use of the
Brief Pain Inventory. Ann Acad Med Singapore 23:129-138,
1994
9. Cosco TD, Doyle F, Ward M, McGee H: Latent structure of
the Hospital Anxiety and Depression Scale: a 10-year system-
atic review. J Psychosom Res 72:180-184, 2012
10. Coyne J, van Sonderen E: No further research needed:
abandoning the Hospital and Anxiety Depression Scale
(HADS). J Psychosom Res 72:173-174, 2012
11. Crits-Christoph P, Newman MG, Rickels K, Gallop R,
Connolly Gibbons MB, Hamilton JL, Ring-Kurtz S,
Pastva AM: Combined medication and cognitive therapy
for generalized anxiety disorder. J Anxiety Disord 24: 1087-
1094, 2011
12. Dunbar M, Ford G, Hunt K, Der G: A confirmatory factor
analysis of the Hospital Anxiety and Depression Scale:
comparing empirically and theoretically derived structures.
Br J Clin Psychol 39:79-94, 2000
13. Dworkin R, Turk D, Farrar J, Haythornthwaite J,
Jensen M, Katz N, Kerns R, Stucki G, Allen R, Bellamy N,
Carr D, Chandler J, Cowan P, Dionne R, Galer B, Hertz S,
Jadad A, Kramer L, Manning D, Martin S, McCormick C,
McDermott M, McGrath P, Quessy S, Rappaport B,
Robbins W, Robinson J, Rothman M, Royal M, Simon L,
Stauffer J, Stein W, Tollett J, Wernicke J, Witter J: Core
outcome measures for chronic pain clinical trials: IMMPACT
recommendations. Pain 113:9-19, 2005
HADS in Acute Back Pain
sponses on the HADS, replicate our findings, and extend
them to confirm its usefulness with other acute pain
conditions.
14. Dworkin RH, Turk DC, Trudeau JJ, Benson C, Biondi DM,
Katz NP, Kim M, Mou J: Validation of the Short-form McGill
Pain Questionnaire-2 (SF-MPQ-2) in acute low back pain.
J Pain 16:357-366, 2015
15. Dworkin RH, Turk DC, Revicki DA, Harding G, Coyne KS,
Peirce-Sandner S, Bhagwat D, Everton D, Burke LB, Cowan P,
Farrar JT, Hertz S, Max MB, Rappaport BA, Melzack R: Devel-
opment and initial validation of an expanded and revised
version of the Short-form McGill Pain Questionnaire (SF-
MPQ-2). Pain 144:35-42, 2009
16. Farrar JT, Young JP Jr, LaMoreaux L, Werth JL, Poole PM:
Clinical importance of changes in chronic pain intensity
measured on an 11-point numerical pain rating scale. Pain
94:149-158, 2001
17. Gallagher RM, Verma S: Mood and anxiety disorders in
chronic pain. In: Dworkin RH, Breitbart WS (eds): Psychoso-
cial Aspects of Pain: A Handbook for Health Care Providers.
Seattle, IASP Press, 2004, pp 589-606
18. Gatchel RJ, Peng YB, Peters ML, Fuchs PN, Turk DC: The
biopsychosocial approach to chronic pain: scientific ad-
vances and future directions. Psychol Bull 133:581-624, 2007
19. Hermann C: International experiences with the Hospital
Anxiety and Depression Scalea review of validation data
and clinical results. J Psychosom Res 42:17-41, 1997
20. Hurley DA, Minder PM, McDonough SM, Walsh DM,
Moore AP, Baxter DG: Interferential therapy electrode place-
ment technique in acute low back pain: a preliminary inves-
tigation. Arch Phys Med Rehabil 82:485-493, 2001
21. Kessler RC, Berglund P, Demler O: The epidemiology of
major depressive disorder: results from the National Comor-
bidity Survey Replication (NCS-R). JAMA 289:3095-3105,
2003
22. Kroenke K, Spitzer RL, Williams JB: The PHQ-9: validity of
a brief depression severity measure. J Gen Intern Med 16:
606-613, 2001
23. Lynch J, Moore M, Moss-Morris R, Kendrick T: Are pa-
tient beliefs important in determining adherence to treat-
ment and outcome for depression? Development of the
Beliefs About Depression Questionnaire. J Affect Disord
133:29-41, 2011
24. Martin CR, Thompson DR: A psychometric evaluation of
the Hospital Anxiety and Depression Scale in coronary care
patients following acute myocardial infarction. Psychol
Health Med 5:193-201, 2000
25. Mayer JM, Ralph L, Look M, Erasala GN, Verna JL,
Matheson LN, Mooney V: Treating acute low back pain
with continuous low-level heat wrap therapy and/or exer-
cise: a randomized controlled trial. Spine J 5:395-402, 2005
26. McNair DM, Lorr M, Droppleman LF: Profile of Mood
States. San Diego, Educational and Industrial Testing Ser-
vice, 1971
27. Monticone M, Baiardi P, Vanti C, Ferrari S, Pillastrini P,
Mugnai R, Foti C: Responsiveness of the Oswestry Disability
Index and the Roland Morris Disability Questionnaire in Ital-
ian subjects with sub-acute and chronic low back pain. Eur
Spine J 21:122-129, 2012
Turk et al
28. Moorey S, Greer S, Watson M, Gorman C, Rowden L,
Tunmore R, Robertson B, Bliss J: The factor structure and fac-
tor stability of the Hospital Anxiety and Depression Scale in
patients with cancer. Br J Psychiatry 148:255-259, 1991
29. Norton S, Sacker A, Done J: Further research needed: a
comment on Coyne and van Sonderens call to abandon
the Hospital Anxiety and Depression Scale. J Psychosom
Res 73:75-76, 2012
30. Roland M, Fairbank J: The Roland-Morris Disability
Questionnaire and the Oswestry Disability Questionnaire.
Spine 25:3115-3124, 2000
31. Roland M, Morris R: A study of the natural history of
back pain. Part I: Development of a reliable and sensitive
measure of disability in low back pain. Spine 8:141-144, 1983
32. Sartorius N, Ustun TB, Lecrubie Y, Wittchen HU: Depres-
sion comorbid with anxiety: results from the WHO study on
psychological disorders in primary health care. Br J Psychia-
try 168(Suppl 30):s38-s43, 1996
33. Spitzer RL, Kroenke K, Williams JBW, Lowe B: A brief
measure for assessing generalized anxiety disorder. Arch
Intern Med 166:1092-1097, 2006
34. Spitzer WO, LeBlanc FE, Depuis M: Scientific approach to
the assessment and management of activity-related spinal
disorders. A monograph for clinicians report of the Quebec
Task Force on Spinal Disorders. Spine 12(Suppl 7S):1, 1987
35. Turk DC, Dworkin RH, Allen RR, Bellamy N,
Brandenburg N, Carr DB, Cleeland C, Dionne R, Farrar JT,
Galer BS, Hewitt DJ, Jadad AR, Katz NP, Kramer LD,
Manning DC, McCormick CG, McDermott MP, McGrath P,
The Journal of Pain 1021
Quessy S, Rappaport BA, Robinson JP, Royal MA, Simon L,
Stauffer JW, Stein W, Tollett J, Witter J: Core outcome do-
mains for chronic pain clinical trials: IMMPACTrecommenda-
tions. Pain 106:337-345, 2003
36. Turk DC, Dworkin RH, Revicki D, Harding G,
Burke LB, Cella D, Cleeland CS, Cowan P, Farrar JT,
Hertz S, Max MB, Rappaport BA: Identifying important
outcome domains for chronic pain clinical trials: an IM-
MPACT survey of people with pain. Pain 137:276-285,
2008
37. US Department of Health and Human Services: Guid-
ance for Industry: Patient-reported Outcome Measures:
Use in Medical Product Development to Support Labeling
Claims. Bethesda, MD, US Department of Health and Hu-
man Services, 2006
38. Veehof MM, ten Klooster PM, Taal E, Westerhof GJ,
Bohlmeijer ET: Psychometric properties of the Dutch Five
Facet Mindfulness Questionnaire (FFMQ) in patients with fi-
bromyalgia. Clin Rheumatol 30:1045-1054, 2011
39. Willke RJ, Burke LB, Erickson P: Measuring treatment
impact: a review of patient-reported outcomes and other ef-
ficacy endpoints in approved product labels. Control Clin Tri-
als 25:535-552, 2004
40. Woolrich RA, Kennedy P, Tasiemski T: A preliminary psy-
chometric evaluation of the Hospital Anxiety and Depres-
sion Scale (HADS) in 963 people living with a spinal cord
injury. Psychol Health Med 11:80-90, 2006
41. Zigmond AS, Snaith RP: The Hospital Anxiety and
Depression Scale. Acta Psychiatr Scand 67:361-370, 1983