CLINICAL

ARTICLE Clinical Trials 2012; 0: 1–10

TRIALS

Development of the Stanford Expectations of

Treatment Scale (SETS): A tool for measuring
patient outcome expectancy in clinical trials
Jarred Younger, Vanisha Gandhi, Emily Hubbard and Sean Mackey

Background A patient’s response to treatment may be influenced by the expecta-

tions that the patient has before initiating treatment. In the context of clinical trials,
the influence of participant expectancy may blur the distinction between real and
sham treatments, reducing statistical power to detect specific treatment effects.
There is therefore a need for a tool that prospectively predicts expectancy effects on
treatment outcomes across a wide range of treatment modalities.
Purpose To help assess expectancy effects, we created the Stanford Expectations of
Treatment Scale (SETS): an instrument for measuring positive and negative treat-
ment expectancies. Internal reliability of the instrument was tested in Study 1. Cri-
terion validity of the instrument (convergent, discriminant, and predictive) was
assessed in Studies 2 and 3.
Methods The instrument was developed using 200 participants in Study 1. Reliabil-
ity and validity assessments were made with an additional 423 participants in Stu-
dies 2 and 3.
Results The final six-item SETS contains two subscales: positive expectancy (a =
0.81–0.88) and negative expectancy (a = 0.81–0.86). The subscales predict a signifi-
cant amount of outcome variance (between 12% and 18%) in patients receiving
surgical and pain interventions. The SETS is simple to administer, score, and
interpret.
Conclusion The SETS may be used in clinical trials to improve statistical sensitivity
for detecting treatment differences or in clinical settings to identify patients with
poor treatment expectancies. Clinical Trials 2012; 0: 1–10. http://ctj.sagepub.com

Introduction
While the majority of treatment outcome – benefi- several treatment modalities (e.g., acupuncture,
cial as well as adverse – is likely treatment specific, pharmaceutical, and psychotherapeutic) [10–12]
some outcome variance is influenced by nonspecific and, in some cases, may be a better predictor of out-
factors [1]. One such nonspecific factor is patient come than actual treatment condition (active versus
treatment expectancy, or the positive and negative sham) [13]. Expectancy can also be a powerful factor
experiences a patient believes that they will have as in the development of adverse treatment outcomes,
a result of receiving treatment. Pretreatment expec- such as nausea from chemotherapy [14–18]. These
tancy has long been recognized as a factor that expectancies may be driven by multiple factors,
shapes treatment outcome [2–9]. The impact of such including a patient’s own preconceived notions or
expectancies on outcome has been demonstrated in the words used by clinicians [19,20].

Department of Anesthesia, Stanford University School of Medicine, Stanford, CA, USA

Author for correspondence: Jarred Younger, Department of Anesthesia, Stanford University School of Medicine, 780
Welch Road, Suite 207F, Palo Alto, CA 94304, USA.
Email: [email protected]

Ó The Author(s), 2012

Reprints and permission: http://www.sagepub.co.uk/journalsPermissions.nav
Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
10.1177/1740774512465064
2 J Younger et al.

The need to measure patient expectancies may be

particularly important in the context of clinical
trials, where expectancies can blur the distinction
between real and placebo conditions [21–24]. It is
also important, given the large percentage of
patients who report adverse side effects from sham
treatments [21,25,26], to be able to identify the
degree to which expectancy contributes to the
occurrence of adverse events. The need to disentan-
gle expectancy from treatment-specific adverse
effects is especially important, given that reported
adverse events during placebo conditions may clo-
sely match those in the active treatment condition
[27].
It is not known how much patient expectancies
influence outcomes, although it is likely quite vari-
able across treatment modalities. The ability to
assess treatment expectancy effects across studies is
hindered by the lack of a standardized assessment
tool of expectancy. Previous attempts at measuring
pretreatment expectancy have largely involved
items that have not undergone reliability or validity
testing. Earlier studies by researchers such as Irving
Kirsch effectively utilized single-item measures to
determine the impact of expectancies on specific
outcomes [28]. These single-item measures have
demonstrated predictive validity and show that
expectancy can be assessed before treatment initia-
tion. However, most of these tools use language that
limits their application to a specific treatment mod- Figure 1. Timeline of instrument development and testing. Three
ality or disease type, and they are not intended to be studies are described in the article.
used outside of a specific treatment context [29–31].
An exception is the Credibility and Expectancy assess expectancy’s impact on both placebo and
Scale, which was developed in 1972 by Borkovec nocebo effects.
and Nau [32]. The instrument yields two subscores: This article describes the development, reliability
credibility and expectancy. While the instrument testing, and predictive validity testing of the Stan-
has shown success in predicting some outcome vari- ford Expectations of Treatment Scale (SETS). The
ables [33], it is not clear that the instrument (parti- instrument was designed to meet all the criteria out-
cularly the credibility subscale) is a useful tool for lined earlier, including being universally applicable
predicting treatment response [34]. Furthermore, across a wide range of treatment modalities (phar-
because the instrument incorporates different
maceutical, psychological, complementary/alterna-
response scales for different items, it is somewhat
tive, surgical/procedural, behavioral, device based,
complex to complete and score.
etc.). We conducted three studies to create the SETS
Because previous questionnaires of pretreatment
and test its reliability and validity (Figure 1). The stu-
expectancy may not meet the particular demands of
dies are described in the following, with each study
clinical trials, we attempted to develop a new instru-
having corresponding ‘Materials and Methods’ and
ment for such measurement. We had six major goals
‘Results’ sections. Following these descriptions is a
in developing the instrument. First, that it be quick
general discussion.
and easy to administer – consisting of as few items
as possible, and using the same response scale for all
items. Second, that it be easy to score and interpret,
Study 1: development and testing of
with as few subscales as necessary, and subscale
scores that are easy to calculate. Third, that it have the initial item pool
strong internal reliability. Fourth, that it have
demonstrated validity in predicting treatment out- Materials and methods
comes. Fifth, that it be applicable to a wide range of
treatment interventions. And sixth, that it measures In Study 1, an item pool was generated and tested
both positive and negative expectancies, in order to for acceptable psychometric properties in 200

Clinical Trials 2012; 0: 1–10 http://ctj.sagepub.com

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
 Patient treatment expectancy scale 3

Table 1. Bivariate (Pearson’s) correlations between all predictor and outcome variables in Study 3

Positive Negative Optimism Anxiety pos_out neg_out

Positive Pearson correlation (r) 1 20.104 0.283* 20.269* 0.326** 0.057

Sig. (two-tailed) (p) 0.383 0.016 0.022 0.006 0.637
Negative Pearson correlation (r) 20.104 1 20.102 0.269* 0.026 0.416**
Sig. (two-tailed) (p) 0.383 0.393 0.022 0.829 0.000
Optimism Pearson correlation (r) 0.283* 20.102 1 20.391** 0.166 0.033
Sig. (two-tailed) (p) 0.016 0.393 0.001 0.173 0.782
Anxiety Pearson correlation (r) 20.269* 0.269* 20.391** 1 20.051 0.011
Sig. (two-tailed) (p) 0.022 0.022 0.001 0.676 0.926
pos_out Pearson correlation (r) 0.326* 0.026 0.166 20.051 1 20.123
Sig. (two-tailed) (p) 0.006 0.829 0.173 0.676 0.316
neg_out Pearson correlation (r) 0.057 0.416* 0.033 0.011 20.123 1
Sig. (two-tailed) (p) 0.637 0.0003 0.782 0.926 0.316

Variables (from left to right) include positive expectancy, negative expectancy, trait optimism, trait anxiety, positive treatment outcome, and negative treat-
ment outcome. Significant correlations are indicated in boldfaced font.
*
Correlation is significant at the 0.05 level (two-tailed).
**
Correlation is significant at the 0.01 level (two-tailed).

participants. The initial pool of items was generated removal, surgery for back pain, hernia repair, and
by soliciting key phrases from physicians, psycholo- tonsillectomies. Ambulatory surgeries are handled
gists, and researchers with experience in clinical by a separate hospital unit, and those patients under-
trials and/or clinical treatments (psychotherapeutic, going minor procedures were not included in the
pharmacologic, and surgical). The clinicians and sample. Surgery was chosen as the target treatment
researchers (N = 19) were invited via email to provide for scale development because it is a discrete inter-
statements that a patient might endorse if he or she vention with a well-defined timeframe and because
expected a strong beneficial response, or a strong participants were highly likely to complete the treat-
adverse response, to an upcoming treatment. ment. To collect survey responses, participants in the
Respondents were not instructed to think of any par- preoperative waiting room were approached and
ticular treatment modality when generating items. asked to complete the 21-item anonymous form.
Both positive response items and negative response The first 200 individuals who agreed to participate
items were included because patients may expect were included. There were no age restrictions, but
beneficial outcomes, while simultaneously expect- the center was an adult surgery unit, so individuals
ing adverse events to occur. The statements received under the age of 18 years were not seen. Question-
from the experts were highly duplicative, resulting naire sheets also asked for optional age and gender
in just 21 unique items (Supplementary Table 1). A information. All procedures were approved by the
seven-point Likert-type response scale was chosen, Institutional Review Board of Stanford University
with the following anchors: (1) ‘strongly disagree’, School of Medicine. Because no personal identifying
(2) ‘moderately disagree’, (3) ‘slightly disagree’, (4) information was collected and the procedure was
‘neither agree nor disagree’, (5) ‘slightly agree’, (6) judged to pose minimal risk, participants were not
‘moderately agree’, and (7) ‘strongly agree’. The required by the Institutional Review Board (IRB) to
seven-point scale was indicated by our previous provide written informed consent.
study in this area that showed a five-point scale to Psychometric analyses were conducted in SPSS
not have appropriate variance at the higher end of version 19 (IBM, New York). Analyses were con-
the scale, and because other researchers have found ducted by the lead author (J.Y.). Psychometric ana-
it to be superior to other options [35]. lyses were performed at the item level to determine
Participants in the initial screening of items were whether the distributions were appropriate for inclu-
patients preparing to undergo surgery at the Stanford sion in the instrument. Unacceptable skew and kur-
Hospital. Placebo effects have been demonstrated to tosis values were a priori set at below 21.5 or above
shape patients’ perception of quality of outcome in +1.5. Factor analysis was carried out using the maxi-
surgeries [36]. Sample size was determined using the mum likelihood approach, with a varimax rotation.
10:1 observation-to-item ratio [37] as a guiding rule. Factor orthogonality was subsequently tested to jus-
Surgery type varied considerably, including proce- tify the use of a varimax rotation. Internal reliability
dures such as hip or knee replacement, tumor of factors was assessed using Cronbach’s a.

http://ctj.sagepub.com Clinical Trials 2012; 0: 1–10

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
4 J Younger et al.

Results tool. The Flesch–Kincaid Grade Level of the entire

scale was 7.8, indicating an eight-grade reading
The average age of the patient sample was 52.8 years level. The Flesch Reading Ease score was 62%, indi-
(standard deviation (SD) = 15.0), with less than 1% cating standard English level.
omitting age information. The sample was 46.2%
female and 43.8% male (with 10% not reporting
gender). Study 2: further reliability testing and
Two items (see Supplementary Table 1) did not initial assessment of predictive validity
meet our criteria for acceptable distribution: items 1
(skewness = 22.2 and kurtosis = 5.5) and 5 (skewness Study 2 was designed to confirm the internal relia-
= 1.6 and kurtosis = 2.0). These items were rejected bility of the two subscales with an independent
from further analysis. group of participants and establish the ability of the
A factor analysis was performed on the remaining scales to predict treatment outcome.
19 items. Two significant factors emerged, with
eigenvalues of 7.6 and 2.5. Factor 1 consisted of
items 2, 4, 7, 8, 9, 11, 20, and 21. Factor 2 consisted Materials and methods
of items 3, 6, 10, 13, 14, 15, 16, 17, and 19. Factor 1
comprised items measuring positive expectancy, The recruitment and data collection process were
while Factor 2 comprised items measuring negative the same as that used in Study 1, with the exception
expectancy. that treatment outcomes were also collected. As with
Scale reliability analyses were used to eliminate the previous studies, a research assistant approached
items not necessary for measuring positive and patients in the preoperative waiting room at the
negative expectancies. Our goal was to reduce the Stanford Hospital and asked if they would be willing
number of items per subscale to three, under the to participate in the study. Additional participants
condition that acceptable internal reliability could were also recruited in the same manner from the
be achieved with that number of items. Three items Pain Management Center at Stanford. Participants
per subscale was a desirable target because we agreeing to the study completed the six-item expec-
wanted participants to be able to complete the tancy instrument, which was then collected by the
instrument very quickly. research assistant. Participants were also given a
Initial internal reliability of the positive expectancy packet that contained outcome measurements and
factor was 0.87. Items were deleted based on their instructed to fill out the outcome measurements 2
weak contribution to internal reliability (item-total weeks after their treatment. The packet contained a
correlations are reported in Supplementary Table 1). self-addressed, stamped envelope so that the post-
Items were removed serially, and the ‘a if item treatment surveys could be returned anonymously.
removed’ output reassessed on the remaining items The pretreatment expectancy instrument was linked
with each iteration. Items 7, 9, and 20 were retained to the posttreatment outcome measures via a unique
through this process, yielding a final a of 0.86. identifier number stamped on the forms. No identi-
The negative expectancy factor had an initial a of fying information was collected on any of the ques-
0.88. Items were removed using the same process as tionnaires. For their participation, patients were
described earlier. After completing the process, items given a US$5 gift card from local retailers. The pay-
10, 15, and 19 were retained, with a final a of 0.81. ment was added to the study because of the addi-
The final SETS version is seen in Figure 2. tional burden of filling out a response packet at
In addition to the six expectancy items, there is a home and mailing it to the laboratory. To preserve
seventh, yes/no item that asks, ‘Have you ever participant anonymity, the gift card was given as a
received this treatment before now?’. This item is prepayment and therefore was not conditional on
intended to allow researchers a way to assess completing and sending the outcome packet.
whether conditioning and prior exposure may be Treatment success was measured with the seven-
contributing to the expectancy–outcome relation- point, single-item Patient Global Impression of
ship. In this study, individuals indicating prior expo- Change (PGIC) scale [38], using the following for-
sure to the treatment were excluded from analyses. mat, ‘Since my treatment, my medical issue is: very
Scores for positive and negative expectancies were much worse, much worse, minimally worse, not
computed. The positive expectancy subscale had a changed, minimally improved, much improved, or
mean of 5.6 (SD = 1.5) and the negative subscale had very much improved’. The PGIC is a commonly
a mean of 3.8 (SD = 1.7). Men and women did not used outcome measure that is applicable to many
differ on scores of positive (t(197) = 1.19, p = 0.234) treatment modalities and medical conditions [39].
or negative (t(197) = 0.65, p = 0.517) expectancy. The instrument is frequently used as a ‘gold stan-
To determine the readability of the scale, the text dard’ for determining treatment outcome, especially
was submitted to a standard readability assessment when pain is a chief complaint [40,41].

Clinical Trials 2012; 0: 1–10 http://ctj.sagepub.com

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
 Patient treatment expectancy scale 5

Figure 2. Demonstration of the final expectancy instrument. An average of items 1, 3, and 5 yields the positive expectancy score, while
an average of 2, 4, and 6 yields the negative expectancy score. Optional demographic information can also be collected. Respondents
can also indicate whether they have prior experience with the treatment.

http://ctj.sagepub.com Clinical Trials 2012; 0: 1–10

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
6 J Younger et al.

Adverse effects were measured with the item, The same pretreatment and posttreatment ques-
‘How severe were the negative or unwanted effects tions were administered as described in Study 3. In
of your treatment?’ rated on an 11-point scale with addition, two questionnaires were added to the pre-
anchors from ‘no negative effects’ to ‘very severe treatment packet. Trait optimism was measured
negative effects’. with the Life Orientation Test–Revised (LOT-R) [48].
The LOT-R is a widely used five-point, 10-item ques-
Results tionnaire. General anxiety was measured with the
anxiety items of the Hospital Anxiety and Depres-
Complete pretreatment expectancy surveys were sion Scale (HADS) [49]. The anxiety scale consists of
collected on 423 individuals. The sample was 49.5% seven items on a four-point scale. The HADS is
female and 41.5% male, with 9% not reporting gen- widely used in clinical populations.
der. Mean age was 53.2 years (SD = 15.8), with less
than 1% not reporting age. A total of 379 of those Results
individuals came from the surgery unit, undergoing
procedures described in Study 1. Again, ambulatory Completed packets were collected from 69 indivi-
surgery patients were not included. In all, 44 of duals (out of 156 administered, for a 46% response
those individuals came from the Stanford Pain Man- rate). Mean age was 53.7 years (SD = 16.5), with
agement Center and underwent treatments such as 100% of participants providing age information.
trigger point injections, acupuncture, ketamine The sample was 45.5% female and 39.1% male, with
infusions, nerve blocks, and pharmaceutical treat- 15.4% not reporting gender. The mean positive
ments. All individuals were analyzed together. Inter- expectancy score was 5.29 (SD = 1.61), and mean
nal reliability was 0.81 for positive expectancy and negative expectancy score was 4.05 (SD = 1.76). The
0.86 for negative expectancy. positive expectancy subscale had an internal relia-
Predictive validity was assessed on 163 individuals bility of 0.88, and the negative subscale had an a of
(of the 423 above, a 37% response rate) who were 0.85. Again, positive and negative expectancy scores
given and completed a posttreatment outcome were not correlated (r = 0.08, p = 0.595).
questionnaire. In all, 25 of those individuals came Correlations between all predictor and outcome
from the Stanford Pain Management Clinic. Positive variables are presented in Table 1. Positive expec-
expectancy was correlated with beneficial outcome tancy was correlated with positive outcome
(r = 0.34, p \ 0.0001), predicting 12% of outcome (r = 0.33, p = 0.006). Controlling for trait optimism
variance. None of the three individual items outper- (with a partial correlation approach) had minimal
formed the positive subscale at predicting outcome impact on the expectancy–outcome relationship
(rs = 0.22, 0.30, and 0.21). Negative expectancy was (r = 0.32, p = 0.009). Trait optimism had no signifi-
correlated with severity of adverse side effects (r = cant relationship with positive outcome (r = 0.12,
0.43, p \ 0.0001), predicting 18% of outcome var- p = 0.312).
iance. None of the individual items outperformed Negative expectancy was correlated with adverse
the total subscale (rs = 0.35, 0.36, and 0.42). Positive side effect severity (r = 0.416, p = 0.0003). Control-
and negative expectancies were not correlated (r = ling for trait anxiety had no impact on the correla-
0.05, p = 0.560). tion, which after partial correlation was unchanged
(r = 0.416, p = 0.0003). Pretreatment anxiety was not
Study 3: convergent, discriminant, and predictive of adverse treatment outcomes (r = 0.011,
p = 0.926).
predictive validity testing
Both subscales of the SETS showed evidence of
convergent validity with the trait measures of opti-
Materials and methods mism and anxiety. Positive expectancy was corre-
Study 3 was designed to confirm the predictive lated (r = 0.28, p = 0.016) with trait optimism.
validity of the SETS in an independent sample of Negative expectancy was correlated (r = 0.27,
participants, as well as test convergent and discrimi- p = 0.022) with trait anxiety. Evidence for discrimi-
nant validity using trait measures of optimism and nant validity was also seen, as individuals with
anxiety. Some research has suggested that optimism greater trait anxiety were less likely to indicate a posi-
is a predictor of placebo response [42–44]. Similarly, tive expectancy of treatment outcome (r = 20.27,
trait anxiety has been identified as a predictor of p = 0.022).
treatment outcome, including pain after surgery
[45–47]. Using the same procedure described in Discussion
Study 3, individuals completed the SETS, but partici-
pants also completed two additional measures: trait In this article, we report the development of an
optimism and anxiety. instrument to measure treatment expectancy. The

Clinical Trials 2012; 0: 1–10 http://ctj.sagepub.com

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
 Patient treatment expectancy scale 7

SETS (Figure 2) is a quick and easy-to-administer established method for directly measuring placebo
tool that provides separate scores for positive and and nocebo processes in clinical trials. While the
negative expectancies. The instrument contains six neurobiological processes of placebo and nocebo
items, which are all coded in the same scale and responses have begun to be identified [7], quantifi-
direction. Most participants are able to complete the able neurobiological measures are impractical in
SETS in approximately 1 min. Scoring is also quick, large clinical trials. However, treatment expectancies
requiring items 1, 3, and 5 to be averaged for posi- (which are more easily measured) may be a primary
tive expectancy, and items 2, 4, and 6 averaged for driver of placebo and nocebo responses [55–57].
negative expectancy. The simple tool is meant to be Response expectancy theory posits that what people
applicable to a wide range of treatment modalities. experience is strongly shaped by what they expect
The positive and negative subscales demonstrate to experience [2,58].
strong internal reliability and are able to predict Factors other than expectancy may also shape pla-
posttreatment response. cebo and nocebo. For example, prior experience
We found the SETS to predict approximately 12% (i.e., conditioning) can affect the placebo response
of positive outcome variance and 18% of negative [5]. Future studies could examine the moderating
outcome variance. While these percentages are mod- influence that prior treatment exposure has on the
est, we note that most of the individuals received a expectancy–outcome relationship. To facilitate that
surgical procedure – other treatment modalities may type of analysis, an additional item on the SETS
involve a greater element of expectancy. Although form asks if the participant has been exposed to that
untested in this initial report, we also suspect that treatment previously. Other drivers of the placebo
the SETS would predict considerably more outcome response can include patient motivation and effort
variance in individuals receiving sham treatments. [59]. The very act of engaging in positive, goal-
Some evidence suggests that expectancy is a more oriented activities (such as completing a treatment
important predictor of sham treatment outcome protocol) may shape placebo response [60]. As pla-
than active treatment outcome [50]. Further study cebo responses are driven by more than just pretreat-
will be needed to determine the predictive validity ment expectancies, this instrument is not intended
of the SETS in those individuals receiving sham or to predict the full extent of the placebo response.
placebo treatments.
In order to verify the object of participants’ posi- Clinical use
tive and negative expectancies, four additional ques-
tions are included at the bottom of the SETS. Item 7 The accurate measurement of pretreatment expec-
asks the participant to list the treatment they will be tancy may be equally important in clinical contexts.
receiving. This question is to confirm that the parti- Patients are becoming increasingly sophisticated at
cipant is thinking of the correct treatment when online medical research and frequently rely upon
completing the SETS. Item 8 asks about specific ben- information found on the Internet to form their
efits the participant expects to receive, and Item 9 expectancies of treatment responses [61,62]. A great
asks about specific fears or worries the participant deal of misinformation may be accessed through
has about the treatment. These items provide addi- blogs and other sources that contain unrepresenta-
tional qualitative data that can be used to determine tive and extreme (both positive and negative)
what specific outcomes the participant was referring accounts of treatment response [63]. Expectancy
to when completing the instrument. Item 10 asks screening could identify those patients with low
whether the participant has ever received the treat- expectations of a beneficial outcome and/or high
ment previously. The interpretation of treatment fears of adverse events. Such patients may be at a
outcome expectancies would be changed if the par- higher risk of poor treatment response, so it would
ticipant had prior experience with the treatment. be advantageous for the clinician to address those
patients’ expectations prior to beginning the treat-
ment regimen [64]. Part of optimal medical treat-
Does the instrument predict placebo and nocebo ment may involve harnessing knowledge of
responses? pretreatment expectancy to enhance outcomes
[65,66] as studies have shown that early interven-
Much effort in conventional clinical trials involves tions to change preconceived notions of treatment
separating treatment-specific effects from nonspeci- response can increase beneficial effects and mini-
fic effects, and controlling for placebo and nocebo mize adverse side effects [65,67].
responses [8,9]. These effects can adversely impact
statistical power [51,52] and may have been grow- Limitations
ing in the recent years, further complicating study
design and leading to a greater number of negative There are a few limitations to this instrument and
trial outcomes [53,54]. Unfortunately, there is no the methods used to develop it. First, the SETS was

http://ctj.sagepub.com Clinical Trials 2012; 0: 1–10

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
8 J Younger et al.

tested mainly on samples of convenience; the vast Funding

majority of participants were surgical patients at the
This study was funded by the Chris Redlich Endow-
Stanford Hospital. Surgical patients represent a sam-
ment in Pain Research, NIH K24 DA029262, and
ple that is appropriate for testing an outcomes mea-
NIH R00 DA023609.
sure because it is simple to identify participants who
are at a pretreatment stage. It is also easy to verify
that the patients complied with and received the Acknowledgments
treatment. However, it should be noted that in
many cases, surgery is part of a more complex treat- We wish to thank Rebecca McCue, Noorulain Noor,
ment regimen. Generalizability will need to be and Barrett Larson for their invaluable assistance
further established by testing the SETS in other med- with the projects.
ication conditions, treatment modalities, and treat-
ment locations. It is expected that the predictive
ability of treatment expectancy will vary consider-
Conflict of interest
ably over different treatment modalities. For exam- The authors have no conflicts of interest to report.
ple, expectancy may predict approximately 40% of
outcome in cognitive therapy for depression [30].
A second potential problem with the SETS is that References
participants may fear negative repercussions if they
do not report full confidence in the treating physi- 1. Colagiuri B. Participant expectancies in double-blind
randomized placebo-controlled trials: Potential limita-
cian. While the introductory paragraphs of the
tions to trial validity. Clin Trials 2010; 7(3): 246–55.
instrument contain language to assuage fears of 2. Kirsch I. Response expectancy as a determinant of experi-
retaliation, it is possible that some respondents will ence and behavior. Am Psychol 1985; 40(11): 1189–202.
still feel compelled to report very high positive 3. Pollo A, Amanzio M, Arslanian A, et al. Response expec-
expectancy. Other patients may have a defensive tancies in placebo analgesia and their clinical relevance.
need to avoid thinking about treatment failure, thus Pain 2001; 93(1): 77–84.
leading to inflated scores of positive expectancy. 4. Kirsch I. Conditioning, expectancy, and the placebo
Other limitations include the short duration of the effect: Comment on Stewart-Williams and Podd (2004).
outcome assessment and the low response rate for Psychol Bull 2004; 130(2): 341–43; discussion 344–45.
5. Stewart-Williams S, Podd J. The placebo effect: Dissol-
the outcome measures (introducing a possible selec-
ving the expectancy versus conditioning debate. Psychol
tion bias). Also, because there are no existing valid Bull 2004; 130(2): 324–40.
measures of treatment expectancy, construct valid- 6. Holm LW, Carroll LJ, Cassidy JD, Skillgate E, Ahlbom
ity is difficult to test. A. Expectations for recovery important in the prognosis
of whiplash injuries. PLoS Med 2008; 5(5): e105.
7. Tracey I. Getting the pain you expect: Mechanisms of
Conclusion placebo, nocebo and reappraisal effects in humans. Nat
Med 2010; 16(11): 1277–83.
We suggest that the SETS be used as a quick assess- 8. Colloca L, Miller FG. The nocebo effect and its relevance
ment of pretreatment expectancy in clinical trials. for clinical practice. Psychosom Med 2011; 73(7): 598–603.
One use of the SETS would be to include baseline 9. Manchikanti L, Giordano J, Fellows B, Hirsch JA. Pla-
scores of positive and negative expectancies as nui- cebo and nocebo in interventional pain management: A
sance covariates in models testing for differences friend or a foe – Or simply foes? Pain Physician 2011;
14(2): E157–75.
between real and sham treatments. Individuals who
10. Linde K, Witt CM, Streng A, et al. The impact of patient
strongly believe that they will benefit from a treat- expectations on outcomes in four randomized controlled
ment may be more likely to report benefit following trials of acupuncture in patients with chronic pain. Pain
a treatment. Likewise, individuals who believe 2007; 128(3): 264–71.
strongly that they will experience adverse effects 11. Oken B, Flegal K, Zajdel D, et al. Expectancy effect:
may be more likely to report adverse effects. To the Impact of pill administration on cognitive performance
extent that these pretreatment beliefs drive out- in healthy seniors. J Clin Exp Neuropsychol 2008; 30(1): 7–
comes, they represent placebo and nocebo responses 17.
and do not accurately reflect the physiological prop- 12. Lewin AB, Peris TS, Lindsey Bergman R, et al. The role
of treatment expectancy in youth receiving exposure-
erties specific to the active treatment. If pretreat-
based CBT for obsessive compulsive disorder. Behav Res
ment expectancy effects are removed statistically, Ther 2011; 49(9): 536–43.
then the models may have greater power to detect 13. Grotle M. Traditional Chinese acupuncture was not
treatment-specific outcome differences – thus redu- superior to sham acupuncture for knee osteoarthritis but
cing the required sample size and resulting cost asso- delivering treatment with high expectations of improve-
ciated with the study. While we did not test the SETS ment was superior to delivering treatment with neutral
in that manner, we look forward to that application. expectations. J Physiother 2011; 57(1): 56.

Clinical Trials 2012; 0: 1–10 http://ctj.sagepub.com

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
 Patient treatment expectancy scale 9

14. Montgomery G, Tomoyasu N, Bovbjerg D, et al. 32. Borkovec TD, Nau SD. Credibility of analogue therapy
Patients’ pretreatment expectations of chemotherapy- rationales. 1972; 3(4): 257–60.
related nausea are an independent predictor of anticipa- 33. Chambless D, Tran G. Predictors of response to cogni-
tory nausea. Ann Behav Med 1998; 20(2): 104–9. tive-behavioral group therapy for social phobia. J Anxiety
15. Watson M, Meyer L, Thomson A, Osofsky S. Psycholo- Disord 1997; 11(3): 221–40.
gical factors predicting nausea and vomiting in breast 34. Devilly GJ, Borkovec TD. Psychometric properties of the
cancer patients on chemotherapy. Eur J Cancer 1998; credibility/expectancy questionnaire. J Behav Ther Exp
34(6): 831–37. Psychiatry 2000; 31(2): 73–86.
16. Montgomery G, Bovbjerg D. Specific response expectan- 35. Diefenbach MA, Weinstein ND, O’Reilly J. Scales for
cies predict anticipatory nausea during chemotherapy for assessing perceptions of health hazard susceptibility.
breast cancer. J Consult Clin Psychol 2001; 69(5): 831–35. Health Educ Res 1993; 8(2): 181–92.
17. Roscoe JA, Bushunow P, Morrow GR, et al. Patient 36. McRae C, Cherin E, Yamazaki TG, et al. Effects of per-
expectation is a strong predictor of severe nausea after ceived treatment on quality of life and medical outcomes
chemotherapy: A University of Rochester Community in a double-blind placebo surgery trial. Arch Gen Psychia-
Clinical Oncology Program study of patients with breast try 2004; 61(4): 412–20.
carcinoma. Cancer 2004; 101(11): 2701–08. 37. Schwab P. Construct validity in organization behavior.
18. Aapro MS, Molassiotis A, Olver I. Anticipatory nausea In Staw M, Cummings L (eds). Research in Organizational
and vomiting. Support Care Cancer 2005; 13(2): 117–21. Behavior. JAI Press, Greenwich, CT, 1980, pp. 3–43.
19. Varelmann D, Pancaro C, Cappiello EC, Camann WR. 38. Guy W. ECDEU assessment manual for psychopharma-
Nocebo-induced hyperalgesia during local anesthetic cology. US Department of Health, Education, and Wel-
injection. Anesth Analg 2010; 110(3): 868–70. fare, Rockville, MD, 1976.
20. Van Laarhoven AI, Vogelaar ML, Wilder-Smith OH, et 39. Dworkin RH, Turk DC, Farrar JT, et al. Core outcome
al. Induction of nocebo and placebo effects on itch and measures for chronic pain clinical trials: IMMPACT
pain by verbal suggestions. Pain 2011; 152(7): 1486–94. recommendations. Pain 2005; 113(1–2): 9–19.
21. Barsky AJ, Saintfort R, Rogers MP, Borus JF. Nonspecific 40. Fischer D, Stewart AL, Bloch DA, et al. Capturing the
medication side effects and the nocebo phenomenon. patient’s view of change as a clinical outcome measure.
JAMA 2002; 287(5): 622–27. JAMA 1999; 282(12): 1157–62.
22. Amanzio M. Do we need a new procedure for the assess- 41. Farrar JT, Young JP, LaMoreaux L, Werth JL, Poole RM.
ment of adverse events in anti-migraine clinical trials? Clinical importance of changes in chronic pain intensity
Recent Pat CNS Drug Discov 2011; 6(1): 41–47. measured on an 11-point numerical pain rating scale.
23. Häuser W, Bartram-Wunn E, Bartram C, Reinecke H, Pain 2001; 94(2): 149–58.
Tölle T. Systematic review: Placebo response in drug trials 42. Scheier MF, Matthews KA, Owens JF, et al. Dispositional
of fibromyalgia syndrome and painful peripheral diabetic optimism and recovery from coronary artery bypass sur-
neuropathy-magnitude and patient-related predictors. gery: The beneficial effects on physical and psychological
Pain 2011; 152(8): 1709–17. well-being. J Pers Soc Psychol 1989; 57(6): 1024–40.
24. Papadopoulos D, Mitsikostas DD. A meta-analytic 43. Morton DL, Watson A, El-Deredy W, Jones AK. Repro-
approach to estimating nocebo effects in neuropathic ducibility of placebo analgesia: Effect of dispositional
pain trials. J Neurol 2011; 259(3): 436–47. optimism. Pain 2009; 146(1–2): 194–98.
25. Pogge RC. The toxic placebo. I. Side and toxic effects 44. Geers AL, Wellman JA, Fowler SL, Helfer SG, France
reported during the administration of placebo medicine. CR. Dispositional optimism predicts placebo analgesia. J
Med Times 1963; 91: 773–78. Pain 2010; 11(11): 1165–71.
26. De la Cruz M, Hui D, Parsons HA, Bruera E. Placebo 45. Rudin A, Wölner-Hanssen P, Hellbom M, Werner MU.
and nocebo effects in randomized double-blind clinical Prediction of post-operative pain after a laparoscopic
trials of agents for the therapy for fatigue in patients with tubal ligation procedure. Acta Anaesthesiol Scand 2008;
advanced cancer. Cancer 2010; 116(3): 766–74. 52(7): 938–45.
27. Amanzio M, Corazzini LL, Vase L, Benedetti F. A sys- 46. Mertens MC, Roukema JA, Scholtes VP, De Vries J. Trait
tematic review of adverse events in placebo groups of anxiety predicts outcome 6 weeks after cholecystectomy.
anti-migraine clinical trials. Pain 2009; 146(3): 261–69. A prospective follow-up study. Ann Behav Med 2011;
28. Price DD, Milling LS, Kirsch I, et al. An analysis of fac- 41(2): 264–69.
tors that contribute to the magnitude of placebo analge- 47. Van Esch L, Roukema JA, Ernst MF, Nieuwenhuijzen
sia in an experimental paradigm. Pain 1999; 83(2): 147– GA, De Vries J. Combined anxiety and depressive symp-
56. toms before diagnosis of breast cancer. J Affect Disord
29. Dennehy EB, Webb A, Suppes T. Assessment of beliefs 2011; 136(3): 895–901.
in the effectiveness of acupuncture for treatment of psy- 48. Scheier MF, Carver CS. Distinguishing optimism from
chiatric symptoms. J Altern Complement Med 2002; 8(4): neuroticism (and trait anxiety, self-mastery, and self-
421–25. esteem): A reevaluation of the Life Orientation Test. J Pers
30. Goossens ME, Vlaeyen JW, Hidding A, Kole-Snijders A, Soc Psychol 1994; 67(6): 1063–78.
Evers SM. Treatment expectancy affects the outcome of 49. Zigmond AS, Snaith RP. The hospital anxiety and
cognitive-behavioral interventions in chronic pain. Clin J depression scale. Acta Psychiatr Scand 1983; 67(6): 361–
Pain 2005; 21(1): 18–26; discussion 69–72. 70.
31. Mao J, Xie S, Bowman M. Uncovering the expectancy 50. Fountoulakis KN, Möller HJ. Efficacy of antidepressants:
effect: The validation of the acupuncture expectancy A re-analysis and re-interpretation of the Kirsch data. Int J
scale. Altern Ther Health Med 2010; 16(6): 22–7. Neuropsychopharmacol 2011; 14(3): 405–12.

http://ctj.sagepub.com Clinical Trials 2012; 0: 1–10

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015
10 J Younger et al.

51. Chen YF, Yang Y, Hung HM, Wang SJ. Evaluation of per- involving inert (placebo) substances. J Psychosom Res
formance of some enrichment designs dealing with high 2008; 65(5): 405–13.
placebo response in psychiatric clinical trials. Contemp 61. Van Woerkum CM. The Internet and primary care physi-
Clin Trials 2011; 32(4): 592–604. cians: Coping with different expectations. Am J Clin Nutr
52. Ivanova A, Qaqish B, Schoenfeld DA. Optimality, sam- 2003; 77(4): 10165–85.
ple size, and power calculations for the sequential parallel 62. Iverson S, Howard K, Penney B. Impact of Internet use
comparison design. Stat Med 2011; 30(23): 2793–803. on health-related behaviors and the patient-physician
53. Enserink M. Can the placebo be the cure? Science 1999; relationship: A survey-based study and review. J Am
284(5412): 238–40. Osteopath Assoc 2008; 108(12): 699–711.
54. Manek NJ, Tiller WA. A new perspective on ‘the placebo 63. Hallingbye T, Serafini M. Assessment of the quality of
effect’: Untangling the entanglement. Med Hypotheses postherpetic neuralgia treatment information on the
2011; 77(4): 614–9. Internet. J Pain 2011; 12(11): 1149–54.
55. Goldstein AP. Participant expectancies in psychotherapy. 64. Sullivan M, Tanzer M, Reardon G, et al. The role of pre-
Psychiatry 1962; 25: 72–79. surgical expectancies in predicting pain and function
56. Howland RH. Understanding the placebo effect. Part 2: one year following total knee arthroplasty. Pain 2011;
Underlying psychological & neurobiological processes. J 152(10): 2287–93.
Psychosoc Nurs Ment Health Serv 2008; 46(6): 15–18. 65. Webb MS, Hendricks PS, Brandon TH. Expectancy
57. Kelley JM. The relation between motivational concor- priming of smoking cessation messages enhances the
dance and response expectancy. J Psychosom Res 2008; placebo effect of tailored interventions. Health Psychol
65(5): 415–16. 2007; 26(5): 598–609.
58. Kirsch I. How Expectancies Shape Experience. American Psy- 66. Colloca L, Miller FG. Harnessing the placebo effect: The
chological Association, Washington, DC, 1999. need for translational research. Philos Trans R Soc Lond B
59. Gaitan-Sierra C, Hyland ME. Nonspecific mechanisms Biol Sci 2011; 366(1572): 1922–30.
that enhance well-being in health-promoting behaviors. 67. Roscoe JA, O’Neill M, Jean-Pierre P, et al. An explora-
Health Psychol 2011; 30(6): 793–6. tory study on the effects of an expectancy manipulation
60. Hyland ME, Whalley B. Motivational concordance: An on chemotherapy-related nausea. J Pain Symptom Manage
important mechanism in self-help therapeutic rituals 2010; 40(3): 379–90.

Clinical Trials 2012; 0: 1–10 http://ctj.sagepub.com

Downloaded from ctj.sagepub.com at University of Birmingham on March 22, 2015