Barsky AJ, Ahern DK. Cognitive Behavior Therapy for HypochondriasisA Randomized Controlled Trial. JAMA. 2004;291(12):1464-1470. doi:10.1001/jama.291.12.1464
Author Affiliations: Department of Psychiatry, Brigham and Women's Hospital and Harvard Medical School, Boston, Mass.
Context Hypochondriasis is a chronic, distressing, and disabling condition that
is prevalent in ambulatory medical practice. Until recently, no specific treatment
has been clearly demonstrated to be effective.
Objective To assess the efficacy of a cognitive behavior therapy (CBT) for hypochondriasis.
Design A randomized, usual care control group design, conducted between September
1997 and November 2001. The individual primary care physician was the unit
of randomization, and all patients clustered within each physician's practice
were assigned to the experimental treatment (individual CBT and a consultation
letter to the primary care physician) or to the control condition. Subjects
were assessed immediately before and 6 and 12 months after the completion
Setting and Participants Participants were 80 patients from primary care practices and 107 volunteers
responding to public announcements, all of whom exceeded a predetermined cutoff
score on a hypochondriasis self-report questionnaire on 2 successive occasions.
Intervention A scripted, 6-session, individual CBT intervention was compared with
medical care as usual. The CBT was accompanied by a consultation letter sent
to the patient's primary care physician.
Main Outcome Measures Hypochondriacal beliefs, fears, attitudes, and somatic symptoms; role
function and impairment.
Results A total of 102 individuals were assigned to CBT and 85 were assigned
to medical care as usual. The sociodemographic and clinical characteristics
of the 2 groups were similar at baseline. Using an intent-to-treat analytic
strategy, a consistent pattern of statistically and clinically significant
treatment effects was found at both 6- and 12-month follow-up, adjusting for
baseline covariates that included educational level, generalized psychiatric
distress, and participant status (patient vs volunteer). At 12-month follow-up,
CBT patients had significantly lower levels of hypochondriacal symptoms, beliefs,
and attitudes (P<.001) and health-related anxiety
(P = .009). They also had significantly less impairment
of social role functioning (P = .05) and intermediate
activities of daily living (P<.001). Hypochondriacal
somatic symptoms were not improved significantly by treatment.
Conclusion This brief, individual CBT intervention, developed specifically to alter
hypochondriacal thinking and restructure hypochondriacal beliefs, appears
to have significant beneficial long-term effects on the symptoms of hypochondriasis.
Hypochondriasis is defined as a persistent fear or belief that one has
a serious, undiagnosed medical illness. Occurring in as many as 5% of medical
outpatients, it is a prevalent, disabling, and chronic condition that has
generally been refractory to psychological and pharmacological treatment and
costly for the health care system.1- 6 Until
recently, there was no empirically validated treatment for the hypochondriacal
patient's somatic symptoms, belief in the presence of an undiagnosed disease,
health-related anxiety, and bodily preoccupation. Although limited by small
samples, lack of randomization and control groups, the absence of long-term
follow-up, limited generalizability, and the absence of validated outcome
measures,7- 12 the
existing treatment literature suggests a role for a variety of psychosocial
interventions, although none has been conclusively validated. In the only
large-scale, rigorous, randomized controlled trial to date, cognitive therapy
and behavioral stress management were both more effective than a waiting list
We have proposed that hypochondriasis can be understood as a self-perpetuating
and self-validating disorder of cognition and bodily perception.14- 16 In
this formulation, personally threatening life events prompt predisposed individuals
to suspect that they have become ill. This suspicion leads them to selectively
attend to benign bodily sensations and health information that confirm their
suspicion and to ignore disconfirmatory evidence. Benign bodily sensations
are thereby amplified and misattributed to the putative disease, which further
substantiates their disease convictions. We have developed a cognitive behavior
therapy (CBT) that specifically targets the cognitive and behavioral amplifiers
of benign bodily symptoms that propel this hypochondriacal cycle of disease
conviction and symptom amplification. Patients are helped to correct faulty
symptom attributions, restructure beliefs and expectations about health and
disease, correct misunderstandings about the medical care process, modify
maladaptive illness behaviors, and learn techniques of selective attention
We hypothesized that the CBT intervention would alleviate the symptoms
of hypochondriasis more effectively than medical care as usual.
A randomized, usual care control group design was used. To avoid a "contamination"
effect, the individual primary care physician was the unit of randomization,
and all patients clustered within each physician's practice were assigned
to the experimental treatment (individual CBT and a consultation letter to
the primary care physician) or to the control condition. Subjects were assessed
immediately before and 6 and 12 months after the completion of treatment.
Subjects were recruited by screening with a self-report hypochondriasis
questionnaire (composed of the Whiteley Index and Somatic Symptom Inventory),
using a predetermined cutoff score selected to identify a sample, half of
whom met Diagnostic and Statistical Manual of Mental Disorders,
Fourth Edition (DSM-IV) criteria for hypochondriasis
and half of whom had subthreshold (subdefinitional) hypochondriasis. The precise
cutoff score of 150 (score range, 52-258) was determined from previous studies3,17- 19 and
was chosen because we believed that many subthreshold individuals were sufficiently
hypochondriacal to be in need of treatment and would be able to benefit from
it. Signed informed consent was obtained in accordance with the Brigham and
Women's Hospital Human Research Committee requirements. Subjects meeting initial
eligibility criteria and exceeding the screening cutoff completed the baseline
research battery, which included a second administration of the screening
questionnaire and a determination of their medical morbidity.
Physicians were randomized immediately following the baseline research
interview by a staff member not connected to the research, using a random
numbers table. Randomization was stratified by physician seniority (house
staff or attending staff) and by practice volume (part-time or full-time for
attending staff, primary care program or traditional internal medicine program
for house staff). All research data were collected by research assistants
who were blind to the patient's treatment status, and the study therapists
and investigators had no foreknowledge of treatment assignment. The therapists
had no role in data collection. Subjects were compensated for each research
interview, but not for undergoing treatment.
Subjects were accrued from 2 sources: (1) successive patients attending
primary care practices in 2 large, academic medical centers, and (2) volunteers
responding to public announcements of a treatment study for "health anxiety
The inclusion criteria were age older than 18 years, English fluency
and literacy, having seen a primary care physician in the last 12 months,
and exceeding the hypochondriasis screening cutoff score on both occasions.
Exclusion criteria included major medical morbidity expected to worsen significantly
in the next 12 months; somatoform pain disorder; psychosis or suicide risk;
and ongoing, symptom-contingent, disability determinations, workers' compensation
proceedings, or litigation.
Cognitive behavior therapy was administered individually in six 90-minute
sessions at weekly intervals. Each session was tightly scripted (manual available
from the authors) and devoted to 1 of 5 factors that cause patients to amplify
somatic symptoms and misattribute them to serious disease: attention and bodily
hypervigilance, beliefs about symptom etiology, circumstances and context,
illness and sick role behaviors, and mood. Each session consisted of educational
information about the symptom amplifiers, an illustrative exercise, and a
discussion to personalize the material presented. The 3 study therapists had
master's or doctoral degrees and prior CBT experience. Treatment sessions
were held in offices in the department of psychiatry.
Because it was considered essential that the patients' ongoing medical
management be coordinated with the CBT to alter their understanding of symptoms,
disease, and medical care, a standardized consultation letter was sent to
each patient's primary care physician. This letter (available from the authors)
contained the following 5 practical suggestions for medical management that
were designed to augment the individual therapy: (1) make improved coping
with somatic symptoms rather than symptom elimination the goal of medical
management; (2) uncouple access to the physician from symptom status by scheduling
regular appointments; (3) provide only limited reassurance; (4) explain the
patient's symptoms using the model of cognitive and perceptual symptom amplification;
and (5) be conservative in medical diagnosis and treatment, within the bounds
of appropriate medical practice.
Treatment fidelity was assessed by auditing audiotapes of randomly selected
therapy sessions from all 3 therapists; adherence to the CBT manual was excellent.
Receipt of the consultation letter was acknowledged by 96.8% of the primary
Outcome Variables. The primary outcome measure was the Whitely Index, a widely used self-report
questionnaire of hypochondriacal attitudes and beliefs, whose validity, reliability,
and sensitivity to change have been demonstrated.20,21 Secondary
outcomes included health-related anxiety, assessed with the Health Anxiety
Inventory, a 14-item, self-report questionnaire that is minimally influenced
by the presence of major medical illness and has good validity, internal consistency,
and reliability.22 The frequency of hypochondriacal
thoughts was assessed with the Hypochondriacal Cognitions Questionnaire, requiring
the patient to rate how often each of 18 disease-related thoughts occurs.
Hypochondriacal somatic symptoms were assessed with the Somatic Symptom Inventory.17,23 The reliability, internal consistency,
and convergent validity of this 13-item questionnaire have been demonstrated
previously.17,18,23- 26 A
clinical diagnosis of DSM-IV hypochondriasis was
made by trained research assistants using the Structured Diagnostic Interview
for Hypochondriasis, a highly structured interview shown in previous work
to have concurrent, convergent, and discriminant validity.27- 29
The tendency to amplify benign bodily sensation and experience it as
noxious, unpleasant, and alarming was assessed with the Somatosensory Amplification
Scale. This questionnaire has good reliability and validity and is sensitive
to change as a result of attention training.14,30- 33
Role impairment and functional status were determined with the Functional
Status Questionnaire.34 This valid and reliable
self-report questionnaire was developed for use in ambulatory medical populations.34,35 It includes subscales assessing intermediate
activities of daily living (eg, doing errands, working around the house) and
social activities (eg, seeing friends, participating in community activities).
Covariates. Psychiatric comorbidity was assessed with the Hopkins Symptom Checklist-90.36,37 This widely used, 90-item instrument
has excellent psychometric properties and provides an overall measure of generalized
psychiatric distress.37 Participant status
indicated whether the subject was a patient accrued in the clinic or a volunteer
recruited from the community. Aggregate medical morbidity was assessed in
2 ways: the Duke Severity of Illness Scale38,39 was
used for patients accrued from primary care practices. This structured audit
of the medical record was conducted by a research physician who was blind
to treatment status. For study volunteers, their primary care physicians provided
2 ordinal ratings of aggregate medical morbidity.
An intent-to-treat approach was used in all analyses. The last-observation-carried-forward
approach was used to impute scores for missing data values. Although the physician
was the unit of randomization, only 11 physicians had more than 1 subject.
Thus, any approach to estimating clustering as an adjustment for physician
effects would likely lead to unstable estimates of the within-cluster correlation.
Therefore, to address the potential effect of clustering on the results, we
selected 1 subject at random from each of these 11 physicians for analysis,
resulting in the exclusion of 24 subjects (n = 163). These results did not
differ in any respect from those obtained with the full sample (N = 187),
which was therefore used for all subsequent analyses. The 2 groups were compared
6 and 12 months after treatment on the primary and secondary outcome variables.
General linear modeling was used as the primary analytic approach, in which
a series of univariate and multivariate repeated measures analysis of covariance
(ANCOVA and MANOVA) models were used. This approach permitted the modeling
of both individual and sets of outcome measures as a function of treatment
effects (between subjects), time of assessment (within subjects), and covariates
(educational level, generalized psychiatric distress, and participant status
[patient vs volunteer]). Hence, the interaction effect of treatment ×
time was of most interest. In cases where the overall MANOVA was significant,
or a priori contrasts were planned, only the ANCOVA results are presented,
using the Greenhouse-Geisser adjustment. Individual, planned contrasts were
calculated for each outcome measure comparing baseline and 6-month, and baseline
and 12-month assessments, respectively. Effect sizes and threshold values
for clinical significance were also derived for each analysis. All tests of
statistical significance were interpreted with a criterion of P<.05. Statistical analyses were performed using SPSS, release 12.0.1
for Windows (SPSS Inc, Chicago, Ill).
The study was conducted between September 1997 and November 2001. A
total of 6307 individuals completed the screening questionnaire, of whom 776
(12.3%) exceeded the cutoff score (Figure
1). Two hundred nineteen individuals declined to participate, 156
proved to be ineligible, and 214 could not be reached subsequently, resulting
in a total of 187 subjects (30.2% of those eligible) who participated. A random
sample (n = 191) of the 589 nonparticipants was compared with those who participated.
The nonparticipants were older (46.4 vs 42.3 years, t376 = 2.96, P = .003), had less education (19.1%
completed college vs 29.4%, χ26 = 42.6, P<.001), and were more likely to be male (38.9% vs 23.5%, χ21 = 10.5, P<.001).
Of the 102 patients in the treatment arm, 63 (61.8%) attended all 6
sessions, 13 (12.7%) attended 4 or 5 sessions, 12 (11.8%) attended 1 to 3
sessions, and 14 (13.7%) attended none. Six-month follow-up was obtained for
85 (83.3%) of the 102 treatment patients and for 76 (89.4%) of the 85 control
patients (87% in person and 13% by telephone). Twelve-month follow-up was
obtained for 92 (90.2%) of treatment patients and 78 (91.8%) of control patients
(84.7% in person and 15.3% by telephone). These attrition rates do not differ
significantly between groups. Six-month follow-up was obtained for 92% of
treatment completers (those attending ≥4 sessions) and 57.6% of treatment
dropouts. Twelve-month follow-up was obtained for 94.7% of treatment completers
and 76.9% of treatment dropouts.
The sociodemographic characteristics of the 2 treatment groups did not
differ significantly (Table 1).
They were predominantly women, middle-aged, and reported a history of hypochondriasis
for approximately 11 years. Educational level and generalized psychiatric
distress did not differ significantly in the 2 groups at baseline but were
used as covariates in the analyses because of their established relationships
to the outcome variables of interest. The treatment and control groups did
not differ significantly in aggregate medical morbidity at baseline.
Eighty patients were recruited from primary care practices and 107 were
volunteers. At baseline, volunteers were significantly more symptomatic and
more disabled than patients on the major outcome measures of hypochondriacal
symptoms (P<.001), health-related anxiety (P<.001), and intermediate activities of daily living
(P<.001). Repeated measures of ANCOVA were performed
on the Whiteley Index modeled as a function of participant status (patients
vs volunteers), treatment (CBT vs usual care) and assessment point (baseline,
6 months, and 12 months), and all 2- and 3-way interaction terms. The 3-way
interaction was significant (F2,362 = 3.38, P = .04), indicating that volunteers had higher Whiteley Index scores
at baseline and experienced a larger treatment effect (difference between
CBT and usual care) at 12 months than the patients (F1,181= 5.04, P = .03), but not at 6 months (F1,181 = 2.37, P = .13). Consequently, participant status (patient or
volunteer) also was included as a covariate in the models.
The results for the Whiteley Index, the primary outcome measure, are
shown in Table 2, using repeated
measures ANCOVA. There was a statistically significant interaction effect
for group (treatment vs control) by assessment time (baseline, 6-month, and
12-month follow-up). Post-hoc tests of within-subjects contrasts disclosed
a statistically significant improvement in the treatment vs control group
at both 6-month (F1,182= 20.1, P<.001)
and 12-month follow-up, compared with baseline (F1,182 = 14.2, P<.001). The treatment effect size for the Whiteley
Index was r = 0.31 at 6 months and r = 0.27 at 12 months.
The secondary outcome measures of hypochondriacal symptoms were analyzed
in the same manner as the Whiteley Index. For hypochondriacal thought frequency,
health anxiety, and somatosensory amplification, the interactions between
group and assessment were statistically significant, indicating a significant
treatment effect on these measures (see Table 2). Post-hoc tests of within-subjects contrasts for hypochondriacal
thought frequency revealed a significantly greater improvement in the treatment
vs control group at 6-month (F1,182= 6.97, P = .009), and at 12-month follow-up (F1,182= 5.64, P = .02). For health anxiety, post-hoc tests of within-subjects
contrasts revealed a significantly greater improvement in the treatment vs
control group at both 6 months (F1,182= 6.73, P = .01) and 12 months compared with baseline (F1,182= 7.01, P = .009). For amplification, individual tests of within-subjects
contrasts revealed a significantly greater improvement in the treatment vs
control group at 6 months (F1,182= 8.47, P =
.004) and at 12 months (F1,182= 7.70, P =
.006). Hypochondriacal somatic symptoms were not significantly improved by
The 2 treatment groups were compared on measures of functional status
using MANCOVA as the omnibus test, with intermediate and social activities
combined as dependent measures, again including baseline educational level,
psychiatric comorbidity, and participant status as covariates. Table 3 indicates a statistically significant interaction effect
for group by assessment time for intermediate activities (F2,364 =
12.32, P<.001), but no statistical significance
for social activities (F2,364 = 2.29, P =
.10). Post-hoc tests of within-subjects contrasts for intermediate activities
of daily living revealed that there was a statistically significant improvement
in the treatment vs control group at both 6 months (F1,182 = 18.44, P<.001) and at 12 months (F1,182 = 18.30, P<.001). The treatment effect size for intermediate
activities at 12 months was r = 0.30. For social
activities, post-hoc tests of within-subjects contrasts revealed that there
was a statistically significant improvement in the treatment vs control group
at 12 months (F1,182 = 3.74, P = .05),
but no significance at 6 months (F1,182 = 3.06, P = .07). The treatment effect size for social activities at 12 months
was r = 0.23.
No therapist effect was found, ie, when the patients treated by each
of the therapists were compared, no significant differences were found for
any of the outcome measures. Although the intervention did not include treatment
for comorbid psychiatric disorder, subjects were free to obtain such treatment
as they or their physicians saw fit. At 6-month follow-up, 20 CBT patients
(19.6%) and 19 control patients (22.4%) had initiated a psychotropic medication
or care with a mental health professional since inception. At 12 months, 6
CBT patients (5.9%) and 5 control patients (5.9%) had initiated a psychotropic
medication or mental health care since their 6-month follow-up. These rates
do not differ significantly between groups.
This 6-session CBT, specifically targeting the cognitive and perceptual
mechanisms thought to underlie hypochondriasis, appears to significantly improve
a range of hypochondriacal symptoms, beliefs, and attitudes. These effects
are evident at 6-month follow-up and persist at 12 months. The effects on
role functioning are not consistently significant at 6 months but emerge at
12 months, the primary end point. These treatment effects are seen using an
intent-to-treat analysis, after adjusting for psychiatric comorbidity, sociodemographic
characteristics, and participant status (patient vs volunteer) at baseline.
The findings are compatible with the only other major trial reported to date
and expand on it by having a control group available for comparison at long-term
Though the magnitude of the treatment effect is modest, it is important
to remember that hypochondriasis generally has been considered a refractory
and chronic disorder (the mean duration of illness was 11 years in this study)
for which there has been no empirically validated treatment. In addition,
this CBT was brief (only 6 sessions) and included no follow-up "booster" sessions.
Finally, patients were not treated in the study for comorbid psychiatric disorder,
and the continued presence of these disorders likely moderated the treatment
The study has several limitations. First, many eligible patients did
not participate, limiting the generalizability of the findings. Those who
did consent to participate might have been more receptive to a psychosocial
approach and hence benefited more from it than those who did not consent.
On the other hand, the study participants might be less severely hypochondriacal
and more responsive to their medical physician's ministrations and/or more
likely to improve spontaneously, which would tend to reduce the treatment
effect. In addition, 25% of the subjects attended less than 4 treatment sessions,
suggesting that future efforts must be directed toward reducing treatment
dropout. This is not unusual for such studies however; Kashner et al,40 for example, found that 56% of somatization disorder
patients randomized to group therapy failed to attend a single session.
Second, we lacked an "attention" control, ie, a generic psychosocial
intervention providing nonspecific attention, support, concern, and positive
expectation. This limits our ability to attribute the treatment effect to
the specific cognitive and behavioral strategies of the intervention. However,
the fact that the cognitive processes thought to underlie the disorder (eg,
hypochondriacal cognitions, health beliefs, amplification) improved with treatment
suggests that the treatment had a specific effect.
Third, considerable improvement occurred in the control group. This
was likely due to the inadvertent inclusion of patients with transient hypochondriasis,
probably because the 2 screening measures were too close to each other in
time (approximately 3 weeks). Additionally, regression to the mean and the
supportive effect of being enrolled in a longitudinal study contributed to
the high rate of spontaneous improvement. A Hawthorne effect may also have
occurred whereby control physicians, having learned of the study, made a greater
effort to help their hypochondriacal patients.
Finally, study subjects came from 2 different sources. Participant status,
however, was included as a covariate in all analyses, and the fact that these
2 groups differed at baseline confers some measure of generalizability on
Hypochondriacal attitudes and concerns improved more than somatic symptoms
did. This finding, although it might seem counterintuitive, was actually expected:
the treatment was intended to improve coping with symptoms rather than curing
them outright ("care rather than cure"). This had both an empirical and a
conceptual basis. Empirically, clinical experience and intervention trials
for a variety of functional somatic syndromes suggest that the patients who
do best are those who learn to compensate for, rather than attempting to eliminate,
their somatic distress. Conceptually, hypochondriacal somatic symptoms cannot
simply be stripped away with symptomatic treatment because they exist for
underlying psychological and interpersonal reasons. This suggests that a realistic
goal in treating hypochondriasis is amelioration of distressing fears and
beliefs and improved coping, rather than the elimination of somatic symptoms
We are unable to partial out the variance in treatment effect between
the CBT and the physician consultation letter. That the latter may have been
beneficial is suggested by 2 studies with somatization disorder patients in
which a psychiatric consultation letter alone resulted in lower health care
costs along with either improved or stable physical functioning.41- 43 This
points to the critical importance of seamlessly integrating the psychosocial
care of these patients with their medical care.
The treatment offered in this study was not attractive to many hypochondriacal
patients, and only 30% of those eligible entered the trial. Hypochondriacal
individuals are by definition convinced of the medical nature of their condition
and therefore psychosocial treatment seems nonsensical to them. Although a
major problem, this should not detract from the fact that those patients who
did undergo treatment benefited from it. And since hypochondriasis is a prevalent
problem in ambulatory medical practice,6,44 this
fraction of hypochondriacal patients still represents a sizeable population.
The treatment must be made more attractive in the future by seamlessly integrating
it into the primary care process and conducting it in the medical setting
(as our treatment was not). The treatment effect could also be strengthened
by increasing the number of sessions to 8 and by adding booster sessions.