The
Relation
of Ulcerative Colitis to Psychiatric A Review of Findings and Methods Carol
Edward
S. North,
L. Spitznagel,
M.D.,
Ray E. Clouse,
Ph.D.,
and David
findings,
The authors reviewed all known English-language literature on the association between psychiatric factors and ulcerative colitis to ascertain the evidence for such an association and evaluate the methods used in these studies. Most of the 1 38 studies contained serious flaws in research design, such as lack of control subjects, unspecified manner of data collection, and absence of diagnostic criteria. Analysis revealed that methodological flaws were significantly related to the fi nding of a positive association between psychiatric f actors and ulcerative colitis. Seven studies represented solid systematic investigation, and all seven failed to fi nd such an association. (Am J Psychiatry 1990; 147:974-981)
S
ince Murray first observed an association between emotional factors and ulcerative colitis in 1930, a voluminous literature addressing this topic has accumulated. At one time, the debate was whether ulcerative colitis is a psychosomatic condition. Some considered ulcerative colitis to have clear psychosomatic origins (1), while others held it to be a purely physiologic condition (2, 3). Intermediate between these two opposing viewpoints are those who believe that ulcerative colitis has a multifactorial etiology and that emotional aspects are partly involved (4-7), and yet others have suggested that psychological distress in ulcerative colitis may merely be an expected result of any such severe and chronic illness (8-10). Part of the reason for this disagreement is the paucity of objective and replicable findings in published reports on the subject. This in turn is related to serious deficiencies in research methods. The early studies lacked systematic data collection and control or comparison groups, elements considered essential and basic to medical research. This work was largely theoretical and consisted of isolated case reports and anecdotal
Received Aug. 16, 1989; revision received Dec. 1, 1989; accepted Jan. 11, 1990. From the Departments of Psychiatry, Mathematics, and Biostatistics and the Division of Gastroenterology, Washington University School of Medicine. Address reprint requests to Dr. North,
Department
of Psychiatry,
Washington
Medicine, 4940 Audubon Ave., St. Louis, Copyright © 1990 American Psychiatric
974
University
MO 63110. Association.
School
of
M.D.,
H. Alpers,
from
which
colitis
is a psychosomatic
accepted
those improved
when Subsequent causative
Factors:
M.D.
it was
conclusions research research
Later
without methods
efforts
mechanism.
that
concluded
disorder.
challenge became
focused
The
ulcerative
investigators
reader
on
(3), even available.
identifying
a
is encouraged
to
re-
view the earlier literature ( 1 1-2 1) to better understand the evolution of this thinking. With this heritage, ulcerative colitis patients have suffered through the years not only with their disease but also with the stigma of the “psychosomatic” label. In each published article, stated conclusions regarding associations between psychiatric factors and ulcerative
colitis
influence
practicing
clinicians
who
read
it.
Thus, each report, whether or not it reflects methodologically sound work, helps to shape the general opinion on the topic. In our review we examined all known English-language literature pertaining to this association to see if the conclusions were influenced by the methods employed and if the presence of an association should be considered proven.
METHOD
Selection
of Articles
A computer search generated references the papers on psychiatric and psychological ulcerative colitis, and a review of Index ings by
provided searching
more.
Additional
through
the
for
many of aspects of Medicus list-
references
bibliographies
were of
gained
all articles until no
so obtained, and this process was continued further references were uncovered. To be included as a research paper for review in this article, a report had to specify that the subjects had ulcerative colitis (as opposed to unspecified inflammatory bowel disease) because Helzer et al. (22, 23) have documented very different findings regarding psychiatrid factors in ulcerative colitis and Crohn’s disease. The size of the study sample also had to be stated explicitly (as opposed to an unspecified number of patients seen over
years
of
experience).
The
decision
was
made
to
include even small studies because these also had an impact on common medical opinion when they were published. If it became apparent from examination of
Am
J
Psychiatry
147:8,
August
1990
NORTH,
the methods of the studies reviewed that the same patient sample (or a subset) was discussed in two or more articles, then these were considered to represent just one study. In such cases, all information was combined into one entry for this analysis, and the number of patients represented the number of nonduplicated subjects from all articles covering that patient subset. We included an article only if it had been published in English or contained an abstract that had been translated into English and included enough information to meet the criteria. A published abstract not attached to a complete article was not included in this review. A study using a comparison group composed of only subjects with Crohn’s disease was considered uncontrolled for the purposes of this review. A study that compared treated and untreated ulcerative colitis subjects but did not have other control subjects was also considered uncontrolled. We constructed a table that listed all studies and selected findings on the association of ulcerative colitis with life events, personality factors, and types and rates of psychiatric disorders reported in each article. (Space limitations prohibit publication here, but this list is available to interested readers on request.) All available published reviews of articles on this subject were collected by the same methods as those for locating research studies. To be included as a review article for review here, an article had to reference at least three research articles meeting our criteria for inclusion in this study. An article presenting data as well as containing a review, if it met the criteria for inclusion as a study, was listed only with the studies and not with the reviews. An article about ulcerative colitis that contained information on a topic besides ulcerative colitis was included in the review. We constructed a table of review articles and noted the number of studies included in each review, percentage of controlled studies in each review, and the conclusions drawn by the author in regard to the association of ulcerative colitis with life events, personality factors, and psychiatric disorders. (This table is also available to interested readers on request.) Data All
papers were carefully reviewed to determine methods and selected psychiatric conclusions. Then, for a statistical analysis of study conclusions in relation to methods, simple tabulations were performed to determine the number of studies with a particular finding. Studies were segregated according to presence or absence of control groups and the type of control group used (i.e., healthy or medically ill). The control group is so central to sound methods that its absence from a study may be considered sufficient evidence of serious methodological deviation. Because of this, presence or absence of control subjects was the only methodological variable statistically analyzed in this review. Therefore, this computation is merely an example of how methodological factors could affect
SPITZNAGEL,
ET AL.
conclusions and illustrates one facet of our broader methodological critique. A traditional meta-analysis of this collection of artides (using all patients reported) was not performed because the methods varied too greatly from study to study, and segregation of studies with similar methods would yield categories too small for effective analysis. Instead, we analyzed the data by giving equal weight to all reports and elected not to delete studies with tiny samples (e.g., five or fewer subjects). Particularly in the earlier years a study involving only a few subjects was considered legitimate research, since at that time case reports were more readily accepted as mainstream medical
literature.
reader
may
100
The
have
resultant
been
even
in-depth disorders.
into
preeffect.
and
post-1970
publication
past
on
on
from
greater
psychiatric
in the
this
as that
with
emphasis
patients
impression
as great
subjects-perhaps
the
a study
of the
considering
studies of individual Separating articles may help identify
RESULTS
We located 1 72 published research reports in our literature search. Thirty-four of these articles were duplicate investigations of the same patients, leaving a total of 138 actual studies under consideration for this review. One group of seven researchers (24-34) was found to have published 1 1 articles over 35 years on what appeared to be subsets of the same group of patients, which gradually grew over the years to include 136 subjects. Of the 138 studies, 1 17 included only adult subjects, 20
included
both.
only
children,
Twenty-five
patients. 2,000.
and
articles
were
referred Only
for psychiatric 34 of the 138
study
groups used
included
reports
ranged sampled
evaluation studies were
of these used comparison ical illness. The remainder
Methodological
one
case
The number of subjects Thirty-one of the studies
“psychosomatic” drome, and/or
Analysis
CLOUSE,
of single
from only
one to patients
and/or treatment. controlled, and
of subjects with control subjects
illnesses, such as irritable healthy control subjects.
19
medwith
bowel
syn-
Observations
study
Am]
Psychiatry
147:8,
August
1990
Following are the methodological deficiencies evident in the published literature on psychiatric factors in ulcerative colitis. 1. Sampling-small number of subjects; gastrointestinal diagnosis not appropriately confirmed; subjects
with
according selection.
inflammatory
to
bowel
specific
disorders
diagnosis;
not
separated
nonrandom,
biased
2. Control groups-none; not appropriate; matched or compared demographically. 3. Data collection-diagnostic criteria not
used
not
or
specified;
reliability/validity;
instruments
lack
not
of
blind
standardized
not or lack
assessment/assessor
975
ULCERATIVE
COLITIS
bias; data not comparable across studies; chart review inadequacies; retrospective. 4. Data analysis-not done; not described. 5. Conclusions-unwarranted on the basis of available data; erroneous assumption of causation from mere association. The first methodological criticisms relate to deficiencies in sampling. It has previously been recognized (3, 8) that the early reports on psychiatric aspects of ulcerative colitis were particularly prone to involve small samples, the most salient examples being the numerous case reports of one or two individuals in psychotherapy. Such tiny samples defy statistical analysis. Reports on larger samples have been criticized for failing to sufficiently describe (35) or to even include (8) statistical data analysis, instead relying on anecdotal reporting based on claims of experience. Another serious criticism of sampling methods relates to bias. Elsewhere (3) it has been noted that many study
samples
were
composed
entirely
of
subjects
re-
begin by selecting a group of psychiatric patients for a study virtually guarantees the finding of a wealth of psychopathology. To avoid selection bias, a consecutive series of ulcerative colitis patients from a nonpsychiatric source is required. A related problem is the failure to sufficiently describe the method of sample selection or randomization procedure (35, 36). Many early reports were also flawed by failure to describe how the patients were assessed gastroenterologically to confirm the diagnosis of ulcerative colitis, leaving the possibility of impure samples, which may have included patients with irritable bowel syndrome rather than ulcerative colitis. A likely contributor to the assumption that psychological factors have an etiologic role in ulcerative colitis is the possibility of coexistence or confusion of this disorder with irritable bowel syndrome, a condition that has documented psychiatric associations (37, 38). Other authors have suggested that readers conceptualize irritable bowel syndrome and ulcerative colitis as “two aspects or two stages of the same affection” (39). Another possible contributor to the presumption of psychosomatic illness is the known association of cramping and diarrhea with stress in normal subjects. This same association occurs in patients with ulcerative colitis. Some authors of early studies drew conclusions about ulcerative colitis from studies of gastrointestinal pathophysiology in normal individuals. While it may be tempting to attribute symptoms of normal stress responses and irritable bowel syndrome to ulcerative colitis, this logical flaw would direct researchers to unwarranted conclusions. Sampling an active clinic population is another methodological error that is particularly relevant in treatment outcome studies. This problem relates to the tendency for patients to consult physicians when they are ill. The statistical phenomenon of regression to the mean would predict that an individual followed over time from an index point of extreme variation from the ferred
976
for
psychiatric
treatment.
Obviously,
to
norm
(i.e.,
starting
physician’s
at the time
office
for
that
treatment
the
patient
visits
of a flare-up
of
the
gastro-
intestinal symptoms) will likely have fewer symptoms in the future. Regression to the mean will be especially relevant in ulcerative colitis because this illness is characterized by a chronic course with intermittent relapses (40, 41). Therefore, studies sampling symptomatic patients
who
visit
physicians’
offices
would
be
expected
to show a decrease in symptoms over time. This improvement could be mistakenly attributed to the experimental treatment applied (e.g., psychotherapy) unless an effect above and beyond that observed in an experimental control (nontreatment) group can be demonstrated or unless statistical allowance for this factor is achieved during data analysis. A serious but very common methodological fault in studies of psychiatric aspects of ulcerative colitis has been the failure to use proper control subjects or to adequately describe them (35). It has been observed (5, 8) that the early studies generally lacked comparison subjects altogether, although studies without control subjects
continue
to
appear
in
print
even
today
(42-
45). It has been
documented (46) that the lifetime preyalence of psychiatric disorder in the general population is about 33%. Therefore, a methodologically equivalent study of ulcerative colitis patients would be expected to show at least this rate by chance alone. Authors of uncontrolled studies who found a lifetime rate of psychiatric disorder in ulcerative colitis patients of