You are on page 1of 8

Journal of Psychosomatic Research 51 (2001) 589 – 596

Anxiety and surgical recovery


Reinterpreting the literature
Marcus R. Munafò*, Jim Stevenson
University of Southampton, Southampton, UK
Received 12 October 2000; accepted 30 May 2001

Abstract

Objective: To critically evaluate the interpretation of the studies were identified by the search strategy, met the inclusion
findings reported in the peer-reviewed literature concerning the criteria and contributed to the review. Conclusions: Associations
association of state and trait anxiety with surgical recovery and between preoperative measures of anxiety and postoperative mood
response to surgery. Methods: The Social Science Citation Index and pain have been consistently reported. Associations with regard
(SSCI), Science Citation Index (SCI), Medline and Psychological to other recovery variables are less consistent. The existing
Abstracts (PsycInfo) databases were searched for studies published evidence does not rule out an interpretation of the results as
since 1981. Reference lists from previous reviews were also reflecting consistent self-reporting bias rather than causal associ-
searched for additional references. Studies that were not in the ation. D 2001 Elsevier Science Inc. All rights reserved.
public domain were not searched for. Results: Twenty-seven

Keywords: Anxiety; State – Trait Anxiety Inventory; Surgery; Surgical patients; Surgical recovery; Prediction; Self-report

Introduction factors such as current (or state) anxiety and depression


might play some part in determining the duration and
An extensive literature exists that attempts to show quality of the recovery period.
relationships between psychological factors, usually meas- Janis [15] proposed that a curvilinear relationship exists
ured preoperatively, and surgical recovery. The main psy- between preoperative ‘‘distress’’ (not necessarily analogous
chological factors of interest are affective, with a great deal to state anxiety) and postoperative recovery. That is, mod-
of emphasis on anxiety, as this is a particularly salient erate levels of ‘‘distress’’ are proposed to result in optimal
feature of the perioperative period [15]. The ability to postoperative recovery, while excessively low or exces-
predict surgical recovery would be of both clinical and sively high levels of ‘‘distress’’ both result in impaired or
theoretical value. suboptimal recovery, although the mechanisms underlying
Surgery represents a major trauma that provokes a these relationships may differ. This moderate level of
relatively stereotyped physiological response [32]. A period ‘‘distress’’ related to optimal postoperative recovery was
of postoperative recovery follows, ranging from a few days termed, by Janis, ‘‘the work of worry,’’ which was hypo-
to several weeks depending, primarily, on the severity of the thesised to prepare the surgical patient for the distress and
preexisting pathology and the surgical procedure performed. suffering associated with postoperative recovery. An in-
There is also variability across patients who have undergone ability to prepare oneself appropriately, reflected in low
a comparable surgical procedure. What accounts for this preoperative ‘‘distress,’’ was suggested to result in greater
difference across patients is unclear, and it is this difference distress postoperatively, for example, because of the shock
that led to the suggestion (e.g., Ref. [15]) that psychological of such unexpected pain, while excessive preoperative
‘‘distress’’ was also supposed to result in greater postoper-
ative suffering, perhaps because of sensitization to noxious
* Corresponding author. ICRF GPRG, Institute of Health Sciences,
University of Oxford, Oxford OX3 7LF, UK. Tel.: +44-1865-226756; fax:
stimuli. More recently, experimental work in psychoneu-
+44-1865-227137. roimmunology has suggested that stress delays wound
E-mail address: marcus.munafo@dphpc.ox.ac.uk (M.R. Munafò). healing and, in addition, that pain has adverse effects on

0022-3999/01/$ – see front matter D 2001 Elsevier Science Inc. All rights reserved.
PII: S 0 0 2 2 - 3 9 9 9 ( 0 1 ) 0 0 2 5 8 - 6
590 M.R. Munafò, J. Stevenson / Journal of Psychosomatic Research 51 (2001) 589–596

endocrine and immune function [18]. Taken together, this attempt to relate this to a constellation of postoperative
suggests the possibility of both biologically and behavior- factors that may be regarded as making up the global
ally mediated associations between preoperative state anxi- concept of postoperative distress and suffering, including
ety and postoperative state. An obvious question that results state anxiety, pain, analgesia consumption, time to discharge
from this thesis is whether ‘‘distress’’ as envisaged by Janis and so on. Consequently, for any review to be able to make
necessarily reflects the same thing as ‘‘stress’’ as inves- consistent comparisons across studies, it is important to state
tigated by Kiecolt – Glaser et al. [18] and ‘‘anxiety’’ as used a particular measure of anxiety that will be used in the
by others. This is tacitly assumed in the literature that has selection of studies. This could potentially include one of
followed, but the basis for this assumption is not clear. the physiological correlates of anxiety, such as salivary
There are a number of physiological, cognitive and cortisol, or one of the several questionnaire measures of
behavioral correlates of anxiety that could conceivably be anxiety. Behavioral measures, such as information seeking
introduced as measures of anxiety [27] if one accepts a or coping behavior, might also be regarded as important.
strong association with subjective anxiety as an indication In the case of this review, only those studies that report
of validity. These may be more closely related to the associations between self-reported state and trait anxiety as
mechanisms that underlie and drive anxiety, but may also measured by the State – Trait Anxiety Inventory (STAI) [36]
be removed from the subjective experience that is relevant and postoperative outcome are included. A large proportion
to the individual. Nevertheless, such an approach would of studies have used this measure of current and disposi-
strengthen the claim that the association between perioper- tional anxiety as both a predictor and outcome measure, and
ative anxiety and recovery is causal and potentially amen- the inclusion of only those studies that use this measure will
able to intervention. simplify the comparison of results across studies. As already
If elevated state anxiety, for example, is reliably shown to mentioned, state anxiety as measured by questionnaires may
predict slower wound healing [18], then it might be hypo- reflect something quite different to the ‘‘worry’’ envisaged
thesized that this is due to the endocrinological and auto- by Janis [30], and this should be borne in mind when
nomic changes associated with elevated state anxiety, and interpreting the results of studies.
their subsequent impact on the wound healing process. The The aim of this paper, therefore, is to critically evaluate
prediction of surgical recovery provides a testing ground for the interpretation of the findings reported in the peer-
hypotheses regarding the relationships that obtain between reviewed literature concerning the association of state and
psychological and behavioral indices and physical health. trait anxiety as measured by the STAI with surgical recovery
Prediction allows for the subsequent testing of hypotheses and response to surgery, and to try to evaluate the validity
regarding causation. and parsimony of these interpretations.
Effective and reliable prediction of surgical recovery
might also provide the potential to facilitate recovery in
‘‘high-risk’’ surgical patients if interventions to modify the Methods
predictor variables thought to be causal were developed.
Indeed, early evidence that this might be possible (e.g., Ref. The review was based upon the Social Science Citation
[13]) has motivated much of the research that has followed. Index (SSCI), Science Citation Index (SCI), Medline and
This would be of clinical benefit given that slower recovery, Psychological Abstracts (PsycInfo) databases. These were
inactivity and so on is strongly associated with subsequent searched for studies published since 1981, the start date of
morbidity, impairment of muscle function and elevated risk the SCI and SSCI, that had the words ‘‘anxiety’’ and
of complication (e.g., deep vein thrombosis) [32]. It is ‘‘surgical’’ or ‘‘surgery’’ or ‘‘patient’’ in the title, abstract
important to make a distinction between clinical and or keywords. In addition, the reference lists from previous
statistical significance: While theoretical models require reviews (e.g., Refs. [18,32]) were searched for additional
tests of statistical significance, this is only of importance references. Studies that were not in the public domain were
in a clinical context in conjunction with evidence for the not searched for.
efficacy of a clinically significant effect. Elevated state Studies generated by the search strategy were checked
anxiety might predict some aspect of surgical recovery to for relevance. Observational studies that used the STAI as
a statistically significant degree, but if this effect is weak an index of current (i.e., state) anxiety and/or dispositional
(i.e., the effect size is small) then the effort and cost (i.e., trait) anxiety prior to surgery were included. Only
associated with intervention might not be justified by the studies that recruited adults (aged 16 years or older) admit-
beneficial impact of such an intervention. In such an ted for inpatient elective surgical procedures were included.
instance the relationship would not be regarded as clinically Intervention studies were not included to concentrate on the
significant. This would also be the case if the risk factor naturally occurring linkages between perioperative anxiety
was not amenable to intervention. and recovery. For each study selected for inclusion, the
The majority of studies that investigate the role of affect following data were extracted:
in surgical recovery use self-report measures of generic state
anxiety as an index of preoperative distress. They then (i) author(s) and year
M.R. Munafò, J. Stevenson / Journal of Psychosomatic Research 51 (2001) 589–596 591

Table 1
Measures of mood
Preoperative Postoperative
Study N Sex Age Type of surgery measures measures Effect
Boeke et al. [2] 111 (93%) M/F 19 – 86 Gallbladder STAI S STAI S r =.30***
Christensen et al. [5] 15 M/F 42 – 64 Abdominal STAI S STAI S Decrease*
De Groot et al. [7] 126 (93%) M/F 20 – 77 Lumbar STAI S STAI S b =.31*
De Groot et al. [9] 126 (93%) M/F 20 – 77 Lumbar STAI S STAI S b =.38***
Duits et al. [11] 217 (71%) M/F 28 – 78 Coronary Bypass STAI S STAI S Decrease***
Duits et al. [12] 217 (71%) M/F 28 – 78 Coronary Bypass STAI S STAI S r =.25***
Jamison et al. [14] 50 (96%) F 18 – 50 Laparoscopic STAI S Psychological reaction r =.39**
STAI T r =.35*
Johnston and Carpenter [16] 73 F 23 – 79 Gynaecological (major) STAI S STAI S F = 9.69**
Kain et al. [17] 53 F 26 – 56 Gynaecological (major) STAI S STAI S r =.47** (1-tailed)
STAI T r =.39** (1-tailed)
Kugler et al. [19] 20 (69%) M/F 19 – 63 Heart Transplant STAI S STAI S r =.11 ns
Manyande and Salmon [23] 40 (75%) M/F 17 – 67 Abdominal STAI S STAI S r =.62**
Martinez-Urrutia [25] 59 M < 65 Not stated STAI S STAI S Decreasea
STAI T + ve assoc.***
Salmon et al. [33] 17 M/F 41 (mean) Middle ear STAI S STAI S Decrease**
Taenzer et al. [38] 40 M/F 20 – 65 Gallbladder STAI S STAI S r =.65***
STAI T r =.53***
Timberlake et al. [39] 121 M/F 35 – 75 Coronary artery bypass STAI S Depression t = 2.33*
STAI T t = 2.16*
Wallace [42] 120 (92%) F Not stated Gynaecological (minor) STAI S STAI S F = 5.5*
Wallace [43] 118 F Not stated Laparoscopic STAI T STAI S r =.22**
a
P not stated.
* P < .05.
** P < .01.
*** P < .001.

Table 2
Measures of pain
Preoperative Postoperative
Study N Sex Age Type of surgery measures measures Effect
Bachiocco et al. [1] 114 (95%) M/F 17 – 78 Thoracic STAI S Analgesia request + ve assoc.**
STAI T + ve assoc.**
Boeke et al. [2] 111 (93%) M/F 19 – 86 Gallbladder STAI S Pain Likert r =.20*
De Groot et al. [8] 126 (93%) M/F 20 – 77 Lumbar STAI S Pain Visual r =.21*
Analogue Scale
Jamison et al. [14] 50 (96%) F 18 – 45 Laparoscopic STAI S Pain Likert r =.46****
STAI T r =.38**
Johnston and Carpenter [16] 73 F 23 – 79 Gynaecological (major) STAI S Pain Visual F = 0.12a
Analogue Scale
Kain et al. [17] 53 F 26 – 56 Gynaecological (major) STAI S Pain Visual r =.29**
STAI T Analogue Scale r =.09a

Manyande and Salmon [23] 40 (75%) M/F 17 – 67 Abdominal (minor) STAI S Pain Visual r =.43**
Analogue Scale
Martinez-Urrutia [25] 59 M Not stated Not stated STAI S Pain questionnaireb r =.27*
STAI T F = 12.31***
Nelson et al. [28] 96 M/F 37 – 84 Coronary artery bypass STAI S Pain questionnaireb r =.41***
STAI T r =.01a
Scott et al. [34] 48 M/F 21 – 73 Gallbladder STAI S Pain questionnairec r =.26*
STAI T r =.37 *
Taenzer et al. [38] 40 M/F 20 – 65 Gallbladder STAI T Pain questionnairec r =.33*
Wallace [41] 118 F Not stated Laparoscopic gynaecological STAI S Pain Expectancy r =.18*
a
P nonsignificant.
b
Melzack and Torgesen Pain Questionnaire.
c
McGill Pain Questionnaire.
* P < .05.
** P < .01.
*** P < .001.
**** P < .0001.
592 M.R. Munafò, J. Stevenson / Journal of Psychosomatic Research 51 (2001) 589–596

Table 3
Measures of physical recovery and behaviour
Study N Sex Age Type of surgery Preoperative measures Postoperative measures Effect
Boeke et al. [2] 111 (93%) M/F 19 – 86 Gallbladder STAI S Length of stay b = .29**
Boeke et al. [3] 58 (95%) M/F 19 – 78 Gallbladder STAI S Length of stay r = .05a
STAI T r = .10a
Boeke et al. [4] 81 M/F Not stated Gallbladder STAI S Length of stay r =.05a
Christensen et al. [5] 15 M/F 42 – 64 Abdominal STAI S change Fatigue Likert change r =.65**
STAI T change r =.38a
De Groot et al. [7] 126 (93%) M/F 20 – 77 Lumbar STAI S Symptom Checklist 90 b =.27**
De Groot et al. [8] 126 (93%) M/F 20 – 77 Lumbar STAI S Physician rated outcome r =.24**
De Groot et al. [9] 126 (93%) M/F 20 – 77 Lumbar STAI S Symptom Checklist 90 b =.28**
Jamison et al. [14] 50 (96%) F 18 – 45 Laparoscopic STAI S Physical state r =.46****
STAI T r =.38**
Johnston and Carpenter [16] 73 F 23 – 79 Gynaecological (major) STAI S Wolfer – Davis scale F = 0.76a
Manyande and Salmon [23] 40 (75%) M/F 17 – 67 Abdominal (minor) STAI S Wolfer – Davis scale r = .05a
STAI T r = .07a
Salmon et al. [33] 17 M/F Not stated Middle ear STAI S Wolfer – Davis scale r = .86**
Simpson and Kellett [35] 45 (94%) M/F > 60 Orthopaedic STAI S Delirium change r = .27a
STAI T r = .15a
a
Stengrevics et al. [37] 94 (82%) M/F Not stated Cardiac STAI S Length of stay
a
STAI T
a
P nonsignificant.
** P < .01.
**** P < .0001.

(ii) number of participants and recruitment rate Table 1). Twelve studies contributed to the review of meas-
(iii) age range of participants ures of pain (see Table 2). Thirteen studies contributed to the
(iv) type of surgery review of measures of physical recovery and behavior (see
(v) preoperative measures Table 3), and two studies contributed to the review of
(vi) postoperative measures measures of physiological correlates (see Table 4).
(vii) effect/relationship Six studies did not specify the age range of participants
[4,25,33,37,41,42]. Most studies included participants
This information was recorded in tabular form, grouped undergoing abdominal, orthopaedic, gynaecological or cor-
by the type of postoperative measure under consideration. onary bypass surgical procedures, with the exceptions of
This included outcome measures of: Bachiocco et al. [1] (thoracic), Kugler et al. [19] (heart
transplant) and Salmon et al. [33] (middle ear). One study
(i) mood [25] did not specify the surgical procedure or procedures
(ii) pain undergone by the participants. Six studies necessarily
(iii) physical recovery/behavior (other than pain) included female participants only because the surgical group
(iv) physiological correlates was undergoing gynaecological procedures [14,16,17,
42,43] and one study included male participants only [25].
Twenty-seven studies conducted in the United States, the No studies reported the basis for sample size selection or
United Kingdom, The Netherlands and Denmark between provided a power analysis, and recruitment rates (i.e., the
1975 and 2000 were identified by the search strategy, met the proportion of those approached who agreed to take part in
inclusion criteria and contributed to the review. Seventeen the study) were also not recorded in several studies
studies contributed to the review of measures of mood (see [4,5,16,17,25,28,33,34,38,39,41,43].

Table 4
Measures of physiological correlates of recovery
Study N Sex Age Type of surgery Preoperative measures Postoperative measures Effect
Salmon et al. [33] 17 M/F Not stated Middle ear STAI S Cortisol r =.02a
STAI S Noradrenaline r =.05a
Salmon et al. [29] 27 M/F 18 – 77 Abdominal (major) STAI T Cortisol r = .55*
STAI T Adrenaline r = .67*
STAI T Noradrenaline r =.53*
a
P nonsignificant.
* P < .05.
M.R. Munafò, J. Stevenson / Journal of Psychosomatic Research 51 (2001) 589–596 593

Results Pain

The studies included form the basis of the review pre- Twelve studies reported on the relationship between
sented below, grouped by the type of outcome measure used. preoperative levels of anxiety and postoperative pain.
The majority of studies used either a Likert scale or a
Mood visual analogue scale to assess postoperative pain. Four
studies [25,28,34,38] used a pain questionnaire, in the first
Seventeen studies reported on the relationship between case the Melzack and Torgesen Pain Questionnaire and in
preoperative levels of anxiety and postoperative mood. the other three cases the McGill Pain Questionnaire. One
All studies used the state subscale of the STAI as an study [41] used a self-report measure of pain expectancy
index of postoperative mood, with the exception of two (assessed prior to surgery), and one study [1] used a
studies [14,39] that used self-report measures of ‘‘psycho- measure of number of analgesia requests. Seven studies
logical reaction’’ and depression, respectively. Six studies reported on the relationship between trait anxiety, meas-
reported on the relationship between trait anxiety, meas- ured preoperatively, and postoperative pain [1,14,17,25,28,
ured preoperatively, and postoperative mood [14,17,25,38, 34,38].
39,43]. Two primary associations emerge from a review of these
Three primary associations emerge from a review of studies. The first is that self-reported preoperative state
these studies. Eleven studies report that self-reported anxiety is positively correlated with self-reported postoper-
preoperative state anxiety is positively correlated with ative pain. Nine studies report correlation coefficients
postoperative state anxiety or an equivalent (‘‘psycho- ranging from .18 to .46. One study [16] reports non-
logical reaction’’ in the case of Jamison et al. [14] and significant results for this association, although this could
depression in the case of Timberlake et al. [39]. Reported simply be a function of small sample size. One study also
correlation coefficients range from .25 to .65, with two reports a significant positive association between preoper-
studies reporting beta coefficients, two studies reporting F ative state anxiety and postoperative analgesia request [1]
statistics and one study reporting t statistics. Five studies without stating the test statistic. The second is that self-
report a second association that self-reported preoperative reported preoperative trait anxiety is also positively corre-
trait anxiety is positively correlated with postoperative lated with self-reported postoperative pain. In this case four
state anxiety or an equivalent. In this case the range of studies report correlation coefficients of between .33 and
reported correlation coefficients is from .22 to .53, with .38 and one study reports an F statistic of 12.31. One study
one study reporting t statistics. Finally, four studies report [1] also reports a significant, positive association without
a significant decrease in self-reported state anxiety over stating the test statistic. Two studies [17,28] report non-
the perioperative period. One study reported a nonsigni- significant correlations.
ficant relationship between preoperative and postoperative In all cases the only measure of pain was taken post-
state anxiety. operatively. For some surgical procedures this is necessary
The majority of studies report correlation coefficients, given the absence of preoperative pain. However, it should
from which it is possible to calculate an effect size index. In be noted that there is, therefore, no control for baseline
the case of the association between preoperative state levels of expression of pain behaviors.
anxiety and postoperative mood the effect size indices range The majority of studies report correlation coefficients,
from .25 to .65, which indicates a moderate to large effect from which it is possible to calculate an effect size index.
size [6]. Three studies report F statistics, of which one In the case of the association between preoperative state
reports sufficient data to calculate an effect size index (.30), anxiety and postoperative pain the effect size indices
which also indicates a moderate effect size [6]. One study range from .18 to .41, which indicates a moderate effect
reports t statistics, from which an effect size of .46 can be size [6] in those studies reporting statistically significant
calculated, which indicates a moderate effect size [6]. A results. One study that reports a significant association
further four studies that report statistically significant results between preoperative state anxiety and postoperative pain
do not report statistics that allow the calculation of an effect does not report statistics that allow the calculation of an
size index. effect size index.
In the case of the association between trait anxiety and In the case of the association between trait anxiety
postoperative mood the effect size indices for the reported and postoperative pain the effect size indices for the two
correlation coefficients range from .22 to .53, which studies for which it is possible to calculate an effect size
indicates a moderate to large effect size [6]. One study index are .37 and .38, which indicates a moderate effect
reports t statistics, from which an effect size of .39 can be size [6] in those studies reporting statistically significant
calculated, which indicates a moderate effect size [6]. One results. Two studies that report a significant association
study that reports a statistically significant association between trait anxiety and postoperative pain do not
does not report statistics that allow the calculation of an report statistics that allow the calculation of an effect
effect size index. size index.
594 M.R. Munafò, J. Stevenson / Journal of Psychosomatic Research 51 (2001) 589–596

Physical state and behavior of self-reported preoperative trait anxiety, however, cor-
relation coefficients ranging in magnitude from .53 to .67
Thirteen studies reported on the relationship between were reported for cortisol, adrenaline and noradrenaline
preoperative levels of anxiety and postoperative physical (negative in the case of the first two and positive in the
recovery and behavior. Four studies [2– 4,37] used length case of the last).
of inpatient stay as an index of physical recovery. Five The correlation coefficients reported for the association
studies [7,9,16,23,33] used questionnaire measures of post- between trait anxiety and postoperative physiological meas-
operative physical state. Other measures used included ures result in effect size indices ranging from .73 to .82,
self-reported fatigue [5], physician-rated physical state which indicates a large effect size [6].
[8], self-reported physical state [14] and change in delirium
as measured by the Mini-Mental State Examination [35].
Six studies reported on the relationship between trait Conclusions
anxiety, measured preoperatively, and postoperative phys-
ical recovery [3,5,14,23,35,37]. This review has shown that preoperative state anxiety has
A review of these studies produces the least clear picture been reported to be associated with postoperative mood and
of the associations between anxiety and postoperative out- pain in a number of studies with some consistency, and to
come. Six studies report only nonsignificant results, while other recovery variables in a smaller number of studies with
another reports some nonsignificant results. Among those less consistency, with a number of studies reporting no
that report significant results, the only outcome measure significant associations.
consistently reported to be associated with self-reported All studies included in this review report linear relation-
state anxiety is the Symptom Checklist 90, with beta ships, with very few actually having included a test for
coefficients of .27 and .28 being reported by two studies. curvilinear relationships, with a few exceptions (e.g., Refs.
A closer inspection of these two papers [7,9], however, [16,42]). In the cases where a test for a curvilinear relation-
suggests that these results might be drawn from a single data ship was included there was no evidence to support such an
set. One study reports a significant beta coefficient for the association. This failure to support Janis’s original hypo-
association between state anxiety and length of stay. One thesis does not mean that the hypothesis is false. In
study reports a significant, positive correlation between state particular, the difficulties associated with self-report indices
anxiety and physician rated outcome. One study reports of state anxiety (see below) and the fact that excessively low
significant, positive correlations between state and trait state anxiety may be both rare and difficult to assess
anxiety and self-reported physical state, and one study accurately may prevent an adequate test being made.
reports a significant, negative correlation between state One shortcoming of research of the kind reviewed above
anxiety and self-reported physical state as measured by that is apparent is the lack of comparability across studies.
the Wolfer –Davis Scale. Of the four studies reporting data Even when studies to be included are restricted to those
on length of inpatient stay, three report nonsignificant employing a single preoperative questionnaire (in this case
associations with preoperative state anxiety. the STAI) there is substantial variation in the type of
Given the inconsistent pattern of results that emerges surgical procedure investigated, the sex and age range of
from an analysis of the studies presented here, it would be participants included, the postoperative measure of recovery
premature to estimate the size of effect of any reported or postoperative state employed, and the statistical analysis
relationships. Much of this inconsistency stems from the use performed. There is also the perennial problem of unreport-
of a variety of postoperative measures of physical state ed nonsignificant findings, which suggests that any rela-
across studies. tionship may be even weaker still.
A more significant problem, however, is the extent to
Physiological correlates which the potential confound of behavioral consistency is
not controlled for in the studies reviewed. For example,
Two studies reported on the relationship between without a baseline index of pain behavior and pain expres-
preoperative levels of anxiety and physiological correlates. sion, any differences in postoperative pain report are
Both studies reported on the associations between anxiety difficult to interpret. Conversely, the fact that preoperative
and levels of cortisol and noradrenaline [29,33]. One and postoperative state anxiety are frequently reported to be
study [29] also reported on the associations between positively correlated may reflect nothing more than behav-
anxiety and adrenaline. One study assessed state anxiety ioral consistency. In this case the behaviors are anxiety-
preoperatively [33], while the other [29] assessed trait related behaviors, which may include responses to items on
anxiety preoperatively. a questionnaire measure. It is important to bear in mind that
The lack of studies makes broad generalizations dif- most of the preoperative and postoperative measures
ficult, but in the case of self-reported preoperative state reported in fact measure a particular behavior or group of
anxiety there were no significant associations with physio- behaviors, and that behaviors tend to show consistency
logical measures (cortisol and noradrenaline). In the case over time.
M.R. Munafò, J. Stevenson / Journal of Psychosomatic Research 51 (2001) 589–596 595

It would seem that there is no reason at present to debatable whether this will address the problem of behav-
interpret the data reported in the studies reviewed here as ioral consistency, since such assessment will inevitably
representing anything more than behavioral consistency. depend on certain aspects of the behavior of the patient.
This may be largely (although not necessarily exclusively) The associations reported between state and trait anxiety
self-reporting consistency, since most measures of post- and postoperative mood and pain suggest that anxiety is a
operative outcome rely on self-report. In a given context potentially important predictor of recovery. However, the
specific behaviors, such as self-report, will be systematic- evidence as it stands is largely consistent with an interpreta-
ally interpreted in a given way, so that even ostensibly tion in terms of consistency in self-reporting. One advance
objective measures of postoperative state, such as time to would be to include measures of state and trait anxiety that
discharge, will be influenced by the self-reported state of are not dependent on self-report. This circumvents the
the patient. This is, in fact, an interesting finding in itself, possible confound of reporting bias and consistency, but it
but the mechanisms that underlie the association between does introduce novel difficulties, such as whether the index
preoperative anxiety and postoperative recovery might of anxiety chosen (such as cortisol) reflects subjective
therefore be quite different to those suggested in many anxiety as felt by the patient.
of the studies that have been reviewed. It is revealing that In addition, given the potential for confounding one
intervention studies have at best resulted in modest bene- might hope for a greater degree of consistency in the
fits for those given an intervention designed to reduce measures used, the population studied and so on. In practice,
anxiety and subsequently improve recovery (e.g., Refs. however, this is likely to be difficult to achieve. It would also
[20,22,26,40]). be desirable to test for nonlinear as well as linear associa-
The only postoperative recovery measures where an tions. The evidence that does exist suggests that linear
explanation in terms of self-reporting consistency would models fit the data better than nonlinear models, but it would
be difficult are physiological correlates of emotional state be desirable to have more data to support this position.
and other physiological measures, such as wound healing. Finally, it should be borne in mind that anxiety is not
Unfortunately, the number of studies that employ such something that exists in isolation but rather operates in a
measures is very limited, and the data that are reported given context and serves to modify an individual’s behavior
suggest that trait anxiety is a more reliable predictor of such and physiology in an adaptive way. There are, therefore, a
outcome measures than state anxiety. If this is the case it is number of related psychological characteristics, such as
unlikely that an intervention designed to ameliorate state coping behaviors, which may be relevant in understanding
anxiety will be successful. Some studies have investigated any associations between anxiety and recovery. Some studies
the relationship between preoperative state anxiety and do report data on these characteristics (e.g., Refs. [10,21]),
intraoperative measures, such as analgesia requirement but often in the context of an intervention design where a
(e.g., Ref. [24]), but it is not clear how this relates to particular coping style is manipulated as part of an interven-
postoperative state, although such variables are certainly tion designed to improve recovery. A clearer understanding
of interest as mediators of recovery. of the associations that obtain in surgical patients between
Given that behavioral consistency and reporting bias is a anxiety, coping and recovery would guide the development
potential confound, efforts to measure and control this would of subsequent intervention studies, which have so far been
be of value. Unfortunately, the measures that might be used comparatively unsuccessful in improving recovery.
to assess baseline expression of particular behaviors are the The role of anxiety in surgical recovery remains unclear
same as those currently used as preoperative predictors of but potentially both theoretically revealing and clinically
outcome. Measures of change, as opposed to absolute values important. For further advances to be made in our under-
of, say, state anxiety might go some way to controlling for standing to be made, however, what will be required will be
this potential confound. However, this will be limited by that an appreciation of the limitations of the research that has
impossibility of measuring baseline levels of, say, pain gone before and an attempt to address these. There are two
behaviors in the case of the majority of surgical populations. questions that result from this review: One is what is
Another alternative would be to rely on assessments of required to improve the quality of research, while another
preoperative and postoperative variables by others (e.g., is whether there are any psychological characteristics other
nursing staff ). Unfortunately, this is equally liable to con- than anxiety that deserve greater attention.
founding. For example, Salmon and Manyande [31] report
evidence that nursing staff systematically underestimate how
well patients feel they are coping with their pain. Moreover, References
the perception by nursing staff that a patient was coping
poorly was associated with negative evaluation in other [1] Bachiocco V, Rucci P, Carli G. Request of analgesics in post-surgical
areas. Including both self-report measures and assessments pain. Relationships to psychological factors and pain-related variables.
Pain Clin 1996;9:169 – 79.
made by others may go some way to countering this problem [2] Boeke S, Duivenvoorden HJ, Verhage F, Zwaveling A. Prediction of
of confounding, but this increase in the number of variables postoperative pain and duration of hospitalization using two anxiety
measured will require an increase in sample sizes, and it is measures. Pain 1991;45:293 – 7.
596 M.R. Munafò, J. Stevenson / Journal of Psychosomatic Research 51 (2001) 589–596

[3] Boeke S, Stronks D, Verhage F, Zwaveling A. Psychological variables [22] Manyande A, Salmon P. Effects of pre-operative relaxation on post-
as predictors of the length of post-operative hospitalization. J Psycho- operative analgesia: immediate increase and delayed reduction. Br J
som Res 1991;35:281 – 8. Health Psychol 1998;3:215 – 24.
[4] Boeke S, Jelicic M, Bonke B. Pre-operative anxiety variables as pos- [23] Manyande A, Salmon P. Recovery from minor abdominal surgery: a
sible predictors of post-operative stay in hospital. Br J Clin Psychol preliminary attempt to separate anxiety and coping. Br J Clin Psychol
1992;31:366 – 8. 1992;31:227 – 37.
[5] Christensen T, Hjortso NC, Mortensen E, Riis-Hansen M. Fatigue and [24] Maranets I, Kain ZN. Preoperative anxiety and intraoperative anes-
anxiety in surgical patients. Acta Psychiatr Scand 1986;73:76 – 9. thetic requirements. Anesth Analg 1999;89:1346 – 51.
[6] Cohen J. A power primer. Psychol Bull 1992;112:155 – 9. [25] Martinez-Urrutia A. Anxiety and pain in surgical patients. J Consult
[7] De Groot KI, Boeke S, Duivenvoorden HJ, Bonke B. Different as- Clin Psychol 1975;43:437 – 42.
pects of anxiety as predictors of post-operative anxiety and physical [26] Mavrias R, Peck C, Coleman G. The timing of pre-operative prepar-
complaints. Pers Individ Differ 1996;21:929 – 36. atory information. Psychol Health 1990;5:39 – 45.
[8] De Groot KI, Boeke S, van den Berge HJ, Duivenvoorden H, Bonke B, [27] Munafò MR. Perioperative anxiety and postoperative pain. Psychol,
Passchier JP. Assessing short- and long-term recovery from lumbar Health Med 1998;3:429 – 33.
surgery with pre-operative biographical, medical and psychological [28] Nelson FV, Zimmerman L, Barnason S, Nieveen J, Schmaderer M.
variables. Br J Health Psychol 1997;2:229 – 43. The relationship and influence of anxiety on postoperative pain in the
[9] De Groot KI, Boeke S, van den Berge HJ, Duivenvoorden HJ. The coronary artery bypass graft patient. J Pain Symptom Manage 1998;
influence of psychological variables on postoperative anxiety and 15:102 – 9.
physical complaints in patients undergoing lumbar surgery. Pain [29] Salmon P, Pearce S, Smith CC, Manyande A. Anxiety, Type A person-
1997;69:19 – 25. ality and endocrine responses to surgery. Br J Clin Psychol 1989;
[10] De Groot KI, Boeke S, Bonke B, Passchier J. A revaluation of the 28:279 – 80.
adaptiveness of avoidant and vigilant coping with surgery. Psychol [30] Salmon P. Psychological factors in surgical recovery. In: Gibson HB,
Health 1997;12:711 – 7. editor. Psychology, pain and anaesthesia. 1st ed. vol. 10. London:
[11] Duits AA, Duivenvoorden HJ, Boeke S, Taams MA, Mochtar B, Chapman & Hall, 1994. pp. 229 – 58.
Krauss XH, Passchier JP, Erdman R-AM. The course of anxiety and [31] Salmon P, Manyande A. Good patients cope with their pain: post-
depression in patients undergoing coronary artery bypass graft sur- operative analgesia and nurses’ perceptions of their patients’ pain.
gery. J Psychosom Res 1998;45:127 – 38. Pain 1996;68:63 – 8.
[12] Duits AA, Duivenvoorden HJ, Boeke S, Taams MA, Mochtar B, [32] Salmon P, Hall GM. A theory of postoperative fatigue: an interaction
Krauss XH, Passchier J, Erdman R-AM. A structural modeling analy- of biological, psychological, and social processes. Pharmacol, Bio-
sis of anxiety and depression in patients undergoing coronary artery chem Behav 1997;56:623 – 8.
bypass graft surgery: a model generating approach. J Psychosom Res [33] Salmon P, Evans R, Humphrey DE. Anxiety and endocrine changes in
1999;46:187 – 200. surgical patients. Br J Clin Psychol 1986;25:135 – 41.
[13] Egbert LD, Battit GE, Welch CE, Bartlett MK. Reduction of post- [34] Scott LE, Clum GA, Peoples JB. Preoperative predictors of postoper-
operative pain by encouragement and instruction of patients. N Engl J ative pain. Pain 1983;15:283 – 93.
Med 1964;270:825 – 7. [35] Simpson CJ, Kellett JM. The relationship between pre-operative anxi-
[14] Jamison R, Parris WCV, Mazson WS. Psychological factors influenc- ety and post-operative delirium. J Psychosom Res 1987;31:491 – 7.
ing recovery from outpatient surgery. Behav Res Ther 1987;25:31 – 7. [36] Spielberger CD, Gorusch RL, Lushene RE. State – trait anxiety inven-
[15] Janis IL. Psychological Stress. 1st ed. New York: Academic tory manual. Palo Alto: Consulting Psychologists Press, 1970.
Press, 1958. [37] Stengrevics S, Sirois C, Schwartz CE, Friedman R. The prediction of
[16] Johnston M, Carpenter L. Relationship between pre-operative anxiety cardiac surgery outcome based upon preoperative psychological fac-
and post-operative state. Psychol Med 1980;10:361 – 7. tors. Psychol Health 1996;11:471 – 7.
[17] Kain ZN, Sevarino F, Alexander GM, Pincus S, Mayes LC. Preop- [38] Taenzer P, Melzack R, Jeans ME. Influence of psychological factors
erative anxiety and postoperative pain in women undergoing hyster- on postoperative pain, mood and analgesic requirements. Pain
ectomy: a repeated-measures design. J Psychosom Res 2000;49: 1986;24:331 – 42.
417 – 22. [39] Timberlake N, Klinger L, Smith P, Venn G, Treasure T, Harrison M,
[18] Kiecolt-Glaser JK, Page GG, Marucha PT, MacCallum RC, Glaser R. Newman SP. Incidence and patterns of depression following coronary
Psychological influences on surgical recovery: perspectives from psy- artery bypass graft surgery. J Psychosom Res 1997;43:197 – 207.
choneuroimmunology. Am Psychol 1998;53:1209 – 18. [40] Wallace LM. Psychological preparation as a method of reducing the
[19] Kugler J, Tenderich G, Stahlhut P, Poisval H, Korner MM, Korfer R, stress of surgery. J Hum Stress 1984;10:62 – 77.
Kruskemper GM. Emotional adjustment and perceived locus of con- [41] Wallace LM. Surgical patients’ expectations of pain and discomfort:
trol in heart transplant patients. J Psychosom Res 1994;38:403 – 8. does accuracy of expectations minimise post-surgical pain and dis-
[20] Manyande A, Chayen S, Priyakumar P, Smith CC. Anxiety and tress? Pain 1985;22:363 – 73.
endocrine responses to surgery: paradoxical effects of preoperative [42] Wallace LM. Pre-operative state anxiety as a mediator of psycholog-
relaxation training. Psychosom Med 1992;54:275 – 87. ical adjustment to and recovery from surgery. Br J Med Psychol
[21] Manyande A, Berg S, Gettins D, Stanford SC. Preoperative rehearsal 1986;59:253 – 61.
of active coping imagery influences subjective and hormonal re- [43] Wallace LM. Trait anxiety as a predictor of adjustment to and recovery
sponses to abdominal surgery. Psychosom Med 1995;57:177 – 82. from surgery. Br J Clin Psychol 1987;26:73 – 4.