Pain 106 (2003) 393–400
`
`www.elsevier.com/locate/pain
`
`Prospective examination of pain-related and psychological
`predictors of CRPS-like phenomena following
`total knee arthroplasty: a preliminary study
`
`R. Norman Hardena,*, Stephen Bruehlb, Steven Stanosa, Victoria Branderb,
`Ok Yung Chungc, Samuel Saltza, Angie Adamsa, S. David Stulbergd
`
`aCenter for Pain Studies, Rehabilitation Institute of Chicago, 345 E. Superior St., Chicago, IL 60610, USA
`bDepartment of Physical Medicine and Rehabilitation, Northwestern University Medical School, Chicago, IL, USA
`cDepartment of Anesthesiology, Vanderbilt University School of Medicine, Nashville, TN, USA
`dDepartment of Orthopedic Surgery, Northwestern University Medical School, Chicago, IL, USA
`
`Received 1 April 2003; received in revised form 7 August 2003; accepted 18 August 2003
`
`Abstract
`
`We hypothesized that preoperative emotional distress and pain intensity would predict the occurrence of signs and symptoms of complex
`regional pain syndrome (CRPS) following total knee arthroplasty (TKA). Depression (Beck Depression Inventory, BDI), anxiety (State Trait
`Anxiety Inventory, STAI), pain (McGill Pain Questionnaire – Short Form, MPQ), and signs/symptoms meeting IASP criteria for CRPS were
`assessed preoperatively, and at 1-, 3-, and 6-months postoperatively in 77 patients undergoing TKA. The prevalence of subjects fulfilling
`CRPS criteria was 21.0% at 1 month, 13.0% at 3 months, and 12.7% at 6 months postoperative. Higher preoperative scores on the STAI
`predicted positive CRPS status at 1-month follow-up ðP , 0:05Þ; with a similar non-significant trend for preoperative BDI scores ðP , 0:10Þ:
`Diagnostic sensitivity for the STAI was good (0.73), with moderate specificity (0.56). Neither measure predicted CRPS at later follow-up
`ðP . 0:10Þ: Greater preoperative pain intensity predicted positive CRPS status at 3-month (MPQ-Sensory and MPQ-Affective; P , 0:01)
`and 6-month (MPQ-Sensory) follow-up ðP , 0:01Þ; but not at 1-month ðP . 0:10Þ: Diagnostic sensitivity was high (0.83 – 1.00), with
`moderate specificity (0.53 – 0.60). Post-TKA patients with CRPS were more depressed at 1-month follow-up ðP , 0:05Þ and more anxious at
`6-month follow-up ðP , 0:05Þ than patients with ongoing non-CRPS pain (all other comparisons non-significant, P . 0:10). Overall, results
`indicate that CRPS-like phenomena occur in a significant number of patients early post-TKA; however, it is not associated with significantly
`greater complaints of postoperative pain. There appears to be a modest utility for preoperative distress and pain in predicting CRPS signs and
`symptoms following TKA, although false positive rates are relatively high.
`q 2003 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
`
`Keywords: Complex regional pain syndrome; Reflex sympathetic dystrophy; Anxiety; Depression; Prospective; Total knee arthroplasty
`
`1. Introduction
`
`Complex regional pain syndrome (CRPS) is a pain
`disorder associated with autonomic dysfunction (Merskey
`and Bogduk, 1994). It can develop following serious trauma
`(e.g. crush or fracture), surgery, or relatively minor injury
`(Bruehl et al., 1999; Galer et al., 1998). Despite increasing
`research interest,
`little is known regarding why some
`patients develop CRPS and others do not, despite similar
`initiating events.
`
`* Corresponding author. Tel.: þ 1-312-238-7878.
`E-mail address: storminmd@aol.com, nharden@rehabchicago.org
`(R.N. Harden).
`
`In one well-controlled prospective study, Schurmann
`et al. (2000) studied patients following distal radial fracture.
`An impaired upper extremity vasoconstrictor response
`occurring post-injury differentiated subjects later develop-
`ing CRPS from those that did not. These results suggest
`autonomic dysfunction may precede development of CRPS,
`consistent with other research indicating a role for altered
`local catecholamine responsiveness (Arnold et al., 1993;
`Baron and Maier, 1996; Birklein et al., 1998; Drummond
`et al., 1991; Harden et al., 1994; Ja¨nig and Baron, 2001;
`Kurvers et al., 1998).
`Psychological factors have been hypothesized to predis-
`pose for or perpetuate CRPS, although the literature on this
`issue is mixed (Bruehl, 2001; Bruehl and Carlson, 1992).
`
`0304-3959/$30.00 q 2003 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
`doi:10.1016/j.pain.2003.08.009
`
`Grun. Exh. 1055
`PGR for U.S. Patent No. 9,408,862
`
`

`

`394
`
`R.N. Harden et al. / Pain 106 (2003) 393–400
`
`Several studies report that CRPS patients have higher levels
`of emotional distress than non-CRPS pain patients (Bruehl
`et al., 1996; Ciccone et al., 1997; Geertzen et al., 1994;
`Hardy and Merritt, 1988) and may experience greater
`premorbid life stress than non-CRPS pain patients (Geertzen
`et al., 1998). Other studies fail to find these differences
`(DeGood et al., 1993; Haddox et al., 1988; Monti et al.,
`1998; Nelson and Novy, 1996). There is one prospective
`study reporting a 91% accuracy of ‘psychiatric risk factors’
`in predicting development of CRPS-like symptoms follow-
`ing surgery for Dupuytren’s contracture (Zachariae, 1964).
`Interpretation of these results is problematic, secondary to
`poor definition of these factors (e.g. ‘sthenic’ personality).
`Psychological distress is often associated with increased
`catecholamine activity (Charney et al., 1990; Light et al.,
`1998), and given the possible role of altered catecholamine
`responsiveness in the pathophysiology of CRPS, a theoreti-
`cal pathway can be described by which psychological
`factors could impact on CRPS development following
`injury (Bruehl, 2001; Harden et al., 1998).
`Testing this hypothesis requires a prospective design,
`which in turn, necessitates study of a population in whom
`CRPS develops with some regularity. Although from 25 to
`40% of patients post-Colles’ fracture may eventually
`experience features suggestive of CRPS (Atkins et al.,
`1989; Bickerstaff and Kanis, 1994), these studies failed to
`use standardized diagnostic criteria and their interpretation
`is problematic. The uncontrolled nature of injury in
`postfracture CRPS makes this model less than ideal. It has
`been proposed that total knee arthroplasty (TKA) might
`provide a useful model for prospectively examining factors
`underlying development of CRPS, given the more con-
`trolled nature of this specific surgical trauma (Stanos et al.,
`2001). Such a model appears viable in that surgery is
`reported to be a precipitating factor in CRPS (Bruehl et al.,
`1999; Galer et al., 1998; Harden et al., 1999). This study was
`designed to explore the utility of this model for examining
`the role of preoperative pain and psychological distress in
`the postsurgical development of CRPS.
`
`2. Method
`
`2.1. Design
`
`This study used a prospective mixed within/between
`subject design.
`
`2.2. Sample
`
`The sample included 77 patients scheduled to undergo
`unilateral TKA. No contralateral TKAs were performed
`on subjects during the study period. All patients had non-
`inflammatory arthritis, with a primary diagnosis of
`osteoarthritis in 93.7% of patients. The right knee was
`affected in a slight majority of cases (54.8%). The sample
`
`was predominately female (61.6%) and Caucasian
`(88.7%), with a mean age of 66.3 ^ 1.29 years. The
`majority of subjects (61%) had at
`least some college
`education. All patients were drawn from the clinic
`population of an orthopedic surgeon highly experienced
`in performing TKAs in an academically affiliated urban
`hospital, and all surgical procedures were performed by
`the same physician (SDS). Due to the prospective nature
`of the study and the elderly sample,
`the sample size
`diminished somewhat across the three follow-up assess-
`ment periods (n ¼ 77; n ¼ 69; and n ¼ 55; respectively).
`Subjects lost
`to follow-up at 6 months did not differ
`significantly from those completing 6-month follow-up
`with regard to any preoperative distress or pain measures
`(all P . 0:10).
`
`2.3. Procedures
`
`Potential subjects were initially approached in the
`presurgical outpatient orthopedic clinic at a preoperative
`visit, approximately 1 month before surgery and informed
`of the purpose of the study and the study procedures.
`Subjects then provided informed consent. Following this,
`patients were asked to complete a baseline packet of
`demographic information and psychometric measures.
`Psychometric measures included the following: (1) McGill
`Pain Questionnaire – Short Form (MPQ; Melzack, 1987), a
`standardized measure of the sensory and affective dimen-
`sions of pain, on which current pain at time of evaluation
`was rated; (2) Beck Depression Inventory (BDI; Beck et al.,
`1961), a widely used standardized measure of depressive
`symptoms on which depressive symptoms in the previous 2
`weeks were rated; and (3) Trait form of the State Trait
`Anxiety Inventory (STAI; Spielberger et al., 1970), a widely
`used standardized measure on which patients rated their
`general tendency to experience anxiety symptoms. Table 1
`displays the intercorrelations among these measures at
`preoperative baseline, with the greatest overlap noted
`between BDI and STAI scores (46% shared variance). In
`addition to completing this psychometric packet preopera-
`tively, patients were again asked to complete these same
`instruments at three planned postsurgical follow-up assess-
`ment periods, approximately 1 month (4.2 ^ 0.14 weeks),
`3 months (13.2 ^ 0.42 weeks), and 6 months (31.4 ^ 1.10
`weeks) following TKA.
`
`Table 1
`Intercorrelations among distress and pain measures at preoperative baseline
`
`Measure
`
`BDI
`
`STAI
`
`MPQ-Sensory
`
`STAI
`MPQ-Sensory
`MPQ-Affective
`
`0.68**
`0.19
`0.38**
`
`–
`0.14
`0.31**
`
`–
`–
`0.67**
`
`**P , 0:01: BDI, Beck Depression Inventory; STAI, State Trait
`Anxiety Inventory; MPQ, McGill Pain Questionnaire – Short Form.
`
`

`

`R.N. Harden et al. / Pain 106 (2003) 393–400
`
`395
`
`At each of the follow-up periods, subjects underwent an
`evaluation by a study physician to determine whether IASP
`diagnostic criteria for CRPS were met. All signs and
`symptoms of CRPS were evaluated and recorded on a CRPS
`database form using procedures we have employed in our
`previous research (see Bruehl et al., 2002, 1999; Harden
`et al., 1999). The IASP criteria for CRPS required the
`presence of an abnormal evoked response or
`report
`suggestive of a pain processing abnormality (e.g. hyper-
`algesia or allodynia), and observation or report of unilateral
`edema or autonomic dysfunction (localized temperature,
`skin color and/or sweating changes). Using a literal
`interpretation of the IASP criteria which require the
`presence ‘at some time’ of the above features, patients
`were considered positive for CRPS if symptoms were
`reported and/or signs were observed on examination
`meeting the two criteria above (Merskey and Bogduk,
`1994). Under normal clinical circumstances, CRPS would
`not be diagnosed if a clear condition other than CRPS could
`account for the patient’s presentation (criterion 4 of the
`IASP criteria). Given that the first follow-up period was
`approximately 1 month following surgery,
`it could be
`argued that any CRPS-like phenomena at this time point
`(e.g. calor, rubor, dolor) might be the result of delayed post-
`surgical healing issues that would not normally merit a
`diagnosis of CRPS. Thus, interpretation of a positive CRPS
`diagnosis at 1-month follow-up in the current study should
`be tempered by this possible caveat. A CRPS diagnosis at
`later follow-up periods would be more likely to represent a
`true clinical diagnosis of CRPS, given that normal healing
`should have been completed by these latter time periods.
`
`2.4. Statistical analysis
`
`All analyses were conducted using the SPSS for
`Windows Version 9 statistical package (SPSS,
`Inc.,
`Chicago, IL). The first set of analyses was prospective in
`nature, and designed to examine whether preoperative pain
`and distress impacted on postoperative development of
`CRPS. Multivariate analyses would have theoretically been
`optimal, given the prospective nature of the study. However,
`due to the degree of subject attrition across follow-up
`periods (reducing power for such analyses) and the
`relatively low rate of occurrence of CRPS in the sample,
`this multivariate approach suffered from significant limi-
`tations. Point-biserial correlations were therefore used to
`examine the relationship between CRPS/non-CRPS status at
`each follow-up time point and scores on preoperative
`measures of pain (MPQ) and distress (BDI and STAI). In
`order to present these results in a clinically meaningful way,
`high/low pain and distress subgroups were derived based on
`post-hoc median splits on each of
`the preoperative
`measures. The percentage of patients diagnosed with
`CRPS within these high/low subgroups is presented in
`tabular form to clarify the nature of the observed effects.
`Diagnostic sensitivity and specificity were also calculated,
`
`with true positive status considered to exist when elevated
`baseline pain or psychological distress was associated with a
`CRPS diagnosis.
`The second set of analyses was conducted using cross-
`sectional data at each follow-up time point, and was
`restricted only to patients continuing to report significant
`pain (at least 30/100 on a VAS) for the given follow-up
`period. t-Tests were used to compare levels of pain and
`distress at follow-up between patients who met criteria for
`CRPS and those who did not meet CRPS diagnostic criteria.
`For each comparison, Levene’s test for equality of variances
`was conducted, and if significant differences in variance
`were present, appropriately adjusted statistical values were
`used in comparisons.
`All means are presented as mean ^ SE. A two-tailed
`criterion of significant of P , 0:05 was used for all primary
`analyses. The maximum available number of subjects with
`valid data were analyzed at each time point.
`
`3. Results
`
`3.1. Prevalence of CRPS post-TKA
`
`The prevalence of a pattern of signs and symptoms
`meeting CRPS diagnostic criteria across follow-up assess-
`ment periods was: 1 month, 21.0%; 3 months, 13.0%; and
`6 months, 12.7%. Table 2 summarizes the frequencies of
`CRPS-related signs and symptoms in the group meeting
`criteria for CRPS and the group not meeting these criteria.
`Not surprisingly, in nearly every case, CRPS signs and
`symptoms were more common in the CRPS group than the
`non-CRPS group. Temperature asymmetry and edema were
`found to be relatively common in both diagnostic groups,
`but were associated with pain processing abnormalities only
`in the CRPS group. Across all assessment periods, the
`temperature asymmetry in the CRPS group reflected almost
`entirely increased temperature on the affected side, with the
`color changes in all cases reflecting erythema. Obvious
`changes in sweating patterns were rare in both groups.
`
`3.2. Can post-TKA CRPS be predicted by preoperative pain
`or psychological distress?
`
`The ability of preoperative pain intensity and psycho-
`logical distress to predict CRPS versus non-CRPS diag-
`nostic status at each follow-up period was next explored.
`Results for the STAI are presented in Table 3. Greater
`preoperative anxiety on the STAI was associated with
`significantly more frequent diagnosis of CRPS at 1-month
`follow-up [rð75Þ ¼ 0:23; P , 0:05]. Nearly 29% of patients
`above the median on preoperative trait anxiety met criteria
`for CRPS, whereas only 10% below the median on anxiety
`were diagnosed with CRPS. CRPS diagnostic sensitivity for
`the STAI was good (0.73), although diagnostic specificity
`was only moderate (0.56). Results for the STAI at all other
`
`

`

`396
`
`R.N. Harden et al. / Pain 106 (2003) 393–400
`
`Table 2
`Signs and symptoms of CRPS (% positive) by diagnostic group across study periods
`
`CRPS characteristics
`
`1-Month follow-up
`Non-CRPS ðn ¼ 61Þ
`
`CRPS ðn ¼ 16Þ
`
`3-Month follow-up
`Non-CRPS ðn ¼ 60Þ
`
`CRPS ðn ¼ 9Þ
`
`6-Month follow-up
`Non-CRPS ðn ¼ 48Þ
`
`CRPS ðn ¼ 7Þ
`
`Self-reported symptoms
`Hyperesthesia
`Temperature asymmetry
`Skin color asymmetry
`Sweating alterations
`Edema
`
`3.3
`68.3
`35.0
`3.4
`83.3
`
`Signs observed on examination
`Hyperalgesia
`0.0
`Allodynia
`0.0
`Temperature asymmetry
`86.4
`% Warmer affected side
`98.0
`Skin color asymmetry
`39.0
`% Erythematous
`95.5
`Sweating alterations
`0.0
`% Increased sweating
`–
`Edema
`84.5
`
`43.8
`68.8
`56.3
`0.0
`87.5
`
`37.5
`31.3
`87.5
`100.0
`50.0
`100.0
`0.0
`–
`68.8
`
`0.0
`63.3
`31.7
`0.0
`70.0
`
`0.0
`0.0
`89.5
`100.0
`15.8
`72.7
`1.8
`100.0
`71.9
`
`66.7
`100.0
`55.6
`11.1
`88.9
`
`22.2
`55.6
`100.0
`87.5
`22.2
`100.0
`11.1
`100.0
`88.9
`
`0.0
`47.9
`16.7
`0.0
`54.2
`
`0.0
`0.0
`54.2
`92.6
`8.5
`80.0
`2.1
`100.0
`47.9
`
`28.6
`71.4
`14.3
`0.0
`100.0
`
`28.6
`28.6
`57.1
`100.0
`28.6
`100.0
`0.0
`–
`85.7
`
`ðP . 0:10Þ;
`follow-up periods were non-significant
`although in a pattern similar to that noted above, twice as
`many patients above the median on the STAI were
`diagnosed with CRPS compared to those below the median
`at 6-month follow-up.
`Results for the BDI are presented in Table 4. A non-
`significant trend similar to the STAI results at 1-month
`follow-up was noted for the BDI [rð74Þ ¼ 0:19; P , 0:10].
`As for the STAI, 28% of patients above the median on the
`BDI were diagnosed with CRPS compared to only 14% of
`those below the median on the BDI. The BDI exhibited both
`moderate diagnostic sensitivity (0.60) and moderate speci-
`ficity (0.62). Analyses of BDI data at the other time points
`were non-significant (all P . 0:10).
`Ratings of preoperative sensory pain intensity on the
`MPQ-Sensory subscale displayed the most consistent
`
`predictive effects (see Table 5). At 3-month follow-up, no
`patients below the median on preoperative MPQ-Sensory
`ratings were diagnosed with CRPS, whereas 18% above the
`median met CRPS criteria [rð65Þ ¼ 0:36; P , 0:01]. A
`comparable difference was observed at 6-month follow-up
`[rð51Þ ¼ 0:34; P , 0:01]. While no patients below the
`median on preoperative MPQ-Sensory ratings were diag-
`nosed with CRPS, 24% above the median met CRPS
`criteria. Diagnostic sensitivity for the MPQ-Sensory sub-
`scale at 3- and 6-month follow-up was very high (both 1.00),
`with moderate specificity in each case (0.54 and 0.60,
`respectively). Results for the MPQ-Sensory subscale at 1-
`month follow-up were non-significant ðP . 0:10Þ: As with
`the MPQ-Sensory subscale, preoperative ratings of affective
`pain intensity on the MPQ-Affective subscale were
`associated with CRPS status at 6-month follow-up
`
`Table 3
`CRPS diagnosis (%) at each follow-up assessment by preoperative anxiety
`levels
`
`Table 4
`CRPS diagnosis (%) at each follow-up assessment by preoperative
`depression levels
`
`Preoperative anxiety (STAI)
`
`Preoperative depression (BDI)
`
`1-Month follow-up*
`Non-CRPS ðn ¼ 62Þ
`CRPS ðn ¼ 15Þ
`
`3-Month follow-up
`Non-CRPS ðn ¼ 62Þ
`CRPS ðn ¼ 6Þ
`
`6-Month follow-up
`Non-CRPS ðn ¼ 47Þ
`CRPS ðn ¼ 6Þ
`
`Low
`
`89.7
`10.3
`
`91.4
`8.6
`
`92.6
`7.4
`
`High
`
`71.1
`28.9
`
`90.9
`9.1
`
`84.6
`15.4
`
`1-Month follow-up †
`Non-CRPS ðn ¼ 61Þ
`CRPS ðn ¼ 15Þ
`
`3-Month follow-up
`Non-CRPS ðn ¼ 62Þ
`CRPS ðn ¼ 6Þ
`
`6-Month follow-up
`Non-CRPS ðn ¼ 46Þ
`CRPS ðn ¼ 6Þ
`
`Low
`
`86.4
`13.6
`
`90.0
`10.0
`
`90.6
`9.4
`
`High
`
`71.9
`28.1
`
`92.9
`7.1
`
`85.0
`15.0
`
`*P , 0:05: Mean STAI scores by median split subgroup: high,
`40.3 ^ 0.99; low, 25.6 ^ 0.57.
`
`†P , 0:10: Mean BDI scores by median split subgroup: high,
`11.7 ^ 0.72; low, 4.2 ^ 0.33.
`
`

`

`R.N. Harden et al. / Pain 106 (2003) 393–400
`
`397
`
`Table 5
`CRPS diagnosis (%) at each follow-up assessment by preoperative sensory
`pain intensity
`
`Preoperative MPQ-Sensory
`
`1-Month follow-up
`Non-CRPS ðn ¼ 62Þ
`CRPS ðn ¼ 14Þ
`
`3-Month follow-up*
`Non-CRPS ðn ¼ 61Þ
`CRPS ðn ¼ 6Þ
`
`6-Month follow-up**
`Non-CRPS ðn ¼ 47Þ
`CRPS ðn ¼ 6Þ
`
`Low
`
`86.5
`13.5
`
`100.0
`0.0
`
`100.0
`0.0
`
`High
`
`76.9
`23.1
`
`82.4
`17.6
`
`76.0
`24.0
`
`reported significantly higher depression levels at 1-month
`follow-up than did patients not meeting CRPS criteria
`[tð47Þ ¼ 2:08; P , 0:05]. Similarly, patients diagnosed with
`CRPS reported significantly greater anxiety on the STAI at
`6-month follow-up compared to those patients without
`CRPS [tð18Þ ¼ 1:87; P , 0:05]. Diagnostic group compari-
`sons on the BDI and STAI at other follow-up periods were
`non-significant (all P . 0:10).
`Pain intensity ratings across diagnostic groups at each
`follow-up period were also examined. Significantly higher
`pain ratings among CRPS patients were not detected at any
`follow-up period, although a non-significant trend towards
`greater MPQ-Sensory ratings among CRPS patients was
`noted at the second follow-up period ðP , 0:10Þ:
`
`*P , 0:05; **P , 0:01: Mean MPQ-Sensory scores by median split
`subgroup: high, 15.8 ^ 1.09; low, 4.1 ^ 0.36.
`
`4. Discussion
`
`[rð49Þ ¼ 0:33; P , 0:05], with a similar non-significant
`trend noted at 3-month follow-up [rð63Þ ¼ 0:21; P , 0:10;
`see Table 6]. Diagnostic sensitivity for the MPQ-Affective
`subscale at 6-month follow-up was high (0.83), with
`moderate specificity (0.53).
`
`3.3. Do CRPS patients differ from non-CRPS patients
`experiencing pain post-TKA?
`
`Data were also analyzed on the cross-sectional relation-
`ship between distress/pain and CRPS status at each follow-
`up period. We hypothesized, based on previous studies, that
`among patients experiencing significant pain post-TKA,
`those diagnosed with CRPS would be more emotionally
`distressed than those not experiencing CRPS. Results of
`analyses regarding this issue are presented in Table 7.
`Ratings of depression on the BDI at 1-month follow-up
`supported this hypothesis. Patients meeting CRPS criteria
`
`Table 6
`CRPS diagnosis (%) at each follow-up assessment by preoperative affective
`pain intensity
`
`Preoperative MPQ-Affective
`
`1-Month follow-up
`Non-CRPS ðn ¼ 61Þ
`CRPS ðn ¼ 13Þ
`3-Month follow-up †
`Non-CRPS ðn ¼ 60Þ
`CRPS ðn ¼ 5Þ
`
`6-Month follow-up*
`Non-CRPS ðn ¼ 45Þ
`CRPS ðn ¼ 6Þ
`
`Low
`
`87.9
`12.1
`
`96.6
`3.4
`
`96.0
`4.0
`
`High
`
`78.0
`22.0
`
`88.9
`11.1
`
`80.8
`19.2
`
`†P , 0:10; *P , 0:05: Mean MPQ-Affective scores by median split
`subgroup: high, 4.2 ^ 0.47; low, 0.4 ^ 0.09.
`
`It was hypothesized that greater psychological distress
`and pain preoperatively would predict development of
`CRPS following TKA. At 1-month follow-up, presence of
`signs and symptoms consistent with CRPS was significantly
`predicted by preoperative trait anxiety levels, with a similar
`non-significant trend for baseline depression levels. Rates of
`CRPS diagnosis were twice as high in subjects with elevated
`preoperative distress than in those with lower distress levels.
`Despite this prediction of CRPS diagnosis at 1-month
`follow-up, psychological distress did not significantly
`predict CRPS diagnosis at 3- and 6-month follow-up. In
`contrast, baseline pain intensity did not significantly predict
`initial post-TKA CRPS diagnosis, but did predict CRPS at
`3- and 6-month follow-up. This predictive effect was noted
`both for the sensory dimension of pain (at 3- and 6-month
`follow-up) and for the affective (suffering) dimension of
`pain (at 6-month follow-up) as assessed by the MPQ. The
`diagnostic sensitivity of preoperative pain intensity was
`high, although specificity was only moderate. In other
`words, elevated preoperative pain was very useful
`in
`predicting future cases of CRPS, but was less useful for
`predicting future non-CRPS cases. Diagnostic sensitivity for
`the measure of trait anxiety was less impressive, although
`still reasonably good (0.73), with specificity found to be
`only moderate. As with baseline pain intensity, trait anxiety
`was better at predicting those who would go on to develop
`CRPS symptomatology than those who would not develop
`such symptomatology.
`The relatively modest and inconsistent predictive utility
`for the baseline distress measures suggest that there may not
`be a large effect of distress on CRPS development following
`TKA. Such predictive effects may have appeared stronger in
`a sample with a substantially larger number of CRPS cases,
`although the practical difficulties of obtaining large
`prospective samples for a condition that occurs relatively
`rarely are somewhat daunting. Even for preoperative pain
`intensity, which showed stronger and longer-term predictive
`utility, the clinical usefulness may be limited by its modest
`
`

`

`398
`
`R.N. Harden et al. / Pain 106 (2003) 393–400
`
`Table 7
`Pain intensity and psychological distress at follow-up by CRPS status among patients experiencing significant ongoing pain post-TKA
`
`Measure
`
`1-Month follow-up
`Non-CRPS ðn ¼ 36Þ
`
`BDI
`STAI
`MPQ-Sensory
`MPQ-Affective
`
`6.9 ^ 0.83
`32.8 ^ 1.37
`10.7 ^ 1.08
`2.1 ^ 0.44
`
`†P , 0:10; *P , 0:05:
`
`CRPS ðn ¼ 13Þ
`
`10.4 ^ 1.67*
`33.6 ^ 2.61
`12.8 ^ 1.61
`2.6 ^ 0.80
`
`3-Month follow-up
`Non-CRPS ðn ¼ 21Þ
`
`7.0 ^ 1.51
`34.1 ^ 1.96
`9.6 ^ 1.08
`2.6 ^ 0.51
`
`CRPS ðn ¼ 7Þ
`
`7.3 ^ 1.88
`31.5 ^ 2.99
`13.2 ^ 2.65†
`2.4 ^ 1.04
`
`6-Month follow-up
`Non-CRPS ðn ¼ 15Þ
`
`5.4 ^ 1.15
`29.9 ^ 1.42
`8.9 ^ 1.29
`2.1 ^ 0.77
`
`CRPS ðn ¼ 5Þ
`
`9.0 ^ 3.61
`37.6 ^ 5.94*
`6.4 ^ 0.60
`1.2 ^ 0.37
`
`specificity (i.e. high false positive rate). Findings of at least
`moderate predictive utility for preoperative pain and distress
`measures are consistent with existing literature. Preopera-
`tive distress and pain intensity have both been shown in
`previous work to be long-term predictors of post-surgical
`non-CRPS chronic pain (e.g. Haythornthwaite et al., 1998;
`Poobalan et al., 2003). Patients with greater pain before
`TKA surgery are at greater risk for heightened postoperative
`TKA pain, irrespective of potentially confounding issues,
`such as preoperative disease severity or postoperative
`complications (Brander et al., 2003). Moreover, work by
`Gracely et al. (1992) would suggest that greater preopera-
`tive pain intensity could alter central pain processing
`pathways, and thereby lead to greater likelihood of post-
`TKA CRPS. In light of these findings, results of this study
`suggest that it may be clinically useful to assess distress and
`pain intensity preoperatively, and if elevated levels are
`found relative to normative data, to implement appropriate
`interventions preoperatively, and to monitor such patients
`more closely for possible CRPS development postopera-
`tively so that early intervention can be initiated. Although
`identifying with certainty patients who will develop
`postsurgical CRPS from preoperative measures does not
`appear feasible, the monitoring approach above might for
`some patients reduce the likelihood of chronic CRPS
`developing.
`Based on prior, albeit inconsistent, evidence that patients
`experiencing CRPS are more distressed than non-CRPS
`pain patients (Bruehl et al., 1996; Ciccone et al., 1997;
`Geertzen et al., 1994; Hardy and Merritt, 1988), it was
`hypothesized that patients meeting CRPS criteria following
`TKA would be more distressed than post-TKA patients
`reporting pain in the absence of a CRPS diagnosis. The
`prospective nature of the current study allowed this
`hypothesis to be examined within the same sample
`repeatedly at several time points in the natural history of
`CRPS. Results indicated that TKA patients meeting CRPS
`criteria at 1-month follow-up were more depressed than
`non-CRPS TKA patients despite comparable pain intensity
`levels. Similarly, CRPS group patients were more anxious at
`6-month follow-up than were non-CRPS patients, despite
`reporting pain levels that were no higher. Comparisons
`between diagnostic groups at other time points were not
`significantly different. Thus, differences in depression and
`
`anxiety levels between CRPS and non-CRPS pain patients
`observed at one follow-up period were not observed at the
`other follow-up periods. Overall,
`these findings do not
`support pervasive and long-term differences in emotional
`distress between CRPS and non-CRPS patients experien-
`cing at least moderate pain following TKA.
`One issue to be considered in interpreting the current
`results is whether
`the patients diagnosed with CRPS
`following TKA were experiencing ‘true’ CRPS. This issue
`is difficult
`to address given the absence of a ‘gold
`standard’
`for CRPS diagnosis and a lack of clarity
`regarding CRPS pathophysiology. However,
`increased
`temperature, erythema, and edema in the affected
`extremity were found to characterize nearly the entire
`subsample diagnosed with CRPS at all follow-up time
`points. This finding might be expected given the results of
`previous prospective work employing careful autonomic
`testing (Birklein et al., 1998). In a sample of CRPS
`patients, acute CRPS (median of 5 weeks, ranging from 2
`to 21 weeks) was found to be associated most frequently
`with locally increased temperature, with a transformation
`to predominately decreased temperature occurring only
`during the chronic phase at two year follow-up (Birklein
`et al., 1998). Cross-sectional research by the same group
`suggests that acute CRPS (median ¼ 5 weeks duration)
`may be distinguished from normal healing following
`trauma by an impaired sympathetic vasoconstrictor reflex,
`although both normal post-traumatic healing and CRPS
`groups typically displayed increased temperature in the
`affected extremity (Birklein et al., 2001). In the current
`study,
`increased temperatures in the affected extremity
`were common among both CRPS and non-CRPS groups,
`as might be expected based on the results of Birklein et al.
`(2001). However, these temperature changes were associ-
`ated with pain processing abnormalities only in the CRPS
`group. Practical limitations in the current study prevented
`reliable longitudinal assessment of sympathetic reflexes,
`and therefore,
`it
`is not possible to use this index to
`strengthen CRPS diagnostic conclusions.
`Despite the issues above, results of the current study
`clearly indicate that the prevalence of CRPS-like phenom-
`ena diminishes noticeably over the first 3 months post-
`TKA, with some stabilization of CRPS prevalence at 6
`months. Although at least moderate ongoing pain may be
`
`

`

`R.N. Harden et al. / Pain 106 (2003) 393–400
`
`399
`
`expected in as many as 50% of TKA patients up to 1 year
`post-surgery (Brander et al., 2003; Dickstein et al., 1998;
`Murray and Frost, 1998), the current results indicate that
`only a much smaller percentage (13%) are likely to be
`experiencing CRPS-like phenomena. Prevalence estimates
`for CRPS post-TKA in the current study are notably
`higher than previously published estimates of less than 3%
`of post-TKA patients based on various non-standardized
`diagnostic criteria (Cameron et al., 1994; Katz et al.,
`1986; Ritter, 1997). These differing prevalence estimates
`are likely to be due at least in part to the differences in
`diagnostic criteria used. The current study is the first to
`apply the IASP diagnostic criteria, which previous work
`indicates may have only moderate specificity and there-
`fore may tend to overdiagnose the condition (Bruehl et al.,
`1999; Galer et al., 1998; Harden et al., 1999). Overall,
`results of this study suggest that CRPS-like phenomena
`may occur following TKA more frequently than pre-
`viously believed. This issue merits further study, and also
`emphasizes the need for routine, close clinical monitoring
`for
`such phenomena in post-TKA patients
`so that
`appropriate intervention may be implemented as early as
`possible.
`There are several limitations to the current study. Inflated
`type I error rate could have impacted on the results, given
`the number of analyses performed. Conservative adjustment
`for type I error rate would have resulted in many of the
`reported results being non-significant. As noted above,
`statistical power issues may also have impacted on the
`results, given the relatively small number of CRPS cases
`developing over the course of the study and the significant
`sample attrition occurring by 6-month follow-up. Such
`attrition could have either increased CRPS prevalence in the
`sample (if patients doing well systematically failed to show
`up for
`follow-up appointments), or decreased CRPS
`prevalence (if CRPS patients systematically left the study
`to seek treatment elsewhere). We have no information to
`indicate that either of
`these types of bias occurred
`exclusively, but note that preoperative pain and distress
`measures did not differ between subjects who completed
`6-month follow-up and those who did not. Given the elderly
`nature of the sample and frequent transportation problems, it
`seems most likely that subject attrition was random. In
`either case, sample attrition issues should not have system-
`atically affected within-subject relationships between pre-
`operative factors and later CRPS status.
`In summary,
`this study demonstrated that greater
`preoperative anxiety modestly predicts initial CRPS symp-
`tomatology following TKA, and that elevated preoperative
`pain intensity predicts the development of longer-term post-
`TKA CRPS. Given the limitations above, this work should
`be considered preliminary. Replication is required in a
`larger sample using a design that includes measures of
`additional psychological constructs and formal testing of
`vasomotor reflex function throughout the study.
`
`References
`
`Arnold JMO, Teasell RW, MacLeod AP, Brown JE, Carruthers SG.
`Increased venous alpha-adrenoceptor
`responsiveness
`in patients
`with reflex sympathetic dystrophy. Ann Intern Med 1993;118:
`619 – 21.
`Atkins RM, Duckworth T, Kanis JA