PAINÒ 150 (2010) 268–274
`
`w w w . e l s e v i e r . c o m / l o c a t e / p a i n
`
`Validation of proposed diagnostic criteria (the ‘‘Budapest Criteria”)
`for Complex Regional Pain Syndrome
`R. Norman Harden a,*, Stephen Bruehl b, Roberto S.G.M. Perez c,d, Frank Birklein e, Johan Marinus d,f,
`Christian Maihofner g, Timothy Lubenow h, Asokumar Buvanendran h, Sean Mackey i, Joseph Graciosa a,
`Mila Mogilevski a, Christopher Ramsden a, Melissa Chont b, Jean-Jacques Vatine j
`a Rehabilitation Institute of Chicago, Chicago, IL, USA
`b Vanderbilt University School of Medicine, Nashville, TN, USA
`c VU University Medical Center, Amsterdam, The Netherlands
`d Trauma Related Neuronal Dysfunction Consortium (TREND), Leiden University Medical Center, Leiden, The Netherlands
`e University Medical Center Mainz, Mainz, Germany
`f Leiden University Medical Center, Leiden, The Netherlands
`g University of Erlangen-Nuremberg, Erlangen, Germany
`h Rush University Medical Center, Chicago, IL, USA
`i Stanford University Medical Center, Stanford, CA, USA
`j Reuth Medical Center, Sackler Faculty of Medicine, Tel Aviv University, Tel Aviv, Israel
`
`a r t i c l e
`
`i n f o
`
`a b s t r a c t
`
`Article history:
`Received 18 November 2009
`Received in revised form 19 March 2010
`Accepted 20 April 2010
`
`Keywords:
`Complex Regional Pain Syndrome
`Reflex sympathetic dystrophy
`CRPS
`RSD
`Diagnosis
`Validation
`
`1. Introduction
`
`Current IASP diagnostic criteria for CRPS have low specificity, potentially leading to overdiagnosis. This
`validation study compared current IASP diagnostic criteria for CRPS to proposed new diagnostic criteria
`(the ‘‘Budapest Criteria”) regarding diagnostic accuracy. Structured evaluations of CRPS-related signs and
`symptoms were conducted in 113 CRPS-I and 47 non-CRPS neuropathic pain patients. Discriminating
`between diagnostic groups based on presence of signs or symptoms meeting IASP criteria showed high
`diagnostic sensitivity (1.00), but poor specificity (0.41), replicating prior work. In comparison, the Buda-
`pest clinical criteria retained the exceptional sensitivity of the IASP criteria (0.99), but greatly improved
`upon the specificity (0.68). As designed, the Budapest research criteria resulted in the highest specificity
`(0.79), again replicating prior work. Analyses indicated that inclusion of four distinct CRPS components in
`the Budapest Criteria contributed to enhanced specificity. Overall, results corroborate the validity of the
`Budapest Criteria and suggest they improve upon existing IASP diagnostic criteria for CRPS.
`Ó 2010 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
`
`literature regarding the disorder now called
`The historical
`Complex Regional Pain Syndrome (CRPS) reflects an array of idiosyn-
`cratic diagnostic schemes [1,4,11,17,28,35]. In response, an interna-
`tional meeting was held in 1993 in Orlando, Florida to develop
`consensus terminology (i.e., CRPS) and standardized diagnostic cri-
`teria to improve clinical recognition of the disorder and facilitate the
`selection of more generalizable research samples [33,36]. Since pub-
`lication of these consensus-based criteria by the International Asso-
`ciation for the Study of Pain (IASP) [23], the extent of their use in the
`clinical setting is unknown, but their application in the research
`setting has been shown to be inconsistent [28].
`
`* Corresponding author. Address: Center for Pain Studies, Rehabilitation Institute
`of Chicago, 446 E. Ontario, Suite 1011, Chicago, IL 60611, USA. Tel.: +1 312 238
`7878; fax: +1 312 238 7624.
`E-mail addresses: nharden@ric.org, hcaporoso@ric.org (R.N. Harden).
`
`Despite the inherent advantages of having standardized, interna-
`tionally-recognized diagnostic criteria for CRPS, it has been sug-
`gested that a lack of proven validity may be a barrier to their use
`by researchers and clinicians [6,13,17]. Incomplete understanding
`of CRPS pathophysiology and the resulting lack of a ‘‘gold standard”
`test make the design of validation studies more challenging [6,28].
`However, studies conducted to date suggest that the IASP criteria
`for CRPS suffer from a lack of specificity [6,10,13]. That is, while
`the IASP criteria may accurately identify most cases of CRPS, they
`also tend to misidentify non-CRPS neuropathic pain conditions as
`CRPS, potentially contributing to overdiagnosis and either inappro-
`priate or unnecessary treatments [6,13]. This inadequate specificity
`results from the fact that the IASP CRPS criteria can be met solely
`based on self-reported symptoms (which can be historical), and
`the use of overly liberal decision rules; for instance requiring only
`the report of edema and pain seemingly out of proportion to the
`injury as sufficient to make the diagnosis [6,10,13,23]. Failure of
`the IASP criteria to incorporate motor and trophic features
`
`0304-3959/$36.00 Ó 2010 International Association for the Study of Pain. Published by Elsevier B.V. All rights reserved.
`doi:10.1016/j.pain.2010.04.030
`
`Grun Exh. 1047
`PGR for U.S. Patent No. 9,283,239
`
`
`1
`
`Grün. Exh. 1009
`PGR for U.S. Patent No. 10,052,338
`
`

`

`R.N. Harden et al. / PAINÒ 150 (2010) 268–274
`
`269
`
`commonly associated with CRPS also may adversely impact diag-
`nostic accuracy [6,13].
`To address these limitations, an international consensus meet-
`ing was held in Budapest in 2003 to review issues related to CRPS
`diagnosis with the goal of recommending improvements to the
`IASP criteria (Ref. [12]; see Appendix I for a list of participants).
`The resulting proposal for modified diagnostic criteria for CRPS
`(the ‘‘Budapest Criteria”) was based primarily on empirically-de-
`rived criteria published previously [6,13]. Research evaluating
`these empirically-derived criteria since their publication in 1999
`indicates they result in improved diagnostic consistency between
`clinicians (kappa = 0.66–0.69) compared to existing IASP criteria
`(kappa = 0.43–0.66) [9]. Moreover, these modified criteria result
`in less frequent diagnoses of CRPS [9,27], potentially reflecting im-
`proved specificity. However, no published studies have yet directly
`compared the current standard IASP criteria to these proposed
`Budapest Criteria vis-à-vis diagnostic sensitivity and specificity.
`In a manner similar to our prior published work [6,10], this study
`sought to compare the relative diagnostic efficiency of these alter-
`native diagnostic criteria in discriminating between CRPS and non-
`CRPS neuropathic pain patients.
`
`2. Method
`
`2.1. Design
`
`An international, multi-site, between-subjects design was used
`to compare the ability of the IASP and Budapest diagnostic criteria
`to distinguish between CRPS-I and non-CRPS neuropathic pain
`patients.
`
`2.2. Subjects
`
`Subjects included a series of 113 CRPS-I patients and 47 patients
`with non-CRPS neuropathic pain (‘‘non-CRPS”) who presented for
`evaluation and treatment at the data collection sites. Due to the
`clinical nature of the sample accrual, matching of CRPS and non-
`CRPS groups in terms of sample size, type of initiating injury, or
`other relevant characteristics was not possible. The CRPS sample
`for this study was restricted to CRPS-I patients to maximize sample
`homogeneity given the small proportion of CRPS-II patients in the
`overall sample (13%) which prevented separate analyses by CRPS
`subtype. Non-CRPS neuropathic pain affecting the limbs appeared
`to be the most appropriate comparison group given that CRPS-I is
`associated with signs and symptoms characteristic of other known
`neuropathic pains (e.g., allodynia, hyperalgesia) and evidence that
`CRPS-I may be associated with some type of peripheral nerve in-
`jury [2,24]. All patients in the CRPS-I group met published IASP cri-
`teria for this disorder [23]. Fracture was the single most common
`initiating event in the CRPS group (41.6%), with surgery and crush
`injuries contributing in an additional 32% of CRPS cases. Distribu-
`tion of CRPS patients across the study sites was: Reuth Medical
`Center (Israel; 31%), University of Erlangen-Nuremberg (Germany;
`16.8%), VU University Medical Center (Netherlands; 15.9%), Univer-
`sity Medical Center Mainz (Germany; 12.4%), Rehabilitation Insti-
`tute of Chicago (US; 10.6%), Leiden University Medical Center
`(Netherlands; 9.7%), and Rush University Medical Center (US;
`3.5%). The two German sites evaluated patients primarily with
`short-term CRPS (mean of
`less than 5 months in duration),
`whereas the other study sites evaluated primarily patients with
`long-term CRPS (all means greater than 30 months in duration).
`Diagnoses in the non-CRPS group included peripheral neuropa-
`thy in a single extremity isolated to a specific nerve distribution
`(45%), radiculopathy (30%), diabetic peripheral neuropathy (15%),
`
`and carpal or tarsal tunnel syndrome (10%). Most common initiating
`events for the non-CRPS pain conditions were surgery (50%) and
`crush injuries (30%). Non-CRPS neuropathic pain disorders were
`diagnosed by presence of persistent pain with clear neuropathic
`etiology supported by relevant testing where appropriate (e.g.,
`EMG and clinical examination consistent with pain and symptoms
`restricted to a specific peripheral nerve distribution following
`known injury to that nerve, extremity pain coexisting with known
`diabetes mellitus, pain in a radicular pattern with disk herniation
`confirmed by MRI, etc.). A lower extremity pain location was signif-
`icantly more common in the non-CRPS patients than in the CRPS
`patients (74.4% versus 47.7%, p < .001). Distribution of non-CRPS pa-
`tients across the study sites was: Reuth Medical Center (Israel;
`12.8%), VU University Medical Center (Netherlands; 23.4%), Univer-
`sity Medical Center Mainz (Germany; 21.3%), Leiden University
`Medical Center (Netherlands; 19.1%), Stanford University Medical
`Center (US; 12.8%), and Rush University Medical Center (US; 10.6%).
`
`2.3. Measures
`
`2.3.1. CRPS database checklist
`In order to insure standardized assessment of signs and symp-
`toms across study sites, a CRPS database checklist similar to that
`used in our past multi-site research work was employed [6,7,13].
`This checklist presented a complete list of the signs and symptoms
`used to diagnose CRPS, as well as other signs/symptoms (e.g., tro-
`phic changes, motor abnormalities) reported to be associated with
`the disorder in previous literature but not incorporated in the IASP
`diagnostic criteria [16,23,31–35]. Based on previous suggestions of
`sensory deficits in CRPS patients beyond the region of pain [30], an
`evaluation of light touch sensitivity (categorized as hypoesthetic,
`normal, or allodynic) was included in the CRPS database checklist
`and was assessed bilaterally on the face, chest, and upper and low-
`er extremities. Categorical measures (e.g., presence or absence)
`were used to assess all signs and symptoms because of the poten-
`tial for decreased inter-rater reliability using interval rating scales
`[15,25]. Written standardized procedures and an instructional vi-
`deo to demonstrate the data collection procedures were provided
`with the checklist to maximize uniform assessment across sites.
`Investigators at all sites were highly proficient in English, thereby
`minimizing the potential impact of language issues. Copies of the
`database checklist and instructions are available from the authors.
`
`2.3.2. Visual analog pain intensity scale
`At all study sites, a 100 mm visual analog scale (VAS) was used
`to assess overall pain intensity. This VAS was anchored with ‘‘no
`pain” and ‘‘worst possible pain” in the patients’ native language.
`
`2.4. Procedures
`
`For all patients in both groups, the study physician conducted
`an evaluation of signs and symptoms using the CRPS checklist de-
`scribed above. This involved obtaining a patient history to assess
`symptoms, as well as conducting a physical examination to assess
`signs. As part of the physical examination, an evaluation of
`mechanical wind-up (to repetitive light pinprick) was conducted
`using a punctate mechanical stimulator (diameter: 0.2 mm; force:
`256 mN) provided to all study sites by one of the authors (C.M.)
`and based on procedures described previously [8,29]. To better
`characterize the degree of temperature asymmetry, temperatures
`in the center of the affected hand (palmar surface) or foot (plantar
`surface) and the contralateral hand/foot were determined while in
`a room temperature environment (minimum 30 min of acclimati-
`zation) using standard infrared (IR) thermometers provided to all
`study sites (Exergen Corp., Watertown, MA). This simple tempera-
`
`2
`
`

`

`270
`
`R.N. Harden et al. / PAINÒ 150 (2010) 268–274
`
`ture assessment methodology was designed solely to provide
`objective documentation of the clinically-determined temperature
`asymmetry used in making the diagnoses. Repeated assessment of
`temperature asymmetry over time would be necessary to optimize
`the accuracy of these temperature evaluations (e.g., [18]).
`Thermal Quantitative Sensory Testing (tQST; Medoc TSA-II, Me-
`doc Inc., Tel Aviv, Israel) data were available for patients at the
`study sites in Israel and Germany, as well as for a subset of patients
`at the Rehabilitation Institute of Chicago and Stanford sites. tQST
`data were available for a total of 61 CRPS patients and 13 non-CRPS
`patients. A standardized protocol was used across all study sites
`obtaining these data. The tQST protocol employed a computer-con-
`trolled 30  30 mm Peltier thermistor probe that was used to eval-
`uate cold and warmth perception thresholds and heat pain
`threshold (mean of three trials each) using the method of limits.
`For upper extremity CRPS, the probe was placed sequentially on
`three adjacent sites on the volar forearm of the affected extremity.
`For lower extremity CRPS, the probe was similarly placed on three
`adjacent sites on the dorsal mid-calf. Prior to each trial, the probe
`was maintained at an adaptation temperature of 32 °C.
`All study procedures were approved by the appropriate ethical
`review boards at participating institutions.
`
`2.5. Statistical analysis
`
`Analyses were conducted using the SPSS for Windows Version
`17 statistical package (SPSS Inc., Chicago, IL). Preliminary analyses
`used t-tests to compare mean values across diagnostic groups and
`the nonparametric phi correlation to evaluate direction and
`strength of associations of categorical measures across groups.
`For correlational analyses of highly skewed continuous variables,
`Spearman’s rho was used to minimize the influence of the skewed
`distribution. The underlying rationale for the approach taken in pri-
`mary analyses is detailed in our similar prior work (see Refs. [6,10]).
`Primary analyses derived measures of diagnostic efficiency (see be-
`low) to provide relative comparisons between the IASP and Buda-
`pest Criteria in distinguishing CRPS from non-CRPS neuropathic
`pain patients. Similar models have been used in validation of diag-
`nostic criteria for headache and psychiatric disorders [21,22].
`In analyses of diagnostic efficiency, IASP criteria were evaluated as
`written and typically applied in clinical practice, i.e., criteria can be
`met by presence of self-reported symptoms or signs noted during
`the physical examination. For the Budapest Criteria (detailed in Ref.
`[14]), both the clinical decision rules (CRPS characteristics present in
`at least 3 of 4 symptom categories and at least 2 of 4 sign categories)
`and research decision rules (CRPS characteristics present in all 4
`symptom categories and at least 2 of 4 sign categories) were evalu-
`ated. Appendix II summarizes the Budapest clinical criteria. Based
`on fulfillment of the various diagnostic criteria as a function of patient
`group membership, several indices of diagnostic efficiency were de-
`rived, including sensitivity, specificity, positive predictive power
`(PPP), and negative predictive power (NPP). Sensitivity is defined as
`true positive rate/true positive + false negative rates, reflecting the
`percentage of true positive (CRPS) cases classified accurately. Speci-
`ficity is defined by true negative rate/true negative + false positive
`rates, and reflects the proportion of true negative (non-CRPS) cases
`classified accurately. Potentially of more importance clinically, given
`the need to maximize probability of correct diagnosis when actual
`disease status is unknown, are PPP and NPP [19]. In this study, PPP
`indicated the probability of accurate categorization to the CRPS group
`based on the diagnostic criteria being tested, whereas NPP indicated
`the probability of accurate categorization to the non-CRPS neuro-
`pathic pain group. Both PPP and NPP are dependent on the prevalence
`of the targeted disorder (CRPS) in the population being examined, and
`these were derived as described by Meehl and Rosen [20]. PPP was
`
`defined as: (CRPS prevalence true positive rate)/((CRPS preva-
`lence true positive rate) + (1 CRPS prevalence false positive
`rate)). NPP was defined as: ((1 CRPS prevalence) true negative
`rate)/((1 CRPS prevalence true negative rate) + (CRPS preva-
`lence false negative rate)). PPP and NPP were derived for scenarios
`in which 50% and 70% of patients referred to rule in or out CRPS actu-
`ally have the disorder, a prevalence range like that which might occur
`in a specialty pain clinic to which suspected cases of CRPS are often
`referred [26]. These four indicators of diagnostic efficiency were con-
`trasted across different criteria to evaluate relative accuracy and
`likely diagnostic utility of each.
`Secondary analyses were conducted to evaluate the extent to
`which each of the diagnostic components included in the Budapest
`Criteria (sensory, vasomotor, sudomotor/edema, and motor/tro-
`phic) contribute to diagnostic accuracy. This was addressed by con-
`trasting the ability of each individual diagnostic component to
`discriminate between CRPS and non-CRPS neuropathic pain patients
`with the ability of all four components to discriminate these groups
`simultaneously. Continuous component scores were derived reflect-
`ing the total number of signs and symptoms observed in each of the
`four categories above for each subject (total score combining all four
`components in CRPS = 12.0 ± 2.59, non-CRPS = 5.4 ± 3.59; t(164) =
`12.84, p < .001).
`These component scores were then included as independent
`variables (individually and in combination) in a series of binary lo-
`gistic regressions, with resulting classification tables used to derive
`sensitivity and specificity values.
`All analyses used the maximum number of cases available and a
`two-tailed probability value of p < .05 was used as the criterion for
`statistical significance. All means are presented as mean ± SD. For
`highly skewed continuous variables, medians are presented with
`interquartile range.
`
`3. Results
`
`3.1. Preliminary analyses
`
`Table 1 summarizes differences between the CRPS and non-
`CRPS groups with regards to demographics and non-diagnostic
`clinical characteristics. The CRPS group was significantly younger,
`more likely to be female, and had experienced their pain condition
`for a significantly shorter time than the non-CRPS patients. Overall
`clinical pain intensity was statistically comparable across groups.
`However, CRPS patients displayed significantly greater acute pain
`sensitivity (lower heat pain threshold) on tQST evaluation com-
`pared to non-CRPS patients. Additionally, CRPS patients were sig-
`nificantly more sensitive to non-noxious warmth (lower warmth
`perception threshold) and cold (higher cold perception threshold)
`during tQST evaluation compared to non-CRPS patients. Sensitivity
`to light touch was comparable between groups for most body
`regions evaluated with two exceptions: significantly greater fre-
`quency of abnormal sensation (hypoesthesia) on the contralateral
`side of the face and the unaffected (contralateral) thigh among
`non-CRPS patients. In general, rates of abnormal light touch sensi-
`tivity were relatively low in both groups.
`Table 2 compares the two groups with regards to CRPS signs and
`symptoms. Significant differences were observed across groups on
`nearly every diagnostic characteristic. Although hypoesthesia and
`altered local reflexes were significantly more common in the non-
`CRPS group, other neurological signs and symptoms usually associ-
`ated with CRPS were more common in the CRPS group. Quantitative
`evaluation of
`temperature asymmetry by IR thermometry
`supported the clinical examination, indicating that the majority of
`CRPS patients with temperature asymmetry exhibited a ‘‘cold CRPS”
`
`3
`
`

`

`R.N. Harden et al. / PAINÒ 150 (2010) 268–274
`
`271
`
`Table 1
`Sample characteristics by diagnostic subgroup.
`
`Variable
`
`Diagnosis
`
`CRPS-I
`(n = 113)
`
`Non-CRPS
`(n = 47)
`
`Table 2
`Diagnostic signs and symptoms by subgroup.
`
`Variable
`
`44.7
`68.1
`Gender (female %)**
`39.3 ± 15.47 53.8 ± 15.28
`Age (years)**
`14.2 (42.1) 41.3 (117.99)
`Pain duration (median (IQR) in months)**
`47.7
`74.4
`Affected extremity (% lower extremity)**
`50.5
`57.1
`Affected side (% right)
`53.3 ± 25.83 49.4 ± 26.38
`VAS pain intensity (0–100)
`Affected side cold perception threshold (°C)*
`28.4 ± 3.78 25.8 ± 5.19
`Affected side warmth perception threshold (°C)** 36.7 ± 3.33 41.5 ± 3.61
`Affected side heat pain threshold (°C)**
`42.3 ± 3.99 46.1 ± 2.48
`Light touch sensitivity (% abnormal)
`Face – affected side
`Face – unaffected side*
`Chest – affected side
`Chest – unaffected side
`Bicep – affected side
`Bicep – unaffected side
`Thigh – affected side
`Thigh – unaffected side**
`
`4.7
`4.7
`4.9
`2.4
`9.5
`4.8
`26.2
`11.9
`
`4.5
`0.0
`3.6
`0.0
`14.2
`0.9
`17.0
`0.9
`
`Note: summary statistics are presented as percentages or mean ± SD. Abnormalities
`in light touch perception reflect either hypoesthetic or allodynic responses as
`judged by the clinician.
`* p < .05.
`** p < .01.
`
`pattern, with the affected extremity on average more than 0.6 °C
`colder than the unaffected extremity, consistent with the diagnostic
`temperature asymmetry cutoff suggested in previous work [5]. The
`CRPS group was noted to have directionally greater mechanical
`wind-up to repetitive light pinprick, although these wind-up data
`failed to achieve even the level of a statistical trend.
`The significantly higher frequency of diminished active range of
`motion (AROM) in CRPS compared to non-CRPS patients noted in
`Table 2 reflects impairments throughout the affected extremity.
`Quantitative goniometric assessment of AROM bilaterally in the
`affected region (i.e., elbow and wrist for upper extremity CRPS, knee
`and ankle for lower extremity CRPS) indicated that elbow/knee flex-
`ion (affected side: 115.3° ± 34.16; unaffected side: 127.0° ± 27.87),
`wrist/ankle flexion (affected side: 35.6° ± 33.30; unaffected side:
`53.9° ± 33.27),
`and wrist/ankle
`extension
`(affected
`side:
`37.4° ± 30.90; unaffected side: 58.8° ± 26.46) were significantly
`reduced in the CRPS affected side compared to the unaffected side
`(ts > 4.1, ps < .001).
`Pain duration may impact the pattern of CRPS characteristics. For
`example, patients with longer duration CRPS displayed a signifi-
`cantly larger number of sensory signs and symptoms (hyperalgesia,
`allodynia, hyperesthesia; Spearman’s rho = 0.22, p < .05) and signif-
`icantly fewer vasomotor signs and symptoms (skin temperature and
`color changes; Spearman’s rho = 0.24, p < .05). Prior work sug-
`gested that CRPS patients over time may transition from a predom-
`inately ‘‘warm CRPS” pattern (affected extremity warmer with
`reddish skin color) to a predominately ‘‘cold CRPS” pattern (affected
`extremity colder with pale or bluish skin color; [3]). Consistent with
`this idea, among CRPS patients in the current study exhibiting nota-
`ble temperature asymmetry detectable on clinical examination,
`those displaying a ‘‘cold CRPS” pattern had experienced CRPS for a
`significantly longer duration than those with a ‘‘warm CRPS” pattern
`(median (IQR) for cold CRPS: 20.1 (38.2) months, warm CRPS: 3.9
`(19.1) months; Mann–Whitney U = 392.00, p < .05). On tQST evalu-
`ation, patients with CRPS of longer duration exhibited significant
`hypoesthesia on evaluation of cold (Spearman’s rho = 0.37,
`p < .01) and warmth perception thresholds (Spearman’s rho = 0.33,
`p < .05) relative to patients with CRPS of shorter duration.
`
`Self-reported symptoms (% yes)
`Hyperesthesia (allodynia, hyperpathia)**
`Hypoesthesia (localized ‘‘numbness”)**
`Temperature asymmetry**
`Skin color asymmetry**
`Sweating asymmetry**
`Asymmetric edema**
`Trophic changes**
`Motor changes**
`
`Signs observed on examination (% yes)
`Hyperalgesia to pinprick**
`Hypoesthesia to light touch*
`Allodynia (any stimulus)**
`Allodynia to cold**
`Allodynia to heat
`Allodynia to light touch
`Allodynia to vibration*
`Allodynia to deep joint pressure**
`Windup to series of 10 pinpricks (0–100)
`Temperature asymmetry by palpation**
`% Affected side colder
`Mean asymmetry by IR thermometry (°C)*
`Skin color asymmetry**
`% Affected side red
`% Affected side blue/pale
`Sweating asymmetry**
`% Affected side increased
`Asymmetric edema**
`Trophic changes (any)**
`Nails
`Hair
`Skin
`Motor changes (any)**
`Weakness
`Tremor*
`Dystonia*
`Decreased active range of motion**
`Altered reflexes in affected area*
`
`Diagnosis
`
`CRPS-I
`(n = 113)
`
`Non-CRPS
`(n = 47)
`
`90.2
`38.5
`86.6
`91.1
`62.5
`89.2
`75.0
`88.3
`
`81.5
`57.7
`70.5
`63.6
`20.8
`68.8
`40.0
`67.6
`16.4 ± 16.88
`69.4
`62.3
`0.62 ± 1.97
`83.9
`41.3
`43.5
`43.8
`80.9
`63.5
`68.5
`42.9
`54.5
`45.5
`79.3
`85.5
`30.1
`26.5
`80.0
`50.8
`
`63.8
`65.0
`38.3
`27.7
`15.2
`40.4
`38.3
`46.7
`
`43.5
`77.5
`29.8
`10.5
`6.7
`52.6
`10.5
`26.7
`13.9 ± 17.61
`14.9
`85.7
`0.11 ± 1.04
`36.2
`35.3
`29.4
`10.6
`60.0
`24.2
`29.8
`27.8
`66.7
`66.7
`40.0
`72.2
`5.6
`5.6
`37.8
`86.7
`
`Note: summary statistics are presented as percentages or mean ± SD. Percentages
`for specific types/direction of allodynia, sweating, temperature, trophic, and motor
`changes reflect percentage of those patients who were positive for this sign cate-
`gory. The negative thermometric asymmetry value for the CRPS group indicates
`that on average the affected side was colder.
`* p < .05.
`** p < .01.
`
`3.2. Diagnostic efficiency
`
`Indices of diagnostic efficiency reflecting the relative ability of
`the different CRPS criteria to discriminate between CRPS-I and
`non-CRPS neuropathic pain patients are summarized in Table 3.
`The current IASP criteria resulted in excellent sensitivity, but poor
`specificity. Table 3 indicates that the Budapest clinical criteria
`retained excellent sensitivity that was nearly identical to the IASP
`criteria, but also displayed much improved specificity compared to
`the latter criteria. Given the intent of the Budapest research criteria
`to maximize specificity (minimize false positives) at the expense of
`sensitivity, it is not surprising that these criteria had the highest
`specificity but also the lowest sensitivity of the various criteria
`examined. Consistent with sensitivity and specificity findings, the
`IASP criteria showed the lowest probability of accurate CRPS diag-
`nosis (PPP) and the Budapest research criteria showed the highest
`probability of accurate diagnosis. The Budapest clinical criteria
`were clearly better in terms of overall diagnostic accuracy (balanc-
`ing PPP and NPP) compared to the IASP criteria. Of note, and some-
`what surprising, was the fact that PPP was only marginally higher
`
`4
`
`

`

`272
`
`R.N. Harden et al. / PAINÒ 150 (2010) 268–274
`
`Table 3
`Comparison of diagnostic efficiency of IASP CRPS criteria versus proposed modified (Budapest) criteria for discriminating between CRPS-I and non-CRPS neuropathic pain.
`
`Diagnostic criteria
`
`Sensitivity
`
`Specificity
`
`Assume 70% CRPS prevalence
`
`Assume 50% CRPS prevalence
`
`IASP
`Budapest clinical
`Budapest research
`
`1.00
`0.99
`0.78
`
`0.41
`0.68
`0.79
`
`PPP
`
`0.80
`0.88
`0.90
`
`NPP
`
`1.00
`0.97
`0.60
`
`PPP
`
`0.63
`0.76
`0.79
`
`NPP
`
`1.00
`0.99
`0.78
`
`Note: positive predictive power (PPP) and negative predictive power (NPP) are dependent on the assumed prevalence of CRPS in the population being considered. For
`illustrative purposes, two scenarios are presented in which either 70% or 50% of patients referred to rule CRPS in or out actually have the disorder. IASP = diagnosis based on
`presence of CRPS signs or symptoms using the International Association for the Study of Pain criteria.
`
`for the Budapest research criteria compared to the Budapest clini-
`cal criteria.
`Analyses were conducted to evaluate the relative contributions
`of each of the diagnostic components included in the Budapest
`Criteria (sensory, vasomotor, sudomotor/edema, and motor/
`trophic) to overall diagnostic accuracy. Table 4 indicates that while
`each of the four individual diagnostic components are reasonably
`sensitive, they are not as specific (0.57–0.71) as the combination
`of all components. Of the four diagnostic components, vasomotor
`characteristics appear to be the most sensitive for distinguishing be-
`tween CRPS and non-CRPS neuropathic pain, but lack the specificity
`of the combined components. Combining all four diagnostic compo-
`nents in diagnostic decision making maximizes sensitivity (0.95),
`but also improves specificity substantially (0.81). This supports
`inclusion of all four components in the diagnostic decision making
`process as suggested in the Budapest Criteria. It should be noted that
`the higher specificity values exhibited in these analyses compared to
`those involving the Budapest Criteria reported in Table 3 resulted
`from use of continuous component scores in the former. While con-
`tinuous sign/symptom scores may optimize statistical prediction,
`they do not reflect the clinical reality of having to set a cutoff for
`making diagnostic decisions as in the actual Budapest Criteria.
`
`4. Discussion
`
`The current study replicated previous findings suggesting rela-
`tively poor diagnostic accuracy for the extant IASP diagnostic crite-
`ria for CRPS. Results indicated that the IASP criteria as written (i.e.,
`criteria can be met by either self-reported symptoms or objective
`signs) were highly sensitive but had poor specificity (0.41). This
`finding is consistent with prior results [6,10], which found specific-
`ity values of 0.36 and 0.27, respectively, for the IASP criteria. These
`findings indicate that current IASP criteria may result in a rela-
`tively high rate of false positive diagnoses, potentially leading to
`unnecessary or inappropriate treatments [12]. Unlike the IASP
`criteria, proposed modified diagnostic criteria (‘‘Budapest Criteria”;
`[12]) require presence of both signs and symptoms of CRPS to
`make the diagnosis, a change that should reduce false positive
`diagnoses.
`Prior work suggested that the Budapest Criteria were associated
`with improved diagnostic consistency between clinicians (kap-
`
`Table 4
`Comparison of the diagnostic efficiency of individual Budapest Criteria diagnostic
`components versus the combination of all diagnostic components.
`
`Criterion
`
`Sensitivity
`
`Specificity
`
`All sign/symptom factor scores
`Sensory factor only
`Vasomotor factor only
`Sudomotor/edema factor only
`Motor/trophic factor only
`
`0.95
`0.83
`0.94
`0.85
`0.86
`
`0.81
`0.57
`0.68
`0.71
`0.67
`
`pa = 0.66–0.69) compared to existing IASP criteria (kappa = 0.43–
`0.66) [9]. To build on this work, the current study provided the first
`direct comparison of the Budapest Criteria to existing IASP diagnos-
`tic criteria for CRPS regarding relative diagnostic efficiency. The
`Budapest clinical criteria provided excellent sensitivity nearly iden-
`tical to that for the IASP criteria (0.99), but with substantially
`improved specificity (0.68). Examination of positive and negative
`predictive power indicated that under conditions in which CRPS
`diagnoses were common (e.g., a clinic receiving many cases of
`suspected CRPS), CRPS diagnoses using the Budapest clinical criteria
`were likely to be accurate 88% of the time, with non-CRPS patients
`correctly diagnosed 97% of the time. These values represent
`improved accuracy in CRPS diagnosis compared to existing IASP
`criteria. Overall, findings in this study suggest that the Budapest
`clinical criteria provide an incremental improvement in diagnostic
`accuracy compared to the current IASP criteria. Findings summa-
`rized in Table 4 suggest that all four diagnostic components included
`in the Budapest Criteria contribute to improved specificity. This im-
`proved specificity and diagnostic accuracy might account for the less
`frequent diagnosis of CRPS using the Budapest Criteria [9].
`A unique feature of the Budapest Criteria is provision of two sets
`of decision rules, one for clinical diagnoses (placing relatively
`greater emphasis on sensitivity) and another for research purposes
`(emphasizing specificity to reduce false positives in research sam-
`ples) [12]. In the current study, the Budapest research criteria dem-
`onstrated the highest specificity of the three sets of criteria
`examined, consistent with their designed purpose. Given that the
`Budapest research criteria decision rules require the presence of
`an extra symptom to reduce false positives, one would expect that
`these decision rules would lead to a notably higher probability that
`a