Psychological factors can cause false pain classification on painDETECT

1 Introduction

The detrimental impact of neuropathic pain (NeP), defined as “pain caused by a lesion or disease of the somatosensory nervous system” [1] on patients’ well-being, health, and quality of life, together with the socio-economic burden, have been well documented [2], [3], [4], [5]. Psychological factors such as depression and anxiety are reported to be much higher in patients with NeP compared to patients with nociceptive pain [5], [6], [7] and when compared to the general population [8], [9]. Hence, early identification and targeted pharmacological treatment of NeP, as well as treatment addressing psychological comorbidities, is crucial in the management of patients suffering from this condition.

While clinical examination in combination with diagnostic tests remains the “gold standard” for the diagnosis of NeP [10], [11], several screening tools have been recommended for its identification and to give guidance to further diagnostic assessment. The most commonly used include the Douleur Neuropathique en 4 questions (DN4) [12], the Leeds Assessment of Neuropathic Symptoms and Signs scale (LANSS) [13], the self-completed version of LANSS (S-LANSS) [14] and the painDETECT questionnaire (PD-Q) [5]. The questionnaires differ in their design; the DN4 and LANSS include a physical sensory examination whereby the S-LANSS and the PD-Q are self-report tools.

The PD-Q was originally developed and validated in a German cohort of patients with nociceptive and NeP, demonstrating a sensitivity and specificity of 85% and 80%, respectively in identifying NeP [5]. The PD-Q has been widely used as a NeP screening tool as well as a tool for characterisation of patients’ pain profiles [15]. Based on PD-Q responses, sub-groups of patients have been identified in large cohorts of patients with painful lumbar radiculopathy/radicular pain [16], diabetic neuropathy and postherpetic neuralgia [17] and fibromyalgia [18]. The characterisation of pain profiles may assist in the proposed mechanism or symptom-based classification approach of NeP conditions [19], [20], [21].

Contrary to the DN4, LANSS and S-LANSS, in which responses to sensory descriptors are binary (yes, no), sensory descriptors in the PD-Q are weighted from 0 to 5, with “0” indicating the person “never feels the sensation” and “5” indicating the person feels the sensation “very strongly”. It is likely that responses to the PD-Q sensory descriptors are influenced by psychological factors, such as catastrophizing or anxiety [22], [23]. These factors could potentially contribute to an overall higher score of PD-Q and a false positive identification of NeP. Consequently, if psychological factors affect the responses to PD-Q, the findings may challenge the use of PD-Q as a tool for sensory phenotyping.

To further explore the relationship between PD-Q and psychological factors we aimed in this study to investigate, in a cohort of patients with persistent pain (i) the relationship between psychological factors (catastrophizing, anxiety, depression and stress) and the total PD-Q score and (ii) if psychological factors are associated with false positive identifications of NeP on the PD-Q compared to clinically diagnosed NeP.

2 Methods

3 Results

3.1 Characteristics of study population

Out of 1,101 patients, 449 patients had to be excluded from the analysis due to missing data in any of the questionnaires (missing data for PD-Q n=244; PCS n=138; DASS n=239, BPI Pain n=99; BPI Interference n=114). The remaining 652 patients (mean age 51 [SD14] years, range 18–88; 57% females) were included (Fig. 1A). The demographic and clinical characteristics for the cohort are shown in Table 1. Based on PD-Q scoring, NeP was likely present in 285 (44%) patients.

Fig. 1:

Flow chart of patients with and without neuropathic pain (NeP) based on the painDETECT questionnaire (PD-Q) (A) and clinical diagnosis (B)

Table 1:

Summary statistics of all variables for all patients (total) and for patients with negative and positive painDETECT results.

	Total n=652			PD-Q negative n=367			PD-Q positive n=285
	n	%	Mean (SD)	n	%	Mean (SD)	n	%	Mean (SD)
Age (years)			51.0 (13.6)			52.6 (14.2)			48.8 (12.4)
Sex
Female	369	56.60		196	53.12		173	46.88
Male	283	43.40		171	60.42		112	39.58
painDETECT	652	100	17.2 (7.5)	367	56.28	11.8 (4.4)	285	43.71	24.3 (4.1)
PCS total			22.6 (13.3)			19.7 (12.2)			26.2 (13.7)
Not catastrophizing	449	68.87		285	63.47		164	36.53
Catastrophizing	203	31.13		82	40.39		121	59.61
DASS depression			15.6 (12.4)			13.4 (11.6)			18.5 (13.0)
Normal	251	38.50		167	66.53		84	33.47
Mild	82	12.58		48	58.54		34	41.46
Moderate	126	19.33		67	53.17		59	46.83
Severe	57	8.74		28	49.12		29	50.88
Extremely severe	136	20.86		57	41.91		79	58.09
DASS anxiety			10.8 (9.1)			8.7 (8.2)			13.4 (9.5)
Normal	279	42.79		194	69.53		85	30.47
Mild	62	9.51		34	54.84		28	45.16
Moderate	136	20.86		66	48.53		70	51.47
Severe	58	8.90		26	44.83		32	55.17
Extremely severe	117	17.94		47	40.17		70	59.83
DASS stress			16.6 (11.0)			14.2 (10.1)			19.7 (11.4)
Normal	325	49.85		219	67.38		106	32.62
Mild	81	12.42		41	50.62		40	49.38
Moderate	90	13.80		48	53.33		42	46.67
Severe	85	13.04		34	40.00		51	60.00
Extremely severe	71	10.89		25	35.21		46	64.79
BPI pain			6.0 (1.7)			5.5 (1.6)			6.7 (1.5)
BPI interference			6.8 (2.0)			6.1 (2.0)			7.6 (1.7)

1. PD-Q=painDETECT; PCS=Pain Catastrophizing Scale; DASS=Depression Anxiety and Stress Score; BPI=Brief Pain Inventory.

3.1.1 Education level

Forty-six percent of patients had either completed or partially completed high school. Thirty-seven percent of patients had undergone additional training after high school (e.g. apprenticeship). Fourteen percent completed either undergraduate or postgraduate university studies.

3.1.2 Employment status

Thirty-three percent of patients were unemployed due to pain. Eighteen percent of the group were retired while 17% worked full time. Some patients were employed part-time (11%) or had home duties (11%) and a smaller proportion (10%) were either students, volunteers, unemployed for other reasons, on paid leave or employment status unknown.

3.1.3 Pain duration and location

Thirty-seven percent of patients reported pain for more than 10 years (Fig. 2). Fifty-five percent of patients reported the lower back/lumbosacral area as their main pain area (Fig. 3).

Fig. 2:

Pain duration for the total patient cohort in years (yrs).

Fig. 3:

Pain location for the total patient cohort.

3.3 Subgroup analysis

3.3.1 Clinical diagnosis compared to the painDETECT classification

The medical records of 285 patients having been identified by PD-Q as NeP positive were reviewed (Fig. 1B). Seven individuals had to be excluded from all summaries and analyses as they had only attended the group education program, and were not individually assessed by a clinician at the pain centre, so no diagnosis was made. Out of the remaining 278 patients (mean age 49 (SD12) years, range 18–81; 62% females), 138 patients (50%) had not been given a clinical diagnosis of NeP (false positive) and 140 patients (50%) had been given a clinical diagnosis of NeP (true positive) (Fig. 1B; Table 3). The patients’ pain presentations are outlined in Fig. 4.

Table 3:

Summary statistics of all variables for patients scoring positive on painDETECT with and without clinically diagnosed neuropathic pain (n=278).

	No clinically diagnosed NeP (false positive)			Clinically diagnosed NeP (true positive)
	n=138 (50%)			n=140 (50%)
	n	%	Mean (SD)	n	%	Mean (SD)
Age (years)			48.0 (12.3)			49.4 (12.4)
Sex
Female	86	50.00		86	50.00
Male	52	49.06		54	50.94
PCS total
Not catastrophizing	75	47.17		84	52.83
Catastrophizing	63	52.94		56	47.06
DASS depression
Normal	34	41.98		47	58.02
Mild	14	43.75		18	56.25
Moderate	20	51.28		19	48.72
Severe	24	50.00		24	50.00
Extremely severe	46	58.97		32	41.03
DASS anxiety
Normal	28	34.57		53	65.43
Mild	15	53.57		13	46.43
Moderate	22	42.31		30	57.69
Severe	28	58.33		20	41.67
Extremely severe	45	65.22		24	34.78
DASS stress
Normal	36	42.86		48	57.14
Mild	16	39.02		25	60.98
Moderate	32	56.14		25	43.86
Severe	27	52.94		24	47.06
Extremely severe	27	60.00		18	40.00
BPI pain			6.7 (1.5)			6.7 (1.5)
BPI interference			7.6 (1.6)			7.5 (1.8)

1. NeP=neuropathic pain; PCS=Pain Catastrophizing Scale; DASS=Depression Anxiety and Stress Score; BPI, Brief Pain Inventory.

Fig. 4:

The number of patients with a clinical diagnosis of neuropathic pain (clinical NeP) and no neuropathic pain (no clinical NeP) for the most common pain presentations (LBP=low back pain; CRPS=Complex Regional Pain Syndrome).

3.3.2 Association of psychological factors with false positive identification of NeP on painDETECT

The percentages of patients catastrophizing and displaying extremely severe depression, anxiety, and stress were higher in the false positive patient group compared to the percentages in the true positive patient group (Table 3). DASS anxiety was found to be significantly associated with a false positive result (p=0.0036) (Table 4). Patients with extremely severe (OR=3.51, 95% CI=1.77–6.96, p=0.0003) or severe (OR=2.62, 95% CI=1.25–5.49, p=0.0109) anxiety were particularly likely to have a false positive PD-Q. Catastrophizing, depression and stress were not found to be significantly associated with a false positive outcome (all p>0.05) (Table 4).

Table 4:

Age and sex adjusted odds ratios, 95% confidence intervals and p-values from analysing neuropathic pain outcome among those with a positive painDETECT result (n=278).

Variables	OR	95% CI	p-Value
PCS total
Catastrophizing vs. Not	1.22	0.75–1.99	0.4168
DASS depression
Mild vs. Normal	1.05	0.45–2.47	0.3439
Moderate vs. Normal	1.4	0.63–3.09
Severe vs. Normal	1.36	0.65–2.85
Extremely severe vs. Normal	1.93	1.00–3.72
Dass anxiety
Mild vs. Normal	2.19	0.92–5.26	0.0036
Moderate vs. Normal	1.38	0.68–2.84
Severe vs. Normal	2.62	1.25–5.49
Extremely severe vs. Normal	3.51	1.77–6.96
DASS stress
Mild vs. Normal	0.85	0.4–1.83	0.2134
Moderate vs. Normal	1.67	0.84–3.33
Severe vs. Normal	1.46	0.71–2.97
Extremely severe vs. Normal	1.96	0.93–4.13
BPI pain
For a one point increase	0.995	0.85–1.17	0.9556
BPI interference
For a one point increase	1.04	0.90–1.20	0.5699

1. OR=odds ratio; CI=confidence interval; PCS=Pain Catastrophizing Scale; DASS=Depression Anxiety and Stress Score; BPI=Brief Pain Inventory. Bold numbers represent statistical significance.

Measures of BPI were similar between the two groups (Table 3), and neither pain (p=0.9556) or interference (p=0.5669) were found to be significantly associated with a false-positive PD-Q outcome (Table 4).

4 Discussion

This study revealed that in a large cohort of patients with persistent pain, psychological factors such as catastrophizing, depression, anxiety, and stress were all associated with a higher score on the PD-Q. For patients classified by PD-Q as having NeP, only 50% of these patients had a clinical diagnosis of NeP. Anxiety measured on the DASS was significantly associated with a false positive classification of NeP.

This retrospective study captured a real life data set of a large patient population with persistent pain. Based on self-reported data the majority of our patients indicated the lower back as their main pain area, followed by the neck/shoulder and widespread body pain. The large proportion of patients with musculoskeletal conditions is reflective of the Australian population, as musculoskeletal conditions and back pain are among the leading causes of total burden of disease in Australia [33]. Considering that 74% of our patients had symptoms for longer than 3 years and pain significantly interfered with their quality of life, it is not surprising that up to one third of patients displayed severe or extremely severe levels of anxiety, depression and stress and behavior of catastrophizing. The PD-Q identified 44% of the total cohort as having likely NeP, which is comparable with results of a previous study in which 37% of 400 chronic pain patients attending a multidisciplinary pain management program were identified by PD-Q as having NeP [31].

There is a large body of literature discussing how persistent pain interferes with psychological and physical function in patients with nociceptive pain and NeP and vice versa [28], [29], [30], [31]. It is difficult to determine the sequence of events, that is, if the psychological traits developed as a consequence to ongoing pain and disability or if they existed prior to development of the pain. Our findings suggest that if someone displays severe anxiety, depression, stress or catastrophizing, they are more likely to score higher on the PD-Q. The strong correlation between the various psychological variables in our study further suggests that any of these traits, be it anxiety, catastrophizing or depression and stress, may drive a higher score on PD-Q.

The impact of psychological factors on PD-Q scoring should be considered when using the PD-Q as a tool for sensory profiling [16], [17], [18], [34]. Patients’ sensory profiles, i.e. symptoms and signs of hyperalgesia, allodynia or numbness may be associated with the specific underlying pain mechanisms [21], [35]. Some evidence is emerging that sub-grouping patients based on their sensory profiles leads to a superior effect of medications in the specific subgroups compared to the entire patient cohort [35], [36] or compared to a control group [37], [38], [39], [40]. In two of these studies pain phenotyping was based on the PD-Q [37], [38]. For the majority of these patients their anxiety and depression score, measured by the Hospital Anxiety and Depression Scale, fell within the normal range, hence psychological factors may not have had an impact on the PD-Q classification. However, this may have been different in previous large epidemiological studies in which a cut-off score of three (moderate) was used to determine a clinically relevant response to PD-Q sensory descriptors [16], [17], [18]. In the patient cohort with fibromyalgia (n=3,035) 15% of patients had severe depression and 18% severe anxiety [18] which may have influenced the patients’ responses to PD-Q descriptors and may have led to a “false patient classification”. In comparison, in patients with painful radiculopathy (n=2,094) [16], diabetic neuropathy or postherpetic neuralgia [17] the percentage of patients with severe depression (4–6%) and severe anxiety (6–9%) was lower. Consequently, one should consider that if patients are wrongly classified, based on PD-Q profiling, they may be at risk of being prescribed an inappropriate medication for their clinical condition.

Out of 278 patients identified by PD-Q as having NeP, 50% did not have a clinical diagnosis of NeP, hence they were false positive. Anxiety was significantly associated with a false positive result. Interestingly, an epidemiological study by Freynhagen et al. in 7,772 patients with low back pain reported that 10.5% of the total population had anxiety, with four times as many patients in the NeP group compared to the nociceptive group (odds ratio 4.5; CI 3.6; 5.6) [5]. This leads to the question of whether these patients, classified as having NeP, did indeed have NeP, or if their anxiety trait influenced their PD-Q scoring. The large proportion of false positive NeP classification in our study suggests that the diagnostic accuracy of PD-Q may be compromised by psychological traits. While several PD-Q validation studies have been conducted [5], [41], [42], [43], most of them did not capture the patients’ psychological profile [5], [42], [43]. The potential concern about the impact of psychological factors on PD-Q plus previous reports of limited diagnostic accuracy of the PD-Q [23], [44], [45], [46], [47], [48] may question the suitability of PD-Q as a screening tool for NeP.

Over 50% of patients who were identified by PD-Q as having NeP had the clinical diagnosis of low back pain without radicular leg pain. These were interpreted by the clinicians as having nociceptive pain. We acknowledge the concept that low back pain may have a neuropathic component [49] and that several studies reported the presence of NeP in patients with low back pain [2], [5], [50], [51], [52], [53], however these reports were based on either a retrospective data review [52] or NeP screening tools [2], [5], [50], [51], [53], including the PD-Q [2], [5], [53]. To our knowledge, no study has yet demonstrated the presence of NeP in patients with low back pain, using objective measures to demonstrate the presence of a nerve lesion.

Numerous studies reported that patients with NeP have higher pain intensity and higher anxiety/depression scores compared to patients with nociceptive pain characteristics [5], [6], [7], however this was not the case in our patient cohort. Pain intensity and interference were equal in patients with and without clinical diagnosis of NeP and, in fact, catastrophizing and severe/extremely severe anxiety as well as depression and stress level showed a higher percentage in the group without NeP compared to the group with NeP. This observation raises the question if patients with NeP might have developed better coping strategies for their pain conditions. One reason may be that patients with NeP have a clear reason for their pain while patients with low back pain often don’t know the cause of their pain, as documented by Daniel et al. [28] comparing psychological and physical function in patients with post-herpetic neuralgia and patients with nociceptive low back pain.

5 Strengths and limitations

The strengths of our study lie in the large sample size of our patient cohort and the robust method of blinded data collection, data entry and analysis. The main limitation relates to the retrospective review of medical records. The treating physicians had based their original pain classification on their clinical assessment findings, results of diagnostic tests and their own clinical judgment. Some records contained more comprehensive information than others. While the clinical diagnosis of NeP by one clinician has been used in several studies as reference standard [13], [14], [41], [54], [55], [56], we are aware that using a standardized classification system such as the NeP grading system [11] by all physicians would have strengthened the design of the study. However, these are the limitations of retrospective data analysis. A further limitation in our study relates to the fact that we encountered a large proportion (40%) of missing data in any of the questionnaires. This may impact on the generalisability of our study results and demonstrates well the difficulties of using self-reported patient data. As for the PD-Q, we encountered missing data in 22% of our patient cohort, which is comparable to findings between 13% and 35% of missing data in other studies [41], [44], [57], [58]. While our study design may have lead us to also explore the number of false negative results on PD-Q, this was not an aim of our study. We assumed that psychological trends will drive a higher score on the PD-Q, but would not drive a lower score (false negative). Furthermore, we encountered resource limitations required for such an analysis.

6 Conclusion

Our study showed that psychological factors including catastrophizing, depression, anxiety, and stress were all influential in producing a higher score on the PD-Q. For patients classified by PD-Q as having NeP, only 50% of these patients were true positives, i.e. they had a clinical diagnosis of NeP. Anxiety was significantly associated with a false positive classification of NeP. Clinicians and researchers should consider the psychological state of the patient before interpreting the results of the PD-Q and/or undertaking sensory profiling based on PD-Q results. The suitability of PD-Q as a screening tool for NeP may be in question.