This cross-sectional study is reported according to the Strengthening the Reporting of Observational studies in Epidemiology (STROBE) Statement.18 From November 2016 to April 2019, people with LBP, visiting Dutch primary care physical therapy practices, were recruited. Ethical approval was provided by the Ethical committee of Maasstad Hospital, Rotterdam, The Netherlands (T2016–38).

Inclusion criteria were: age between 18 and 65 years, having had LBP for at least one week with or without referred pain in one or both lower extremities and having a mean pain intensity reported on the 0–100 mm Visual Analogue Scale (VAS) of ≥ 30 mm during the last week. Exclusion criteria were: pregnancy; LBP after surgery or trauma; psychiatric diagnosis determined by a psychiatrist; people with fibromyalgia, chronic fatigue syndrome, or rheumatoid arthritis; LBP due to referred pain from internal organs; and inability to write or read Dutch.

Participants were screened for eligibility according to the Dutch LBP guideline of the Royal Dutch Society for Physiotherapy by their treating physical therapist and were informed about this study.19 Patients received an information leaflet about the objectives and content of the study. The first author (HdB), an experienced physical therapist trained in performing QST-measurements, executed all measurements. HdB was blinded to the medical record and all other study data of the participants. Prior to enrollment, all participants provided written informed consent.

First, participants provided demographic information about sex, age, most painful site of their LBP, and the present type of pain (by self-completing the questionnaire “painDETECT”). The other questionnaires: SBT, CSI, Tampa scale for kinesiophobia, Pain Catastrophizing Scale, Roland Morris Disability Questionnaire, and VAS for pain severity were also self-administered. To avoid test order effect, questionnaires were administered in an alternately structured order: participant A started with questionnaire 1 (and ended with questionnaire 7), participant B started with questionnaire 2 (and ended with questionnaire 1), participant C started with questionnaire 3 (and ended with questionnaire 2) and so on. After completing the questionnaires, the study was continued with the QST-measurements.

The Start Back screening Tool (SBT) identifies to what extent patients with LBP seen in primary care are at risk for a poor prognosis. It classifies patients in one of three risk levels (low-, medium-, high-risk).14 The SBT consists of nine questions of which eight questions need to be answered with “true/false” and one on a 5-point Likert-scale.20 The Dutch version generates sufficient valid and reliable data and shows substantial reproducibility.20

The Central Sensitization Inventory (CSI) measures CS related symptoms and consists of 25 CS related questions, to be answered on a 5-point Likert-scale (0 = never, 4 = always). The Central Sensitization Inventory Symptom Severity Calculator determines patient related severity level (low, medium, high) of the CSI. The test-retest reliability of the Dutch version is excellent.21 The intraclass correlation coefficient (ICC) varies from 0.88 to 0.91.21 The sensitivity and specificity of the questionnaire is 81 % and 75 %, respectively.22

The Tampa scale for kinesiophobia (TSK) assesses kinesiophobia. It consists of 17 statements that need to be scored on a 4-point Likert scale (1 = highly disagree, 4 = highly agree).23 A cut-off score of ≥37/68 indicates kinesiophobia.23 The Dutch version of the TSK (TSK-DV) shows high internal consistency (Cronbach's alpha range 0.68, 0.80), good construct and criterion validity, and an ICC of >0.7.24,25

The Pain Catastrophizing Scale (PCS) measures the degree of pain catastrophizing by assessing three related subscales: magnification, rumination, and helplessness.26 It consists of 13 statements that need to be scored on a 5-point Likert scale (0 = not at all, 4 = all the time).26 A cut-off score of ≥30/52 is used to establish “pain catastrophizing”.26 The PCS shows good test-retest reliability and internal consistency.27,28

The Roland Morris Disability Questionnaire (RDQ) assesses the functional status of people with LBP.29 This questionnaire contains 24-items in which the answer options are “yes/no”.30 Total score is calculated by adding up the number of “yes” answers ranging from 0 (=no disability) to 24 (= maximal disability).31 Higher scores indicating a higher level of disability.31 The Dutch version of the RDQ is a reliable questionnaire (ICC= 0.91).29

The Visual Analogue Scale (VAS) measures pain intensity of people with LBP. This self-reported scale has a horizontal line of 100 mm. The score varies from 0 mm (no pain) to 100 mm (unbearable pain).32 Participants are requested to mark the line that corresponds most closely with the pain intensity they currently experience.32 Despite its low content validity, the VAS is recommended to measure and report pain in clinical trials in people with LBP.32,33 The ICC varies from 0.73 to 0.85.34

The painDETECT discriminates between nociceptive and neuropathic pain mechanisms.35 It consists of seven questions about the quality of neuropathic pain symptoms with a 5-point Likert-scale (0=never, 5=very strongly), one question about radiating pain answered by “yes/no”, and four pictures that describe the pain course pattern. The participant has to mark the picture that describes the pain course best. A total score ≤ 12 is predominantly nociceptive pain and a total score of ≥ 19 is predominantly neuropathic pain.35 It is a valid, reliable screening tool and it has been adequately translated into Dutch.35,36 The ICC is 0.91.37

To determine the testing site for the QST-measurements, participants indicated the most painful side of their LBP. If the pain was the same between sides, the right side was chosen. QST-measurements were performed at the following locations: thumb, lower back (2 cm lateral to the spinous process of L4), and gastrocnemius muscle-tendon junction. During the measurements the participants lied prone and the locations were marked. The measurements started with heat pain threshold (HPT). After a five-minute pause, pressure pain threshold (PPT) was measured. After another five minutes, temporal summation (TS) was assessed. After another five minutes pause, conditioned pain modulation (CPM) was assessed. The whole procedure took one hour.

Heat Pain Threshold (HPT) measurements were conducted with an increasing stimulus (1 °C /sec, 32 °C baseline and 50 °C cut-off, 8 cm2 thermode), using the TSA 2001-II (MEDOC, Israel). Participants were instructed to say ‘stop’ when the sensation first became uncomfortable.38 The measurements were performed twice on each location (i.e., thumb, lower back (2 cm lateral to the L4 spinous process), and the gastrocnemius muscle-tendon junction) with an interval of 30 s. For analysis, the mean scores of each location were calculated and used. Indication for enhanced sensitivity in the somatosensory system is lower HPTs at non-segmental level.39 HPT measurements have proved to be of acceptable reliability.40

Pressure Pain Threshold (PPT) measurements were performed with a handheld pressure algometer (Wagner Instruments, FDX 50 Algometer, Greenwich, USA,). The circular probe has a 1 cm diameter and the measurement was performed by applying an increasing stimulus (1 kg/s). The instruction was to say ‘stop’ when the sensation first became uncomfortable.38 PPTs were also assessed twice at the same locations as the HPT with an interval of 30 s and the mean of each location was calculated and used for analysis. Lower remote PPTs at non-segmental levels are also indicative of CS.39 PPT measurements show acceptable test-retest reliability.40

Temporal summation (TS) was measured by applying 10 consecutive identical nociceptive stimuli at the same test locations described above.41 Pain sensation will increase if the neuronal output amplifies, which mediates TS.41 The stimuli were applied at the previously determined mean PPT intensity. This pressure was maintained for one second before starting the measurement. At a rate of approximately 2 kg/s for each stimulus, the pressure was increased. Stimuli were presented with an interstimulus interval of one second. By means of a verbal numeric rating scale (0=no pain, 10 =unbearable pain) for the pain intensity, the participants had to rate the first, fifth, and tenth stimulus. The result for TS is calculated by the difference between the tenth and the first verbal numeric rating scale score. The reliability is acceptable.40

Endogenous pain inhibition is measured by means of the Conditioned Pain Modulation (CPM) paradigm42 using the Thermo Scientific™ VersaCool™ refrigerated circulating bath (ThermoFisher Scientific, Newington, U.S.A.) and the pressure algometer. The painful conditioning stimulus was done by immersing the participant's hand, contralateral to the most painful side, in a cold water bath of 12 °C. Participants were instructed to keep their hand in the water as long as bearable for a maximum of two minutes. On each test location, PPTs were taken as the test stimulus, once their hand was removed out of the water. In order: 1) thumb, 2) lower back, 3) gastrocnemius muscle-tendon junction. Mean score of each location was calculated and used for analysis. For drawing a conclusion about endogenous pain inhibitory capacity, post-conditioning scores were subtracted from pre-conditioning scores. The reliability is acceptable.40

This study was performed as a secondary analysis of a case-control study which investigated whether differences in QST-measurements exist between people with acute and chronic LBP and healthy controls.43 For data analysis IBM SPSS Statistics for Windows Version 25.0 (Armonk, NY: IBM Corp.) was used. Demographic data were presented as mean (standard deviation). The participants were divided into acute (0–12 weeks) and chronic LBP (≥ 12 weeks).44 The entire dataset was checked for completeness. Incomplete questionnaires were removed for analysis. Outliers were identified by checking boxplots. Only the outliers who exceeded at least 1.5 times the inter quartile range,45 were removed from the data analysis. Cohen's kappa was calculated estimating the agreement in identifying subgroups identified by the SBT and CSI (categorical variables): whether participants belonging to the SBT high risk level, were in the CSI high severity level. The interpretation of Cohen's kappa was according to Landis and Koch46 (<0.00 poor, 0.00–0.20 slight, 0.21–0.40 fair, 0.41–0.60 moderate, 0.61–0.80 substantial, 0.81–1.00 almost perfect). Spearman's correlation calculation was done between the CSI and painDETECT to determine if participants with positive painDETECT were a possible confounder. The interpretation was according to Schober et al.47 (0.00–0.10 negligible, 0.10–0.39 weak, 0.40–0.69 moderate, 0.70–0.89 strong and 0.90–1.00 very strong). A linear contrast analysis was performed. Conditionally, group mean of the three risk/severity levels of the SBT and CSI need to differ significantly. Therefore, one-way ANOVA was applied, based on the significance of the Levene's test (p < 0.01). If it met this requirement linear contrast analysis was performed. Dependent variables were: QST measurements and questionnaires (except PainDETECT). Independent variables were: risk/severity levels of SBT or CSI and participants divided into the acute and chronic LBP group. Adjustments were made for sex and age.

One hundred participants with LBP with a mean age of 42.36 (SD 10.84) years participated. Forty-four participants were male and 56 were female. Forty-seven participants had acute and 53 had chronic LBP (Table 1). On the painDETECT, 57.1 % of the participants scored negative (nociceptive pain), 23.5 % scored ambiguous (unclear pain mechanism), and 19.4 % of participants scored positive (neuropathic pain). The CSI scores correlated moderately with the painDETECT scores (r = 0.47; p < 0.001).47 During recruitment two participants reported that this study was too time consuming and therefore discontinued study participation. One participant appeared to be insufficiently proficient in the Dutch language and was excluded. Two more participants were excluded: one suffering from fibromyalgia and one having had low back surgery. The recruitment ended on reaching the required 100 participants. Four SBT, one CSI, and one painDETECT questionnaires were excluded from further analysis due to incompleteness. Due to technical problems, the HPT measurements consisted of 91 data points instead of 100.

The level of agreement (95 % CI) between the SBT and CSI was 0.10 (−0.03, 0.23), which indicates no agreement.46

All questionnaires showed positive significant linear trends across the risk levels (TSK, PCS, VASmean, RDQ, all p < 0.001) with adjustments for sex and age. Both HPTthumb and HPTgastroc showed negative significant linear trends across the risk levels (HPTthumb, p = 0.03; HPTgastroc, p = 0.005). Both PPTL4 and PPTgastroc showed negative significant linear trends across the risk levels (PPTL4, p = 0.02, PPTgastroc, p = 0.04). None of the TS measurements showed significant linear trends across the risk levels. Only the δCPMthumb showed a positive significant linear trend across the risk levels (δCPMthumb, p = 0.04) (Table 2).

All questionnaires showed positive significant linear trends across the severity levels (TSK and PCS, p < 0.001; VASmean,p = 0.01; RDQ, p = 0.001) with adjustments for sex and age. All PPT measurements showed negative significant linear trends across the severity levels (PPTthumb, p = 0.009, PPTL4, p = 0.012, PPTgastroc, p = 0.001). The δCPMthumb showed a positive significant linear trend across the severity levels (p = 0.03) (Table 3).