Influence of Response Criterion on Nociceptive Detection Thresholds and Evoked Potentials

Pain scientists and clinicians search for objective measures of altered nociceptive processing to study and stratify chronic pain patients. Nociceptive processing can be studied by observing a combination of nociceptive detection thresholds and evoked potentials. However, it is unknown whether the nociceptive detection threshold measured using a Go-/No-Go (GN) procedure can be biased by a response criterion. In this study, we compared nociceptive detection thresholds, psychometric slopes and central evoked potentials obtained during a GN procedure with those obtained during a 2-interval forced choice (2IFC) procedure to determine 1) if the nociceptive detection threshold during a GN procedure is biased by a criterion and 2) to determine if nociceptive evoked potentials observed in response to stimuli around the detection threshold are biased by a criterion. We found that the detection threshold can be higher when assessed using a GN procedure in comparison with the 2IFC procedure. The average P2 component in the central evoked potential showed on-off behavior with respect to stimulus detection and increased proportionally with the detection probability during a GN procedure. These data suggest that nociceptive detection thresholds estimated using a GN procedure are subject to a response criterion.


Introduction 44
Pain scientists and clinicians search for objective criteria to identify impaired nociceptive 45 processing for the purpose of stratification and treatment of chronic pain patients (Mouraux & 46 Iannetti, 2018). With this aim, nociceptive processing of patients is usually evaluated using a 47 combination of neurophysiological and psychophysical testing. In this field, there is a recent 48 renewed interest in the assessment of mechanical, thermal and electric detection thresholds. 49 However, the interpretation of these thresholds could alter depending on the procedure through 50 which these thresholds are measured. 51 Recently, we developed a method to assess nociceptive processing by quantifying the effect 52 nociceptive stimulus properties on detection probability and cortical evoked potentials (EPs). In 53 this method, we stimulate nociceptive afferents in the skin by intra-epidermal electric stimulation 54 with a specialized electrode (Steenbergen et al., 2012). This method selectively activates 55 nociceptive afferents in the skin provided that low stimulation currents are used, for which a limit 56 of twice the detection threshold was proposed as a rule of thumb . Stimulus 57 amplitudes are centered around the detection threshold by an adaptive psychophysical method of 58 limits (Doll, Veltink, & Buitenweg, 2015) and the electroencephalogram (EEG) is recorded in 59 response to each stimulus. This allows us to record the combination of nociceptive detection 60 thresholds and evoked potentials in response to nociceptive stimulation. We recently showed that 61 nociceptive detection thresholds of single-pulse and double-pulse intra-epidermal electric stimuli 62 can be used to observe peripheral and central changes of nociception following deafferentation by 63 capsaicin (Doll et al., 2016 (Steenbergen, 2012). Two stimulus types were 122 used during the experiment: 123 • One square pulse with a pulsewidth of 210 μs. 124 • Two square pulses with a pulsewidth of 210 μs and an inter-pulse interval of 10 ms. 125 126

Familiarization 127
Participants were instructed to press and hold a button. For familiarization with the sensation of 128 intra-epidermal stimuli, participants were stimulated with a series of pulses with a stepwise (0.025 129 mA) increasing amplitude and instructed to release the button when a stimulus was clearly 130 perceived for at least two times. For an initial estimate of the detection threshold for each stimulus 131 type, participants were stimulated with a series of pulses with a stepwise (0.025 mA) increasing 132 amplitude and instructed to release the button when any sensation was perceived that they ascribed 133 to stimulation. 134 135

Go/No-Go Procedure 136
Participants were seated upright in a chair and asked to focus on the site of stimulation. Detection 137 thresholds were estimated and tracked using an adaptive procedure (Doll et al., 2015). Participants 138 were instructed to press and hold a button, and to briefly release the button when any sensation 139 was perceived that they ascribed to stimulation (Fig. 1). For the adaptive procedure, the stimulus 140 amplitude was randomly picked from a vector of 5 stimulus amplitudes with a stepsize of 0.025 141 mA initialized around the initial estimate of the detection threshold. The vector of amplitudes was 142 decreased by 0.025 mA when a stimulus was reported as detected and increased by 0.025 when the participant did not release the response button. This process was repeated independently for 144 every stimulus type, with the order of stimulus type randomized, for a total of 130 stimuli per type. The scalp EEG was recorded at 32 channels (international 10/20 system) using a REFA amplifier 171 (TMSi B.V., Oldenzaal, the Netherlands) with a sampling rate of 1024 Hz. Participants were asked 172 to fix their gaze at a spot on the wall. Electrode impedance was kept below 20 kΩ. 173 174

Nociceptive Detection Threshold 175
The nociceptive detection probability was estimated by global optimization of the negative log-176 likelihood using an implementation of the GlobalSearch algorithm (Ugray et al., 2007) in 177 combination with an interior-point algorithm to find local minima (Coleman & Li, 1996) in Matlab. 178 In the case of the Go-/No-Go procedure (Equation (1) In the case of a 2IFC procedure (Equation (2)), this function was adapted to account for a 50% 182 guessing rate at low stimulus amplitudes. 183 Detection probability for a go/no-go procedure: Detection probability for a 2-interval forced choice procedure:

Nociceptive Detection Threshold 203
A typical example of an experiment with the GN and the 2-IFC procedure is displayed in Fig. 2. 204 During the GN procedure, the detection threshold for single-pulse stimuli was larger than the 205 detection threshold for double-pulse stimuli. Both thresholds showed a small increasing drift over 206 time. During the 2-IFC procedure, the thresholds were equal for single-pulse and double-pulse 207 stimuli. Drift over time was small or not present. The detection thresholds and slopes for all 25 208 participants are displayed in Fig. 3. The detection threshold for single-pulse stimuli during a 2-IFC 209 procedure was significantly lower than the detection threshold for single-pulse stimuli during a 210 GN procedure. The psychometric slope for single-pulse stimuli during a 2-IFC procedure was 211 significantly larger than the psychometric slope for single-pulse stimuli during a GN procedure.

Evoked Brain Activity 227
Grand average evoked potentials at Cz-M1M2 acquired during both procedures are displayed in 228 Fig. 4. There was a significant contrast between evoked potentials in response to detected and non-229 detected stimuli in the GN procedure, and correct and incorrect trials in the 2IFC procedure. For 230 the GN procedure, the evoked potential was significantly larger than baseline for detected as well 231 as non-detected stimuli. For the 2IFC procedure, the evoked potential was only significantly larger 232 than baseline for correct trials. Note that the average evoked potential for correct trials (2IFC) was 233 lower than the average evoked potential for detected stimuli (GN), but might be confounded by 234 inclusion of trials that were not consciously perceived but simply guessed correctly. 235 Grand average evoked potentials at Cz-M1M2 for several levels of detection probability 236 are displayed in Fig. 5. There was a significant effect of detection probability on the evoked 237 potential during both procedures and for both stimulus types. While the average evoked potential 238 during a GN procedure appears graded with stimulus intensity, the average evoked potential during 239 a 2IFC procedure remains low until high levels of detection probability are reached, i.e. a detection 240 probability larger than 0.875. Both phenomena are more clearly visible in Fig. 6, where the average amplitude of the major positive peak between 380 and 420 ms, the P2, is displayed. Here, the 242 average P2 appears to increase almost proportional with respect to detection probability during the 243 GN procedure. Note that this proportional increase with detection probability can be attributed to 244 two phenomena: 1) The average P2 for detected stimuli is at almost every point significantly larger 245 than the average P2 amplitude for non-detected stimuli, leading to an increased average P2 over 246 all stimuli when more stimuli are detected. 2) There is an increasing trend in the average P2 for 247 both detected and non-detected stimuli, leading to a further increase in the average P2 over all 248 stimuli with respect to detection probability. Similar to previous figure, the average P2 during the 249 2IFC procedure remains low until a probability larger than 0.875 is reached.

Discussion 270
In this study, we observed nociceptive detection thresholds, psychometric slopes and central 271 evoked potentials obtained during a GN and a 2IFC detection procedure. The differences observed 272 between both procedures in nociceptive detection threshold and in evoked responses include 273 important clues about how nociceptive detection might work, and how the threshold obtained 274 during these procedures can be interpreted. 275 The first objective of this study was to determine if the nociceptive detection threshold 276 during a GN procedure is biased by a response criterion. We found that the detection threshold for 277 single-pulse intra-epidermal electric stimuli is significantly higher, and the psychometric slope 278 significantly lower, during a GN procedure in comparison with a 2IFC procedure. In contrast, we 279 found that the threshold for double-pulse stimuli does not differ significantly between procedures. 280 This result implies that for some types of stimuli the nociceptive detection threshold measured 281 during a GN procedure reflects evoked neural activity exceeding a response criterion, rather than 282 the presence of sensory evidence itself. Equal detection thresholds for double pulse stimuli 283 between the GN and the 2IFC procedure indicate that the extend to which the observed detection 284 threshold is influenced by the response criterion also depends on stimulus properties, and that the 285 bias of the detection threshold introduced by a criterion might be lower for high signal-to-noise 286 ratio stimuli such as the double-pulse stimulus in this experiment. In addition, a significant 287 difference was observed between single-and double-pulse stimuli during a GN procedure, while 288 no significant difference was observed between detection thresholds for single-and double-pulse 289 stimuli during a 2IFC procedure. Although a small difference between the single-and double-pulse 290 threshold might go unnoticed due to estimation errors, it is clear that the large difference between 291 both stimulus types in a GN procedure almost completely disappears during 2IFC. The reason for 292 this discrepancy between both tasks remains unclear without more sophisticated psychophysical 293 modeling, which is out of the scope of this study. However, these results warrant the development 294 of novel psychophysical models that are tailored to the process of nociception in future studies. 295 One of the potential factors that might help explaining such a difference would be the presence of 296 spontaneous neural activity influencing both the response criterion and psychometric slope of the 297 participant. More importantly, formulation of psychophysical models that are connected to 298 neurophysiological mechanisms can lead to more insight in the interpretation of the detection 299 thresholds measured in a clinical or research setting. 300 The second objective of this study was to determine if the presence of a response criterion is 301 reflected in the nociceptive evoked potentials observed in response to stimuli around the detection 302 threshold. We measured a significant central evoked response at Cz-M1M2 during both procedures 303 for detected stimuli (GN) and correctly reported trials (2IFC). We also measured a significant 304 evoked response to non-detected stimuli (GN), which was absent for incorrectly reported trials 305 (2IFC). We found that the evoked P2 response is proportionally graded with detection probability 306 during a GN procedure. At the same time, we observed that the P2 response during a 2IFC 307 procedure for stimuli with the same detection probability (corrected for guessing rate), remains 308 low until a large detection probability is reached. The P2 response to detected and non-detected 309 stimuli show that we might be looking at a mostly dichotomous response, where the response is 310 much larger for detected stimuli than for non-detected stimuli. The visual evoked P3 response is 311 considered a key marker of conscious access to sensory evidence (