During the observation period, 144 potentially eligible patients were admitted to the interdisciplinary surgical ICU at the University Hospital Ulm, Germany. After simple randomization and enrollment into the study, 30 patients had to be excluded for the following reasons before a final data analysis could be performed: 2 patients died before obtaining written consent, 10 patients had not yet signed post-hoc informed consent at the time of data analysis, in 6 patients the data set was incomplete, 7 patients could not be measured in the specified 12-hour interval after admission to the ICU. 5 other patients were excluded after checking the NFRT due to an incorrect measurement (Fig 1). Finally, a total of 57 patients in each group was included in the further analysis (see Table 2). In Group B, patients were older and mortality was higher compared to Group A. Furthermore, compared to group A, group B patients had statistically significantly more frequent BPS point values of 4, which indicates slight pain. Men were over-represented in both groups. In group A, patients were sedated deeper on average compared to Group B and received more remifentanil.
The following comments on the results refer exclusively to Group A, as Group B served only as a control group:
In detail, the 57 patients in group A were 61 years old. 44 were male, 13 female. The mean length of stay at the ICU was 10 days. 6 patients died during their stay in the ICU. The NFRT was measured at least twice a day in all patients of Group A, preferably with repeated measurements (interval 30-60 minutes after the first measurement). In addition, BPS and RASS were recorded at each NFRT assessment. In 51 out of 57 patients, the pain assessment predominantly resulted in a BPS score of 3. 21 of these patients had a BPS score of 4 in at least one measurement. Higher BPS scores (> 4) could only be detected in 11 patients (Table 2). In terms of a reliable data analysis, the further evaluation of the data was performed in patients with a BPS of 3 or 4.
Difference between NFRTs at BPS 3 and BPS 4
The NFRTs of all patients in group A show that with n = 57 patients and 297 measurements the median threshold is 32.00 mA (IQR = 16.85 - 52.30 mA). Patients whose NFRT was measured at a BPS point value of 3, correspondingly classified as free of pain, have a statistically significantly higher threshold (n = 51 patients, 210 measurements, median threshold 36.00 mA, IQR = 20.50 - 60.00 mA, Mann-Whitney rank-sum test - p-value: 0.048). With a BPS point value of 4, corresponding to slight pain, a median NFRT of 26.90 mA (IQR = 16.85 - 52.3 mA, measurements = 53) was determined in 28 patients. In comparison to the overall population, there is no statistically significant difference (Mann-Whitney rank- sum test - p-value: 0.079). However, a comparison of the measured stimulus thresholds between BPS 3 and BPS 4 indicates that patients with assumed absence of pain had a statistically significantly higher NFRT than patients with mild pain (Mann-Whitney rank-sum test - p-value: 0.005). The results are displayed in Figure 3.
In a second step, univariate analysis was used to determine whether the NFRT is statistically affected by demographic factors. The univariate analysis did not show a statistically significant effect on NFRT for any of the investigated parameters (Table 3). Since the results of the univariate models exhibited no significance, we did not run a more complex multivariable regression model afterwards. In addition, a mixed linear model was used to investigate the effect of the applied analgesics and sedatives on the NFRT in group A (Table 4).
Calculated Effects of the used analgesics and hypnotics on the NFRT in a mixed model
In patients with measured NFRTs at BPS 3 and 4 (Fig. 4), it could be shown in the mixed model that an increase of 1 mg/h in remifentanil would have led to a statistically significant increase in the NFRT by 56.29 mA (estimate = 56.29, SE = 21.90, p-value = 0.011). At BPS point values of 3, a 1mg/h increase in remifentanil dosage would have resulted in a statistically significant 60.03 mA increase in NFRT (estimate = 60.03, SE = 24.38, p-value = 0.015). At BPS point values of 4, a 1 mg/h increase in remifentanil would have resulted in a 75.30 mA increase in NFRT. Due to the number of available readings, the result is not statistically significant (estimate = 75.30, SE = 66.43, p-value = 0.267). If patients had a BPS score of 4, a dose increase of sufentanil by 1 µg/h would have resulted in a statistically significant decrease of the NFRT by -6.8 mA (estimate = -6.800, SE = 1.467, p-value = 0.017). At BPS point values of 3 and of 3 and 4 when considered together, there was no statistically significant effect of a dose increase of sufentanil on the NFRT (BPS 3 & 4: estimate = -0.014, SE = 0.904, p-value = 0.988, BPS 3: estimate = 0.166, SE = 0.972, p-value = 0.866). A dose increase of metamizole would not have led to statistically significant changes in the NFRT in the calculation model. A potential increase of metamizole by 1 mg/h did not lead to statistically significant changes in the NFRT in the calculation model. Likewise, no statistically significant change in the NFRT could be achieved by bolus administration of piritramide (1 mg). The decrease in the NFRT at a BPS point value of 3 can be explained as a statistical effect due to the small number of measured values.
Neither an increase in propofol nor lormetazepam dose by 1mg/h increased the NFRT in the mixed-model in a clinically relevant manner. The calculated statistical significance of an increase in propofol dose is interpreted as a purely statistical effect.
NFRT at different RASS values
Comparing the measured median NFRTs at different RASS values, it is seen that the deeper the sedation, the higher the corresponding stimulus threshold. This is graphically depicted in Figure 4, with the corresponding statistical analysis. Since several measurements were performed per patient, analogous to the BPS-NFR measurements, a simple correlation analysis is not statistically valid. A multivariate analysis of the correlation between NFRT and RASS was not performed, since it was demonstrated in the mixed model calculation that a change in the dose of i.v.-hypnotics had no significant effect on NFRT.