Survey of Opioid-Induced Respiratory Depression in Critically Ill Patients Without Mechanical Ventilation and Construction of a Risk Nomogram

To investigate the current status of opioid-induced respiratory depression (OIRD) and potential risk factors in critically ill patients without mechanical ventilation in the intensive care unit (ICU) and to construct a risk nomogram to predict OIRD. A total of 103 patients without (or who were weaned from) mechanical ventilation who had stayed for more than 24 h in the ICU between June 1, 2021 and September 31, 2021, were included. Patient data, including respiratory depression events, were recorded. The least absolute shrinkage and selection operator regression model were used to select features that were then used to construct a prediction model by multivariate logistic regression analysis. A nomogram was established for the risk of respiratory depression events in patients without mechanical ventilation. The discriminatory performance and calibration of the nomogram were assessed with Harrell’s concordance index and a calibration plot, respectively, and a bootstrap procedure was used for internal validation.


Introduction
Moderate-to-severe acute or chronic pain is experienced by 50-77% of patients in the intensive care unit (ICU) [1][2][3]. Severe pain is associated with changes in heart rate and blood pressure, dyspnea, respiratory distress, enhanced catabolism, and immune dysfunction and has adverse neuropsychologic effects such as anxiety, depression, and impaired sleep that result in poor outcomes for patients [4,5]. Intravenous opioids are used as rst-line therapy for non-neuropathic pain in critically ill patients, including those without mechanical ventilation; these patients may have multiple-organ dysfunction and abnormal drug metabolism, placing them at higher risk of respiratory depression and death from opioids [6]. Opioidrelated adverse events can lead to longer hospital stays, increased hospitalization costs, readmission to hospital, and increased mortality [7]. However, most research on opioid-induced respiratory depression (OIRD) has focused on postoperative patients [8] and little is known about the incidence in critically ill patients without mechanical ventilation.
In this study, we try to investigate the possible risk factors for respiratory depression in patients in the ICU without mechanical ventilation (include discontinuing MV) receiving opioid analgesia. Moreover, we constructed a risk nomogram to identify patients at risk for OIRD so that timely and appropriate interventions can be implemented.

Study design and patient selection
Postoperative patients without mechanical ventilation (include discontinuing MV) who were admitted to the ICU of Nanfang Hospital of Southern Medical University, Guangzhou, China., between June 1, 2021, and September 31, 2021, were enrolled. Inclusion criteria were age >18 years and patients while received opioids for analgesia treatment for >24 h. The following opioids were used in the ICU: hydrochloric acid, remifentanil, fentanyl citrate, sufentanil citrate, morphine sulfate, hydromorphone hydrochloride, and butorphanol tartrate. Morphine was permitted for explosive pain relief and combination with nonopioid analgesics was allowed. Exclusion criteria were patients with severe intracranial or spinal neurologic diseases or coma caused by vascularity, infection, intracranial expansion, or injury; history of opioid addiction; patients with missing data; and patients who did not provide informed consent. The study design was approved by the appropriate ethics committee of Nanfang Hospital, Southern Medical University. We interviewed the patient or the patient's family member via phone to obtain consent, and all patients informed consent for study participation. This study was performed in accordance with the ethical standards of the Helsinki Declaration.

Data collection and de nitions
Data were collected from the electronic medical record database of Nanfang Hospital, Southern Medical University, Guangzhou, China. The data included patient identi cation number, name, age, height, weight, body mass index (BMI), chronic underlying diseases, types of surgery, kidney function, liver function, blood gas analysis, comorbidities, Acute Physiologic Assessment and Chronic Health Evaluation II score, heart rate, respiratory rate, duration of mechanical ventilation (the duration of MV for patients who have not received MV is 0), the dosage of sedatives and analgesics, and the occurrence of respiratory depression.
Management of analgesia and sedation was based on the 2018 Pain, Agitation/sedation, Delirium, Immobility (rehabilitation/mobilization), and Sleep (disruption) guidelines [9] and Guidelines of Sedation and Analgesia in ICU of Chinese Adults (2018) [10]. Analgesia was determined based on a Numerical Rating Scale score of <4 points for patients who were conscious and Critical-Care Pain Observation Tool score of <3 points for those who were unable to express themselves.
With adequate analgesia, the depth of sedation was determined based on physiologic parameters. A target-guided sedation strategy was implemented using the Sedation-Agitation Scale (SAS). Light sedation was de ned as a SAS score of 3-4 points. Propofol, dexmedetomidine, and midazolam were used for sedation. The depth of sedation was assessed every 2 h and titrated in order to achieve a satisfactory effect. If the SAS score was <2 points, the drug dose was gradually reduced or sedation was stopped.
Patients without mechanical ventilation were de ned as those who were admitted to the ICU but were not given or were weaned from mechanical ventilation (ie, successfully extubated). These patients received no oxygen therapy (no use of nasal catheters, masks, or non-invasive ventilators and no high-ow nasal cannula oxygen therapy). Correspondingly, if a patient is weaned after 28 hours of mechanical ventilation, it is recorded as 28h, and if another patient has never been mechanically ventilated, it is recorded as 0h.
Respiratory depressionA respiratory depression event was de ned as the occurrence of 1 or more of the following changes (See Supplement Material 1) [15]: decreased oxygen saturation and/or hypercapnia caused by hypoventilation; decreased respiratory rate; and apnea and/or asphyxia.

Endpoints
Occurrence of respiratory depression was the primary endpoint. Risk factors causing respiratory depression were the secondary outcomes.

Statistical analyses
Qualitative variables are expressed as frequencies and percentages. For quantitative variables, the Kolmogorov-Smirnov test was used to distinguish whether the data conformed to a normal distribution. Normally distributed continuous variables are expressed as mean ± standard deviation, while nonnormally distributed continuous variables are expressed as median (interquartile range). Categorical variables are expressed as a number and percentage.
Sample size evaluation was performed based on events per variable. Potential prognostic variables were identi ed by least absolute shrinkage and selection operator (LASSO) regression, and selected variables were further analyzed in a multivariate logistic regression model. Variables with prognostic signi cance were used to construct a respiratory depression prediction model and incorporated into a nomogram. Harrell's concordance index (C-index) and calibration curve were used to assess the discriminatory performance and calibration of the model, respectively.

Patient characteristics
All patients admitted to the ICU of our hospital between June 1, 2021 and September 31, 2021 were screened and 103 patients were ultimately included in the study. A total of 49 patients (47.6%) were considered to have respiratory depression events (Fig. 1). The demographic and clinical characteristics of the patients are shown in Table 1.

Risk factors for respiratory depression
A total of 36 variables that were considered as potential predictors of respiratory depression were identi ed with the LASSO regression model. Disease type, sepsis, duration of mechanical ventilation, and PaO 2 /FiO 2 were associated with respiratory depression when the optimal λ value was 0.0817 (Fig. 2).
According to the results of the logistic regression analysis, we selected 4 signi cant factors (Table 2) as input variables of a nomogram for respiratory depression events in ICU patients without mechanical ventilation while receiving opioids for analgesia (Fig. 3). The output variables were the corresponding scale score of the input variables, total points, and predicted probability of respiratory depression in the patients.

Discriminatory performance and calibration of the nomogram
The C-index of the risk factor model was 0.902 (95% con dence interval [CI]=0.842-0.961) (Fig. 4A); the corrected C-index using 1000 bootstrap samples was 0.873, which demonstrated that the model had a good discriminatory performance. The calibration curve showed consistency between predicted and actual results (Fig. 4B).

Discussion
In the present study, 47.6% of critically ill patients without mechanical ventilation while receiving opioid analgesics developed respiratory depression. The logistic regression analysis identi ed 4 factors associated with respiratory depression. One of the risk factors was disease type; speci cally, patients with cardiopulmonary and respiratory disease who underwent surgery had a longer stay in the ICU because of delayed tracheal extubation and a higher rate of respiratory depression. Therefore, more attention should be paid to airway management in these patients. Surgery for otorhinolaryngologic or oral and maxillofacial diseases are not only more likely to cause airway blockage by secretions or aspiration, but can also cause damage to the airway and result in its collapse. Respiratory depression can also occur as a result of compression and bandaging of the head and neck or glossoptosis after surgery [16]. In general, the risk of respiratory depression is lower after abdominal surgery than after operations that are directly related to the airway. However, patients with abdominal diseases are often transferred to the ICU due to intraoperative bleeding, prolonged duration of operation, and unstable vital signs; thus, the occurrence of respiratory depression may be indirectly associated with circulation status and changes in intra-abdominal pressure. Finally, as orthopedic or trauma surgery is usually at a site that is far away from the airway or respiratory tract and is performed under local anesthesia, patients have stable vital signs and have a lower risk of respiratory depression. Notably, we did not examine whether there are differences in respiratory depression between local and general anesthesia, as demonstrated by a metaanalysis of 12 retrospective studies on postoperative opioid-related respiratory depression [8]. In our study, the multivariate logistic regression model revealed that patients undergoing surgery for cardiopulmonary and respiratory diseases (odds ratio [OR]=32.833, 95% CI=4.189-725.164) were at higher risk for respiratory depression when receiving opioids after surgery compared to orthopedic and trauma surgery patients.
Sepsis is a systemic in ammatory response triggered by infection that can lead to multiple organ dysfunction syndromes and even death [17]; mortality from sepsis ranges from 10-52% [18][19][20].
Pulmonary vascular endothelial damage during sepsis interferes with capillary blood ow and increases vascular permeability, which can cause interstitial or alveolar pulmonary edema, ventilation-perfusion mismatch, and hypoxemia [21]. Additionally, sepsis is usually accompanied by increased energy expenditure from respiration and is complicated by encephalopathy or a decreased level of consciousness [21,22]. Therefore, it is important to stabilize the respiratory system in cases of sepsis, which may require tracheal intubation and mechanical ventilation. We found that sepsis (OR=6.898, 95% CI=1.756-33.000) was an independent risk factor for respiratory depression in non-ventilated critically ill patients treated with opioid analgesics in the multivariate logistic analysis.
Patients admitted to the ICU without mechanical ventilation are at high risk of OIRD. Such patients are usually admitted for advanced circulation and respiration monitoring. In our analysis, patients admitted to the ICU who did not receive mechanical ventilation (OR=20.757, 95% CI=2.409-502.222) had a greater risk of respiratory depression than those while received mechanical ventilation for ≤1 day. Baseline PaO 2 /FiO 2 (OR=7.350, 95% CI=2.483-24.286) was also an independent risk factor for respiratory depression events in the former group. Thus, clinicians should prioritize the management of these patients, for example by supplementing oxygen therapy with assisted ventilation or delaying tracheal extubation in order to accelerate the recovery of respiratory function.
Most studies on OIRD have focused on patients undergoing routine surgery and only a few have examined critically ill patients without mechanical ventilation. A recent meta-analysis of 12 retrospective studies including 841,424 surgical patients showed that age, sex, BMI, diabetes, and kidney disease were not associated with postoperative OIRD [8], which is consistent with the ndings of our study; it also showed that COPD and obstructive sleep apnea syndrome are risk factors for OIRD. However, the analysis examined chronic lung diseases, which have a broad de nition and include patients with chronic bronchitis, who have acceptable respiratory compensation; therefore, the authors were unable to determine whether underlying lung diseases are a risk factor for OIRD. It is also worth mentioning that our study and the meta-analysis differed in terms of the de nition and observation time of OIRD, analgesic program, and patient distribution. The meta-analysis adopted a stricter de nition of OIRD with a different rescue standard (ie, using naloxone to reverse OIRD). In contrast, as we aimed to establish a means of identifying early symptoms of respiratory depression to enable timely intervention, we used a standard de nition of OIRD (ie, decreased pulse oxygen saturation, increased arterial carbon dioxide, and changes in the mode of assisted ventilation). In terms of analgesic drugs, the meta-analysis mostly included postoperative patients whose morphine equivalent daily dose (MEDD) (either self-controlled or administered via the intravenous, oral, epidural, etc routes) was 24.7±14 and 18.9±13 mg in OIRD and non-OIRD patients, respectively. In our study, all patients were continuously administered the analgesic intravenously with an electronic micropump. The median hourly MED10 was 0.6 mg, equivalent to 144 mg MEDD. Thus, MED10 was much higher in critically ill patients without mechanical ventilation than in patients undergoing routine surgery. In addition, the highest risk period for OIRD in the meta-analysis was the rst 4 h after surgery, during which 34% of all critical respiratory events occurred; OIRD typically occurred in the rst 12-24 h (85%), but most were not extubated within the rst 4 h after surgery and 48% (50/103) of patients were mechanically ventilated for >1 day, with 62% (31/50) reporting respiratory depression events.
Higher rates of opioid accumulation and infusion and combined use of sedative drugs can increase the risk of respiratory depression. Different types of opioids have a characteristic rate of metabolism, which can alter the risk of respiratory depression [8]. In this study, although the total amount of analgesic and sedative drugs and treatment duration were similar between patients with and those without respiratory depression, the rate of respiratory depression differed between the two groups (P=0.024) when the cumulative dose of analgesic drugs was classi ed as >50 (MED10) and ≤50 (MED10); that is, in the respiratory depression group, the >50 (MED10) subgroup had a higher rate of respiratory depression than the ≤50 (MED10) subgroup (61% vs 39%). However, as the multivariate logistic regression analysis showed that >50 (MED10) was not an independent risk factor for respiratory depression, it was not included in the nal regression model. Various analgesic drugs are used in the ICU. For patients with a higher risk of respiratory depression, doctors usually administer butorphanol (which has a lower risk of respiratory adverse effects), reduce the rate or dosage of opioid infusion, or combine opioids with other medications such as nonsteroidal anti-in ammatory drugs. Thus, selection bias by clinicians is a potential reason for our observations. Finally, in the subgroup analysis based on fentanyl (See Supplement Material 2), we found that patients treated with the combination of fentanyl and butorphanol had fewer respiratory depression events and used a lower dosage of opioids than those treated with fentanyl alone (P<0.05), which is in accordance with previous studies. However, we did not adjust for confounding factors between groups and the results should therefore be interpreted with caution. Randomized controlled trials are needed in order to clarify whether the fentanyl-butorphanol combination bene ts critically ill patients without mechanical ventilation.
Respiratory depression induced by opioids is mainly due to the activation of µ receptors, which can be modulated by δ receptors [23]. Unlike µ receptors, activation of δ and κ receptors can relieve chest wall muscle stiffness. The κ receptor agonist butorphanol was shown to have fewer respiratory adverse effects than other opioids [24][25][26]. Tapentadol, a new µ receptor agonist, has a strong analgesic effect and low risk of respiratory depression [27]. Further studies on the clinical manifestations and causes of OIRD will help to determine the optimal pharmacologic intervention for pain management based on the characteristics of opioid receptors.
Our study had some limitations. Firstly, the sample size was relatively small, which undermined the reliability of the multivariate logistic regression analysis. Moreover, because of the small sample size, we used only internal veri cation instead of random split veri cation of the bootstrap analysis to validate our model. A multicenter study with a larger cohort is needed for external validation and optimization of our model. Finally, the degree of respiratory depression was not classi ed in the nal multivariate regression model. Studies based on a larger population can provide more data for predicting the occurrence of respiratory depression.

Conclusion
Sepsis, duration of mechanical ventilation, and oxygenation index were independent risk factors for respiratory depression events. The predictive model showed good discriminatory performance and calibration and can be used by clinicians to identify at-risk patients so that timely and appropriate interventions can be implemented.

Declarations
Ethics approval and consent to participate: The study design was approved by the appropriate ethics committee of Nanfang Hospital, Southern Medical University. We interviewed the patient or the patient's family member via phone to obtain consent. This study was performed in accordance with the ethical standards of the Helsinki Declaration.
Consent for publication: Not applicable.
Availability of data and material: All data generated or analysed during this study are included in this published article.
Competing interests: The authors declare that they have no competing interests.  Nomogram of respiratory depression events in critically ill patients receiving opioid analgesic treatment. MV mechanical ventilation. Abdominal surgery includes gastrointestinal, urologic, hepatobiliary, and obstetric/gynecologic surgery.