Prognostic factors in acute poisoning with central nervous system xenobiotics: development of a nomogram predicting risk of intensive care unit admission

Abstract

Background

Acute intoxication with central nervous system (CNS) xenobiotics is an increasing global problem. Predicting the prognosis of acute toxic exposure among patients can significantly alter the morbidity and mortality. The present study outlined the early risk predictors among patients diagnosed with acute exposure to CNS xenobiotics and endorsed bedside nomograms for identifying patients requiring intensive care unit (ICU) admission and those at risk of poor prognosis or death.

Methods

This study is a 6-year retrospective cohort study conducted among patients presented with acute exposure to CNS xenobiotics.

Results

A total of 143 patients’ records were included, where (36.4%) were admitted to the ICU, and a significant proportion of which was due to exposure to alcohols, sedative hypnotics, psychotropic, and antidepressants (P = 0.021). ICU admission was associated with significantly lower blood pressure, pH, and HCO₃ levels and higher random blood glucose (RBG), serum urea, and creatinine levels (P < 0.05). The study findings indicate that the decision of ICU admission could be determined using a nomogram combining the initial HCO₃ level, blood pH, modified PSS, and GCS. HCO₃ level < 17.1 mEq/L, pH < 7.2, moderate-to-severe PSS, and GCS < 11 significantly predicted ICU admission. Moreover, high PSS and low HCO₃ levels significantly predicted poor prognosis and mortality. Hyperglycemia was another significant predictor of mortality. Combining initial GCS, RBG level, and HCO₃ is substantially helpful in predicting the need for ICU admission in acute alcohol intoxication.

Conclusion

The proposed nomograms yielded significant straightforward and reliable prognostic outcomes predictors in acute exposure to CNS xenobiotics.

Introduction

Acute drug poisoning is a significant global problem accounting for multiple morbidities and mortalities. Acute intoxication with central nervous system (CNS) xenobiotics is an increasing global problem, with a special focus in the Middle East.¹ In a 10-year study in Saudi Arabia, 70% of children and 30% of adults were hospitalized for acute exposure to CNS depressants.² Recent studies have demonstrated an increase in hospital admissions due to overdose of CNS xenobiotics, notably during times of crises, such as the COVID-19 pandemic.^3,4

CNS xenobiotics function through different mechanisms and affect neurotransmitters through diverse pathways. CNS depressants are a group of CNS xenobiotics that suppress nervous system functions, inducing sedation in mild doses and coma in large doses.⁵ Gamma-aminobutyric acid potentiation is a common mechanism involved in many CNS xenobiotics, such as benzodiazepines and barbiturates.⁵ On the other hand, psychostimulants, such as amphetamines and cannabinoids, ameliorate other neurotransmitters including serotonin, dopamine, acetylcholine, and norepinephrine neurotransmission.^6,7

In acute drug poisoning, predicting prognosis is critical and can significantly alter the morbidity and mortality outcomes.⁸ Although more severe initial presentations are encountered among poisoned patients than in classical pathological conditions, the prognosis tends to be better, mainly with early intervention and guidance.⁹

Nomograms are evaluation systems that have been used as early as 1975, when Rumack and Matthew established the famous acetaminophen nomogram.¹⁰ Risk prediction nomograms are practical tools that use the already available clinical and laboratory data to predict diverse outcomes. Previous studies have reported multiple advantages of nomograms in terms of their practicality, rapidity, and suitability for educational purposes without the need for computers.^11,12 Nonetheless, nomograms were limited to the prediction of other pathological and postoperative conditions.¹³ Nowadays, few nomograms have become adopted in clinical toxicology, and the few studies on them have conveyed promising findings.¹⁴

Despite the widespread incidence of exposure to CNS xenobiotics intoxication, no published bedside evaluation systems predicting the severities and outcomes have been established. Given the serious effects of these xenobiotics and the numerous exposure opportunities, the current study aimed to identify early findings (clinical and laboratory factors) that could be used as significant risk predictors among patients who overdosed on CNS xenobiotics admitted to the emergency room. Moreover, this study aimed to establish bedside nomograms for identifying patients in need of intensive care unit (ICU) admission and those at risk of poor prognosis or death.

Patients and methods

Study design and setting

The current study is a 6-year retrospective cohort study. Data were obtained from the medical records of adult patients who visited the King Fahad Medical City (KFMC) Emergency Department, Riyadh, Saudi Arabia, and were diagnosed with acute exposure to CNS xenobiotics between January 2016 and December 2021. KFMC has a strategic location in the heart of Riyadh city, the capital of Saudi Arabia, with a total capacity of 1,200 beds. KFMC Poison Control Center is considered to be one of the first poison control centers in Saudi Arabia. The Poison Control Center is staffed by Poison Information Specialists, Clinical toxicologists, and backup Medical Toxicologists and offers services 24 h a day throughout the year. The center covers the second healthcare cluster in Riyadh, Saudi Arabia, serving tertiary and secondary hospitals, small peripheral hospitals, and >40 primary healthcare centers.¹⁵

Sample size

According to Vittinghoff and McCulloch, the concept of event per variable (EPV) of 10 is acceptable for logistic regression.¹⁶ The minimum sample size calculated was 140 participants; thus, we enrolled 143 patients. Logistic regression involves stepwise analysis, resulting in only independent variables with a large effect size.^17,18 Therefore, a lower rule of thumb, such as EPV of 10, was relevant and was subject to a case of medium-to-large effect size.

Inclusion criteria

The current study was conducted among adult patients who were diagnosed with acute CNS medication poisoning during the mentioned period. The diagnosis was based on the history of exposure, recognizing the drug container, clinical examination, and laboratory evaluations, including urine dipstick screening for illicit drugs using the ARCHITECT ci4100 system (Abbott Laboratories, Chicago, IL, United States). The diagnosis was confirmed qualitatively using Gas Chromatography Coupled Mass Spectroscopy (GCMS-QP2010 Ultra system (Shi- madzu, Kyoto, Japan).¹⁹ The availability of quantitative analysis (drug-level assessment) was not considered as one of the inclusion criteria. This type of investigation was not part of the routine investigations requested for all admitted patients. However, whenever quantitative assessment was accessible, it was included. Patients were included regardless of the manner of exposure (accidental, intentional, addiction, or undetermined exposure circumstances).

According to International Classification of Disease, 10th revision (ICD-10), 6 main xenobiotic categories were included and coded. The 19th chapter of ICD-10 describes injury, poisoning, and other consequences of external causes. Codes T36–T50 are assigned for drugs, medicaments, and biological substances, while T51–T65 codes describe the toxic effects of substances chiefly nonmedicinal. T40, T42, T43, T43.3, T43.4, T43.5, T43.6, and T51 were included in the current study.²⁰

• (1) Sedative hypnotics, antiepileptics, and antiparkinsonian drugs: This group involves mainly Barbiturates and Benzodiazepines (Code T42 ICD-10).

• (2) Narcotics and psychodysleptics (hallucinogens), including natural opium alkaloids, synthetic, and semisynthetic opium derivatives (Code T40 ICD-10).

• (3) Antipsychotics and neuroleptics, including phenothiazines, butyrophenone, indole, thioxanthene derivatives, lithium, diazepines, and benzamides (Codes T43.3, T43.4, T43.5 ICD-10).

• (4) Psychotropic and antidepressants, including selective serotonin reuptake inhibitors (SSRIs), monoamine oxidase inhibitors, non-SSRIs, and other unclassified antidepressants (Code T43 ICD-10).

• (5) Alcohols, including any alcohol containing products, including ethanol, methanol, and other unspecified alcohols (Code T51 ICD-10).

• (6) Psychostimulants, including amphetamine, and cannabinoids (Code T43.6 ICD-10).

Exclusion criteria

Patients that met ≥1 of the exclusion criteria were excluded from the study. The exclusion criteria were patients aged <18 years upon admission, patients with incomplete medical records, and patients with concomitant head trauma or other regional injuries. Patients who coingested substances that do not fall under the categories of CNS xenobiotics and those with a query diagnosis were also excluded.

To accurately assess the prognosis and outcomes of every included xenobiotic on individual basis, patients who reported exposure to ≥1 category of CNS xenobiotics were excluded from the current study (i.e. patients exposed to an overdose of alcohol and antipsychotics simultaneously). Additionally, to increase the accuracy of the obtained findings, those who attended the hospital dead were excluded.²¹Figure 1 is a flowchart that describes the eligibility criteria of the enrolled patients.

Fig. 1.

Flowchart of inclusion and exclusion criteria of the patients enrolled in the current study.

Data collection

An anonymous case report form was filled for every enrolled patient to record the demographic data (e.g. age and sex). Exposure history included the category of xenobiotics used, manner of exposure, delay interval from the time of exposure to receiving emergency treatment, and the length of hospital stay. Vital data upon presentation (systolic and diastolic blood pressure, heart rate, respiratory rate, and axial temperature in °C) and the main presenting complaints were reported.

According to the KFMC protocols, the Glasgow Coma Scale (GCS) score and Poison Severity Score (PSS) were calculated for each case. The patients were scored on the GCS as mild (13–15), moderate (9–12), or severe (3–8),²² and on the PSS, as none, mild, moderate, and severe.²³ As the original PSS is complex and incorporates diverse lab investigations which cannot be accurately estimated upon admission, multiple studies modified it according to their contexts.^24–28

The PSS was calculated for every included patient based on the criteria illustrated in Table 1. This modified PSS was adopted from Davies et al. (2008) with slight modifications. Criteria used in this score were (Respiratory [intubated], Neurological [Consciousness status, seizures], and Cardiovascular system [pulse and systolic blood pressure]).²¹ However, we graded the consciousness status into 4 subcategories (fully conscious, somnolence, coma, and deep coma).²⁴ Moreover, we classified the severity grades as per the original PSS into the grade none (grade 0) that refers to normal findings, grade 1 that refers to mild, grade 2 that refers to moderate, and grade 3 that refers to severe. We did not include grade 4, as it refers to those who presented dead (who were already excluded from the current study).²¹

Table 1.

The modified PSS used to assess the intoxication severity in the studied patients.

Body system/intoxication severity	Grade 0 (none)	Grade 1 (mild)	Grade 2 (moderate)	Grade 3 (severe)
Respiratory  • Intubated	• No	• No	—	• Yes
Neurological  • Consciousness status  •Seizures	• Fully conscious • No	• Somnolence • No	• Coma —	• Deep Coma. • Yes
Cardiovascular  • Bradycardia (pulse)  • Tachycardia (pulse)  • Hypotension (systolic blood pressure)	• No • No • No	• >50 pbm • <140 pbm. • >100 mm Hg	• 41–50 pbm • 141–180 pbm • 81–100 mm Hg	• <40 pbm • >180 pbm • <80 mm Hg

The modified PSS could be calculated based on the worst finding in ≥1 of the 3 main body systems (respiratory, neurological, and cardiovascular). It offers the advantage of being easily calculated without including laboratory investigations or several complex measurements.

Laboratory evaluations included serum electrolyte levels (Na, K, and Cl levels in mmol/L), arterial blood gas analysis (including anion gap, pH, PO₂, and PCO₂), HCO₃, random blood glucose (RBG), liver transaminases (aspartate transaminase [AST] and alanine transaminase [ALT]), serum bilirubin, urea and creatinine levels, and complete blood count.²⁹ Out of the studied xenobiotics, we were only able to obtain the quantitative assessment of ethanol blood concentration (using GCMS at cutoff >10 mg/dl).

According to guidelines adopted at KFMC, all cases received supportive fluid therapy and decontamination in cases of recent ingestion. Antidotes, particularly naloxone, flumazenil, and fomepizole were provided if indicated. The patients were grouped according to the primary in hospital outcome, which is ICU admission, into: group I—patients discharged without ICU admission, and group II—patients admitted to the ICU. Moreover, 2 secondary outcomes, namely, prognosis and mortality, were investigated. Prognosis was considered to be poor if patients encountered ≥1 of the following complications: acute kidney injury, respiratory failure, blindness, and cardiovascular complications in terms of arrhythmia. Patients wholly cured without any of these complications were considered to have good prognosis.

The obtained data, including demographic data, exposure history, scoring, and clinical and laboratory evaluation, were reported initially upon admission. By contrast, the reported therapeutic regimens, length of hospital stay, and outcomes (ICU admission, prognosis, and mortality) were reported after the patient was discharged from the hospital.

Compliance with ethical standards

The present study was conducted after obtaining approval from the Institutional Review Board (IRB) of KFMC (approval number IRB00010471) and the College of Medicine of Dar Al-Uloom University (approval number Pro21110002), Riyadh, Saudi Arabia. As stated in the Declaration of Helsinki 1964 and its later amendments, the medical records were handled anonymously, and the patient’s confidentiality was preserved using a coding system for every case report form. The IRB committees waived the requirement for informed consent due to the retrospective and noninvasive study design.

Statistical analysis

Data were analyzed using Statistical Package of Social Sciences version 28 (Inc., Chicago, IL, United States). Qualitative variables were expressed as numbers and percentages, whereas quantitative variables were expressed as mean ± standard deviation (SD) or as median and interquartile range (IQR). The independent t-test and Mann–Whitney U test were employed to analyze quantitative variables, whereas the chi-squared (χ²) test, Fisher’s exact test, and Monte Carlo test were employed for qualitative variable analysis. All variables were evaluated as ICU admission predictors initially via univariate analysis. The factors that showed a significant predictive value (P < 0.05) were further subjected to multivariate analysis.

Multivariate backward binary logistic regression was employed to ascertain the independent predictors of ICU admission and to determine the predictors of the secondary outcomes (prognosis and mortality). To analyze predictors of the need for ICU admission, the regression model was performed collectively by considering all studied CNS xenobiotics and then separately for sedative hypnotics, antiepileptics, and antiparkinsonian drugs (T42) and alcohol (T51) (the 2 most frequently encountered drug categories in the studied cohort). Receiver operating characteristic (ROC) curves, area under the curve (AUC), sensitivity, and specificity were calculated for the significant predictors of ICU admission identified in the multivariate analysis.

Using the STATA/SE software version 17.0, nomograms were created to predict the need for ICU admission, prognosis, and mortality among patients exposed to CNS xenobiotics and to then to predict ICU admission among patients with acute alcoholic intoxication. The nomogram consisted of 3 scales: 1 scale conforming to every predictor, a total score scale, and a probability scale. To use the nomogram, the values obtained for each patient were checked against the scales for each predictor and scored correspondingly. Then, the total score was calculated as the sum of the individual scores for each predictor. Finally, the probability for the outcome to be predicted (i.e. need for ICU admission, prognosis, and mortality) was calculated using the probability scale by correlating to the total score calculated in the previous step. A Kattan-style nomogram, which is appropriate for a binary logistic regression model, was adopted.¹²P < 0.05 was considered to be statistically significant.

Results

The current study included 143 patients. The demographic and exposure data of the patients are presented in Table 2. About 31.9% of group I and 36.5% of group II were aged 25–35 years. There were more men than women in both groups. However, the studied groups had no significant age or sex variations. Fifty-two patients (36.4%) were admitted to the ICU. Alcohol was the most frequently used substance among all patients (46.2%, n = 66) and in each group (49.5% of patients in group I and 4.4% of patients in group II). Out of the 66 patients admitted with alcohol intoxication, 54 patients (81.8%) ingested ethyl alcohol (ethanol), and 12 patients (18.2%) were admitted following exposure to methyl alcohol (methanol). Exposure to methanol was more significantly associated with ICU admission, as all patients diagnosed with acute methanol intoxication were admitted to ICU compared to 16.7% (n = 9) of those exposed to ethanol (P < 0.001).

Table 2.

Demographic and exposure data and presenting complaints of included patients according to the need for ICU admission.

	ICU admission				Total		Test of sig	P value
	Group I (n = 91)		Group II (n = 52)
	No	%	No	%	no	%
Age group  • 18–< 25  • 25–< 35  • ≥35	41 29 21	45.4 31.9 23.1	18 19 15	34.6 36.5 28.9	59 48 36	41.3 33.6 25.2	1.52	0.466
Sex  • Female  • Male	33 58	36.3 63.7	18 34	34.6 65.4	51 92	35.7 64.3	0.03	0.843
Poison type  • Sedative hypnotics  • Narcotics  • Psychostimulants  • Alcohols  • Psychotropic and antidepressants  • Antipsychotics	16 8 7 45 5 10	17.6 8.8 7.7 49.5 5.5 11.0	12 3 0 21 11 5	23.1 5.8 0.0 40.4 21.2 9.6	28 11 7 66 16 15	19.6 7.7 4.9 46.2 11.2 1.5	12.80	0.021a,MC
Exposure manner  • Intentional  • Accidental  • Addiction  • Undetermined	43 17 30 1	47.3 18.7 33.0 1.1	30 11 11 0	57.6 21.2 21.2 0.0	73 28 41 1	51.0 19.6 28.7 0.7	2.99	0.393MC
Presenting complaint (≥1)  • Seizures  • GIT symptoms  • Disturbed consciousness  • Respiratory distress	23 46 26 55	25.3 51.1 28.6 60.4	23 23 38 41	44.2 44.2 73.1 78.8	46 69 64 96	32.2 48.6 44.8 67.1	5.44 0.62 26.50 5.08	0.020a 0.429 <0.001a 0.024a

MC, Monte Carlo test.

Group I includes patients not admitted to ICU, while group II includes ICU admitted patients.

^aSignificant.

Sedative hypnotics were the second-most frequently used substances in both groups. Following alcohol, significantly more ICU admissions were caused by exposure to sedative hypnotics (23.1%), psychotropics, and antidepressants (21.2%). However, out of all psychotropics and antidepressants cases (n = 16), most patients (68.8%; n = 11) required ICU admission. No patients were admitted to the ICU after exposure to psychostimulants (P = 0.024).

More than half of the patients (51.0%, n = 73) were admitted due to intentional self-poisoning. For 1 patient, the manner of exposure could not be determined. The rates of ICU admissions due to accidental exposure and addiction were similar (21.2%, n = 11). The genuine presenting complaint among the studied patients was respiratory distress. Significantly more patients suffered from respiratory distress in group II (78.8%) compared to 6.4% of group I. Disturbed consciousness and seizures were significantly more common in group II (P < 0.05), as presented in Table 2.

Table 3 presents the patients’ vital data, laboratory test results, and scoring upon presentation. Group II patients exhibited significantly lower systolic and diastolic blood pressure than group I patients (P < 0.05). Heart rate, respiratory rate, and axillary temperature did not significantly vary between the 2 groups. In laboratory evaluations, RBG, PCO₂, AST, urea, and creatinine levels were significantly higher, whereas pH and HCO₃ were significantly lower in group II. Acid–base disturbances were significantly more frequent among group I (53.8% vs. 31.9%; P = 0.010). The median GCS in group II was 11.5 versus 15 in group I (P < 0.001). PSS was significantly higher, and GCS was significantly lower in group II.

Table 3.

Vital signs, laboratory investigations and scoring of the studied patients upon hospital presentation according to the need for ICU admission.

	ICU admission				Test of sig	P value
	Group I (n = 91)		Group II (n = 52)
	Mean ± SD		Mean ± SD
Systolic Blood Pressure, mm. Hg	12.9 ± 18.9		106.4 ± 24.4		3.93	<.001a
Diastolic Blood Pressure, mm. Hg	74.7 ± 15.0		68.6 ± 17.3		2.20	0.029a
Heart Rate/min.	97.3 ± 19.4		96.4 ± 29.9		0.20	0.836
Respiratory Rate/min.	21.0 ± 3.1		21.2 ± 5.9		0.18	0.856
Axillary Temperature Co	36.7 ± .4		36.6 ± .6		0.72	0.470
RBG mmol/L	5.8(4.9–7.2)		8.1(5.35–12)		3.25	0.001a
Na + mmol/L	138.2 ± 3.7		138.1 ± 5.7		0.15	0.877
K+ mmol/L	3.9 ± .6		4.0 ± .8		1.31	0.191
CL+ mmol/L	104.9 ± 5.3		105.1 ± 5.7		0.25	0.803
pH	7.3 ± .2		7.2 ± .3		3.21	<0.001a
PCO2 mm. Hg	38.6 ± 1.0		45.2 ± 13.9		3.01	0.004a
PO2 mm. Hg	97.3 ± 5.4		95.3 ± 7.4		1.70	0.091
HCO3 mEq/L	2.2 ± 3.5		15.5 ± 3.5		7.72	<0.001a
Anion gap	18.6 ± 6.8		19.4 ± 6.7		0.74	0.456
Acid base balance: no, % No disturbance Disturbance	62 29	68.1 31.9	24 28	46.2 53.8	6.66	0.010a
ALT unit/L: median (IQR)	20 (15–380)		23 (16–3)		0.40	0.648
AST unit/L: median (IQR)	29 (21–38)		39 (28–49.3)		2.26	0.023a
Bilirubin level μmol/L: median (IQR)	7 (4.10–12)		8.3 (15.4–12.2)		0.77	0.441
Urea level mmol/L: median (IQR)	3.9 (3.2–5)		4.95 (3.4–6.5)		2.95	0.003a
Creatinine level μmol/L: median (IQR)	64 (51–8.8)		85 (53.5–114.8)		3.08	0.002a
WBCsa 103: median (IQR)	7.9 (5.9–1.7)		9.4 (6.4–12.8)		1.73	0.082
RBCsa106	4.8 ± .7		4.7 ± .8		1.30	0.196
Hb g/dL	13.5 ± 2.3		13.2 ± 2.7		0.71	0.476
Plateletsa 103	294.1 ± 90.1		270.4 ± 91.0		1.50	0.134
GCS: median (IQR)	15 (13–15)		11.5 (6.3–15)		6.01	<0.001a
Mild Moderate Severe	79 9 3	86.8 9.9 3.3	22 8 22	42.3 15.4 42.3	38.92	<0.001a
PSS: no, % Non Minor Moderate Severe Fatal	4 51 32 4 0	4.4 56.0 35.2 4.4 0	0 8 18 26 0	0.0 15.4 34.6 5.0 0	48.35	<0.001a,MC

Group I includes patients not admitted to ICU, while group II includes ICU admitted patients. AST, aspartate aminotransferase; WBCs, white blood cells count; RBCs, red blood cells count.

^aSignificant.

The mean value of ethanol blood concentration among patients diagnosed with acute ethanol intoxication was about (222.1 ± 66.3 mg/dL). Comparing groups I and II yielded no significant variations in the ethanol blood concentration between group II (231.7 ± 91.9 mg/dL) and group I (219.5 ± 63.1 mg/dL, P = 0.885).

Significant variations in therapeutic regimens were observed according to the status of ICU admission. Patients in group II exhibited a significant need for antidotal and vasopressor therapies as well as for mechanical ventilation and hemodialysis (P < 0.05). Complications, such as acute kidney injury and respiratory failure, were significantly more frequent with ICU admission (P < 0.001). Unlike blindness, which was a significant complication among patients admitted to the ICU (P = 0.040), arrhythmia did not significantly vary between groups I and II (P = 0.698). Overall, 63.5% of group II patients had poor prognosis (P < 0.001). Similarly, all mortalities in the study population were among group II (P < 0.001). Furthermore, a significantly greater delay interval until treatment and an increased length of hospital stay were observed in group II (P < 0.001) (Table 4).

Table 4.

Therapeutic regimens, complications, and secondary outcomes (prognosis and mortality) of the studied patients according to the need for ICU admission.

	ICU admission				Total		Test of sig	P value
	Group I (n = 91)		Group II (n = 52)
	No	%	No	%	No	%
Antidote	26	28.6	25	48.1	51	35.7	5.48	0.019a
Vasopressor	0	0.0	35	67.3	35	24.6	8.39	<0.001a
Mechanical ventilation	0	0.0	33	63.5	33	23.1	75.07	<0.001a
Hemodialysis	7	7.7	18	34.6	25	17.5	16.62	<0.001a
Respiratory failure	0	0.0	33	63.5	33	23.1	75.07	<0.001a
Acute kidney injury	3	3.3	14	26.9	17	11.9	17.63	<0.001a
Blindness	7	7.7	10	19.2	17	11.9	4.20	0.040a
Arrhythmia	20	22.0	10	19.2	30	21.0	0.15	0.698
Prognosis  • Good prognosis (complete cure)  • Poor prognosis (≥1 complication)	71 20	78.0 22.0	19 33	36.5 63.5	90 53	62.9 37.1	24.41	<0.001
Mortality	0	0.0	9	17.3	9	6.3	16.80	<0.001a,FE
Delay time (hours): median (IQR)	10 (4–24)		24 (12.5–48)		-		3.29	<0.001a
Length of hospital stay (hours): median (IQR)	24 (24–72)		72 (36.5–138)		-		4.51	<0.001a
ICU stay (hours): median (IQR)	-		48 (24–84)		-		-	-

Group I includes patients not admitted to ICU, while group II includes ICU admitted patients. FE, Fisher’s exact test.

^aSignificant.

Table 5 presents the results of the multivariable backward logistic regression analysis for predicting the need for ICU admission for patients exposed to CNS xenobiotics. The prognostic model consisted of (−0.36) HCO₃ + (−0.53) GCS + (1.84) PSS + (−3.04) pH. The model significantly predicted the need for ICU admission (χ² = 101.46, P < 0.001) and exhibited a good contribution (R² = 69.6%) to outcome and accuracy (87.4%). Furthermore, the model was fit (optimistic-adjusted Hosmer and Lemeshow test, P = 0.755) to predict the need for ICU admission. In analyzing the predictors for sedative hypnotics, antiepileptics, and antiparkinsonian drugs, the model found ICU admission to be significantly associated with HCO₃ level (−0.58), whereas for alcohol, the model identified HCO₃ (−0.58) + RBG (0.28) + GCS (−1.67) as predictors. The main predictors of poor prognosis and mortality were low HCO₃ and high PSS (P < 0.05). A high RBG was a significant predictor of mortality but not of prognosis.

Table 5.

Backward binary logistic regression predictors of ICU admission, prognosis, and mortality in the study population.

	P value	Exp(B)	95% CI
			Lower	Upper
Predictors of ICU admission for all studied CNS xenobiotics
• HCO3	<0.001a	0.69	0.59	0.81
• GCS	0.001a	0.58	0.43	0.80
• PSS	0.015a	0.04	0.01	0.55
• pH	0.032a	6.31	1.17	34.05
Predictors of ICU admission for sedative hypnotics, antiepileptics, and antiparkinsonian drugs
• HCO3	0.028a	0.47	0.241	0.92
Predictors of ICU admission for alcohol
• HCO3	0.007a	0.56	0.368	0.85
• RBG	0.007a	1.33	1.079	1.63
• GCS	0.014a	0.19	0.050	0.71
Predictors of mortality for all studied CNS xenobiotics
• HCO3	0.015a	0.79	0.65	0.95
• PSS	0.024a	6.15	1.26	29.91
• RBG	0.030a	1.14	0.0	1.94
Predictors of prognosis for all studied CNS xenobiotics
• PSS	0.005a	3.55	1.43	7.60
• HCO3	0.049a	0.92	0.84	0.99

CI, confidence interval.

^aSignificant.

The diagnostic performance of the significant parameters as predictors of ICU admission (HCO₃, GCS, PSS, and pH) was evaluated via ROC analysis. The parameters were found to significantly predict ICU admission (P < 0.001) and showed good discriminating power to correctly classify patients for ICU admission (AUC = 0.848, 0.799, 0.725, and 0.729 for HCO₃, GCS, PSS, and pH, respectively). The best cutoff points of these predictors showed a higher specificity than sensitivity for HCO₃, GCS, and pH (73.0 vs. 82.0, 50 vs. 97.0, and 42.0 vs. 87.0, respectively), whereas PSS showed a higher sensitivity and lower specificity (85.0 vs. 6.0) as presented in Table 6 and Fig. 2. Figures 3–6 present generalized risk prediction nomogram for predicting the studied outcomes.

Table 6.

ROC curve analysis and validity of the significant predictors of ICU admission.

	AUC, 95% CI	Cutoff value	Sensitivity, 95% CI	Specificity, 95% CI	Accuracy, 95% CI
HCO3	0.848 [0.782–0.941]	≤17.1	73.0 [59.0–84.0]	82.0 [73.0–89.0]	79.0 [71.0–85.0]
GCS	0.779 [0.693–0.866]	≤11	5.0 [36.00–64.0]	97.0 [9.0–99.0]	8.0 [72.0–86.0]
PSS	0.725 [0.640–0.810]	Moderate to severe	85.0 [71.0–93.0]	6.0 [5.0–7.0]	69.0 [61.0–77.0]
pH	0.729 [0.641–0.818]	≤7.2	42.0 [29.0–57.0]	87.0 [78.0–93.0]	71.0 [62.0–78.0]

Fig. 2.

ROC curves of pH, HCO₃, Glasgow coma scale, and PSSs as predictors of ICU admission among the studied patients.

Fig. 3.

Risk prediction nomogram for the need for ICU admission in patients with acute exposure to CNS medication. An example of calculating the probability of ICU admission in a patient presented with acute CNS medication. An imaginary line is drawn from the value of each predictor to the score line. Afterward, the sum of these values on the score line determines the total score. The calculated probability of ICU admission is calculated by correlating the probability scale to the predetermined value on the total score scale. In this case, the probability of ICU admission could be assessed as follows: moderate PSS that has a score of 1.6; GCS = 11 that has a score of 1 and HCO₃ level (mEq/L) = 13.6 that has a score of 2, and pH = 6.82 that has a score of 1. Total score = 1.6 + 1 + 2 + 1 = 5.6 and probability of ICU admission = 0.96.

Fig. 6.

Risk prediction nomogram for ICU admission in patients with acute exposure to alcohols. Using the information in this figure, the probability of ICU admission in a patient after consuming toxic dose of alcohol could be assessed as follows: GCS = 12 has a score of 1.9; RBG (mmol/L) = 7.5 has a score of 1; and HCO₃ level (mEq/L) = 10 has a score of 3.4. Total score = 1.9 + 1 + 3.4 = 6.3 and probability of mortality = 0.9.

Discussion

The current study aimed to investigate the factors associated with ICU admission after acute exposure to CNS xenobiotics and to establish risk prediction nomograms as simple and reliable tools for predicting the need for ICU admission, poor prognosis, and mortality. Apart from the first nomogram developed by Rumack and Matthew (1975)²⁹ for managing patients with acetaminophen intoxication, using nomograms in acute drug exposure is limited. Few studies have established predictive nomograms in the context of acute poisoning. While some of these studies investigated nonpharmacological agents, such as pesticides, or enrolled all patients irrespective of the type of poison, the present study is the first to introduce risk prediction nomograms as outcome predictors among patients exposed to CNS xenobiotics. Moreover, the current study targeted predicting 3 critical outcomes in the form of ICU admission requirement, prognosis (morbidity), and mortality. Table 7 summarizes the studies that have utilized nomograms for patients with acute toxic exposure.

Table 7.

Risk prediction nomograms for patients who presented with acute toxic exposure in the literature.

Study, year	Country	Number of patients used to develop the nomogram	Type of exposure	Factors included in the nomogram	Predicted outcomes
The current study	Saudi Arabia, KFMC	143	CNS xenobiotics	PSS, GCS, HCO3, pH, RBG level	ICU admission, prognosis (complications), and mortality
Elgazzar et al. 202130	Egypt, Tanta University Poison Control Center	1260	All pharmacological and nonpharmacological agents	HCO3, systolic and diastolic blood pressure, respiratory rate, O2 saturation, GCS, pulse, potassium level, and alleged circumstances of poisoning	ICU admission
Dong et al. 202112	The First Hospital of Jilin University and the Lequn Hospital of the First Hospital of Jilin University, Northern China	440	Organophosphorus compounds	Age, white blood cells count, albumin, cholinesterase, blood pH, and lactic acid levels	Severity of poisoning
Lu et al. 202157	China	About 34 out of 80	Paraquat	Rad-score, blood paraquat concentration, creatine kinase, and serum creatinine	Prognosis and mortality
Amirabadizadeh et al. 20208	Iran, Vali-e-Asr Hospital in Birjand	267	All pharmacological and nonpharmacological agents	Age, GCS, white blood cells count, sodium, and serum creatinine levels	Mortality
Buckley et al. 202058	Australia	194	Lithium	Estimated Glomerular Filtration Rate and lithium concentration	expected lithium concentration >1 mmol/L in 36 h as 1 of EXTRIP criteria to judge decision of hemodialysis
Farzaneh et al. 201836	Iran, Imam Khomeini Hospital	68	Aluminum phosphides	GCS, systolic blood pressure, urine output, HCO3, and serum creatinine level	Mortality
Lionte et al. 201759	Romania	180	All pharmacological and nonpharmacological agents	Gender, age, initial lactate, K+, initial CKMB (MB isoenzyme of creatine kinase), the QTc interval, and DT (E wave velocity deceleration time)	Mortality
Lachance et al. 201560	CHU de Québec-Hôtel Dieu de Québec Hospital, Canada	28	Methanol	Initial methanol concentration	Hemodialysis time
Dugandzic et al. 198961	Ottawa Civic Hospital, Canada.	55	Salicylates	Serum salicylate, time of ingestion, and clinical presentation	Severity of poisoning and management decision
Dugandzic et al. 198961	Literature review	129	Noncardiotoxic drugs	QT interval and heart rate	Risk of arrhythmia in the form of torsade de pointes
Waring et al. 198662	United Kingdom	541	Antidepressants	QT interval and heart rate	Risk of arrhythmia in the form of torsade de pointes
(Goldfarnk et al. 1986)63	Bellevue hospital, United States	17	Opioids	Initial bolus of naloxone in mg, infusion rate, 15-min bolus, Target Cp-30 min, Cp-30 min, and Cpss, besides considering the clinical presentation	Continuous infusion of naloxone
Rumack and Matthew et al. 197526	United States	—	Acetaminophen	Plasma acetaminophen concentration and time at >4 h postingestion	Hepatoxicity and need for N-acetyl cysteine

Acute exposure to CNS xenobiotics is a global problem that has increased over the last decades. About 40% of calls to poison control centers were due to exposure to CNS drugs.^4,30,31 Moreover, exposure to this medication category is the leading cause of mortalities from acute toxic exposure.³² In the current study, 36.4% of presented cases required ICU admission, which is a relatively high proportion compared with other studies^33–35. The increased proportion of ICU admission in the current study is attributed to a different set of inclusion criteria. Indeed, this variation reflects the severity and seriousness of this category of xenobiotics.

As demonstrated in the current study, the cases of exposure to CNS xenobiotics mostly occurred intentionally and involved young adults. Intentional exposure to CNS xenobiotics agrees with the previously conducted studies.^30,36 Given the predominance of intentional exposure (51%) and considering that alcohol and sedative hypnotics were the most used substances, serious concerns are raised. However, the literature has reported conflicting results regarding preferences for ≥1 drug categories associated with suicide attempts. The variations are attributed to the availability and accessibility of the substances. Nevertheless, CNS depressants are among the most commonly used substances for suicidal self-poisoning.³⁷

The current study showed a significant association between the need for ICU admission and respiratory distress, pronounced disturbed consciousness, or seizures. Certainly, significant respiratory and CNS distress upon ICU admission indicate the severity of the patient’s condition.³⁶ In agreement with the current study, acute exposure to a high dose of alcohol, which is the most commonly used substance in the current study, was associated with respiratory failure that mostly requires ICU admission.³⁸ Although these conditions are not limited to CNS xenobiotics,³⁴ they are well known for potentiating these effects compared with other drug categories.

This study demonstrated that some laboratory results on admission were among the most significant outcome predictors. HCO₃ level was the most significant predictor of ICU admission, poor prognosis, and mortality. Furthermore, pH < 7.2 was considered as a significant predictor of ICU admission. The significant role of HCO₃ as an outcome predictor (ICU admission) is congruent with the findings of Elgazzar et al., who reported the same findings after exposure to aluminum phosphide and other drug types.³⁴ Additionally, it was reported that high anion gap metabolic acidosis (low pH and HCO₃) was a significant predictor of multisystem organ failure for patients of acute methanol exposure,³⁹ a mortality predictor after exposure to aluminum phosphides,⁴⁰ and morbidity and mortality predictor, particularly with psychotropic and narcotics.⁴¹

The current study highlights the significant power of HCO₃ as a predictor of ICU admission for patients with alcohol intoxication. In agreement, it was reported that metabolic acidosis (low pH and HCO₃ levels) was significantly associated with death following acute methanol exposure.⁴² Similarly, a case series reporting an acute intoxication of 15 individuals due to ingestion of alcohol-based hand sanitizers containing methanol mentioned that all patients developed high anion gap metabolic acidosis and low bicarbonate serum level ranging from <5 to 13 mEq/L. The mentioned study proposed an association between acid–base disturbances and unfavorable outcomes following acute alcohol intoxication.⁴³

Fig. 4.

Risk prediction nomogram for poor prognosis in patients with acute exposure to CNS xenobiotics. In this case, the probability of poor prognosis could be assessed as follows: HCO₃ level (mEq/L) = 1.9 that has score of 3.5, and PSS of grade none that score of 0. Total score = 0 + 3.5 = 3.5 and probability of poor prognosis = 0.07.

Fig. 5.

Risk prediction nomogram for mortality in patients with acute exposure to CNS xenobiotics. Using the information in this figure, the probability of mortality could be assessed as follows: RBG (mmol/L) = 9.2 has a score of 1.9; moderate PSS = has a score of 5.6; and HCO₃ level (mEq/L) = 9.5 has a score of 6.5. Total score = 1.9 + 3.7 + 6.5 = 14 and probability of mortality = 0.13.

Multiple mechanisms are involved in drug-induced metabolic acidosis. The accumulation of acidotic metabolites and ketone bodies (formic acid in alcohol), imbalance in ATP consumption and production (the antiepileptic valproic acid), and impaired renal clearance of acids (propylene glycol-containing medication) are some proven mechanisms. Based on these mechanisms, metabolic acidosis is evidently a result, not a cause, of injury to different body systems.⁴⁴

Another significant predictor of ICU admission identified is the RBG level on admission. The current study showed that RBG level is a significant predictor of ICU admission among patients acutely intoxicated with alcohol. Furthermore, the initial blood glucose level was a significant predictor of mortality among patients acutely exposed to CNS xenobiotics. The association observed between hyperglycemia and complications or mortality after alcohol exposure agrees with the findings of previous reports.^39,45 Consistent with the current findings, hyperglycemia after acute drug poisoning, regardless of the drug type, was associated with poor outcomes and a higher mortality rate in another study.⁴⁶

In strong alignment with the current study, Zadeh et al. described hyperglycemia as a significant mortality predictor in acute methanol intoxication and reported a positive correlation between blood glucose level and base deficit.⁴⁷ Though the mechanism of hyperglycemia with methanol is unclear, pancreatitis is mostly blamed for that, notably during the early hours of admission.⁴⁸ However, to prove this mechanism, reporting autopsy finding in methanol fatalities is suggested. Activation of counterregulatory hormonal secretion as a result of methanol-induced stress might offer a justification for methanol-associated hyperglycemia.⁴⁹ Apart from alcohol, we cannot ignore the strong association between hyperglycemia and susceptibility to infections, reduced immune response, dehydration, electrolyte disturbance, or multisystem organ failure in acute traumatic and hypoxic events.⁵⁰

Apart from considering the metabolic profile and the RBG level on admission, the current study highlights the role of patient scoring using PSS and GCS upon admission. The current study demonstrated that PSS is a significant predictor of ICU admission, poor prognosis, and mortality. Moderate-to-severe scores were highly suggestive of the need for intensive care.

PSS is a tool that has been described as an accurate and reliable tool for assessing the likelihood of poison severity and complications.⁵¹ In agreement with the current study, a higher PSS (>2, moderate to severe) was associated with lower survival (100% mortality) among drug-poisoned patients.⁵² An earlier study conducted in Saudi Arabia among methanol-intoxicated patients yielded similar findings. PSS exhibited a significant predictive power for unfavorable outcomes after exposure to methanol intoxication. Surprisingly, the GCS did not show a similar predictive power, which strongly agrees with the current study findings.³⁹

Inconsistent with the current study, multiple studies have yielded controversial results regarding the efficiency of using PSS in its current form to evaluate the severity of intoxicated patients.^8,51 That’s why we adopted the modified PSS encompassing the clinical finding and excluding metabolic findings and laboratory investigations that had been analyzed separately.

Acute drug poisoning induces a series of biochemical alterations in CNS reactions, resulting in brain damage. Therefore, most patients with acute drug poisoning encounter a change in their mental state and consciousness level, which are reflected in the GCS evaluation.⁵³ The current study proposed a GCS score <11 as the cutoff for ICU admission and described GCS as reliable predictor for ICU in acute alcohol intoxication. In the context of acute drug poisoning, various studies have demonstrated diverse GCS cutoffs for different outcomes. In agreement with the current study, scores <10 were mostly associated with complications,⁵³ whereas scores <8 indicated the need for intubation with antidepressant overdose.⁵⁴ In pediatrics, GCS scores <8 suggest a high risk of mortality.⁵⁵ The variable cutoffs are attributed to the differences in the type of drug, study population, and predicted outcomes.⁵⁶

The significance of GCS in evaluating the severity of acute toxic exposure cannot be denied. The results of the current study concur with multiple studies, where GCS was employed as a predictor of recovery in poisoned patients admitted to the ICU,⁵⁷ as an indicator of the need for mechanical ventilation in patients with antidepressant overdose,⁵⁴ and as a delayed outcome predictor after acute drug poisoning.⁵⁵ Others found that GCS was a significant predictor of mortality and ICU admission after acute pesticide and other types of poisoning.^8,34,40

Regarding alcohol intoxication, low GCS was a risk factor for mortality and poor prognosis, and GCS was negatively correlated with the length of hospital stay in Asian patients diagnosed with methanol intoxication.⁵⁸ Similarly, Tümer et al. found that mental confusion (low GCS) was a factor determining ICU admission and mortality in methanol intoxication.⁵⁹ In disparity, previous literature reported that although about half of patients diagnosed with acute methanol intoxication presented with severely reduced GCS < 8, GCS failed to predict brain injuries and death.⁶⁰ This discrepancy could be explained if we knew that the latter study included traumatized patients with acute alcohol intoxication.

These conflicting reports have raised concerns about the accuracy of GCS in evaluating the condition of poisoned patients. The use of GCS in evaluating acute exposure to sedative hypnotics and opioids might give misleading conclusions. Patients who scored 3 might fully recover within days.⁸ A state of sedation or paralysis may impede proper scoring.⁶¹ Interrater variations and rapidly changing scores upon intervention are among the most common limitations of using GCS for evaluating the severity of drug poisoning.^62,63

As aforementioned, the exclusive use of either PSS or GCS in evaluating the condition of poisoned patients is inadequate and might be misleading. The current study offers an advantage of combining PSS and GCS as 2 of 4 outcome predictors that allow the monitoring of relationships between different variables and increasing the reliability and validity of these scores.

Conclusion

The current study demonstrated that the need for ICU admission could be determined using the initial HCO₃ level, blood pH, PSS, and GCS score. HCO₃ level < 17.1 mEq/L, pH < 7.2, moderate-to-severe PSS, and GCS < 11 are significant predictors of ICU admission. Moreover, high PSS and low HCO₃ level are significant predictors of poor prognosis (in terms of complications) and mortality. Hyperglycemia is another significant predictor of mortality. Combining initial GCS, RBG level, and HCO₃ is substantially useful in predicting the need for ICU admission in patients presented with acute alcohol intoxication. A multivariable risk prediction nomogram for evaluating different outcomes that combines the mentioned factors yields a significant, straightforward, and reliable outcome prognostic tool for evaluating cases of acute exposure to CNS xenobiotics.

Strength and limitations

The current study adds to the existing knowledge and fills a gap in the literature on the management of exposure to CNS xenobiotics. Furthermore, it is the first study to develop a risk prediction nomogram for different outcomes (need for ICU admission, prognosis, and mortality) among patients acutely exposed to CNS xenobiotics. The previous studies included either a single category of xenobiotics or all cases of acute toxic exposure to predict one single outcome. However, conducting the study retrospectively in 1 center with a relatively small number of patients are the main limitations.

Recommendations

We recommend future multicenter studies involving larger data sets. Given the frequent presentation of multiple drug combination, particularly among addicts or suicidal patients, future research investigating the influence of multiple coingestion on different outcomes is highly recommended. Nevertheless, investigating the association between the other studied xenobiotics levels and ICU admission might yield other useful outcome predictors. Furthermore, external validation for the proposed models is recommended to evaluate the model’s generalizability.

Funding

This work was funded by the Deanship of Postgraduate and Scientific Research, Dar Al-Uloom University, Riyadh, Saudi Arabia.

Conflict of interest statement: The authors declare that they have no competing interests.

Data availability statement

Data are available upon reasonable request from the corresponding author.

Authors’ contributions

All authors contributed equally to the study in the form of conceptualization, data curation, data analysis, interpretation, writing original draft, and review and editing the final draft. The corresponding author is the one responsible for fund acquisition.