Background

The severity of poisoning depends on many factors, such as the type and dose of xenobiotics, patient characteristics (age, sex, comorbidities), clinical features at hospital presentation (level of consciousness, blood pressure, pulse rate, respiratory rate, temperature), time to treatment, poisoning circumstances (intentional or accidental; coingestions) and/or laboratory findings (electrolyte imbalances, coagulation abnormalities, renal function etc.) [1,2,3,4,5]. In addition to these predictors, the ICU mortality rate after poisoning also depends upon the class of medications/chemicals to which a patient is being exposed (e.g. opioids, sedatives, street drugs, etc.…), which differs between low- and middle-income countries and countries such as the USA, Australia or European countries. ICU mortality in the USA and European ICUs is ranging from 0 to 6% depending on the study [4, 6,7,8,9].

Previously published studies have reported conflicting data on ICU admission and mortality rates. Comparisons between these studies is difficult because they often lacked common methodology and definitions, were relatively small single-center retrospective studies, or missed information on exposure and on treatment. Additional file 1: Table S1 in the Supplement contains information on the source, number of centers, population, age range, number of patients, most prevalent intoxications, important findings and limitation per study.

The limitations of these previous studies and the lack of multinational database of ICU patients with severe poisonings in Europe were the basis for our prospective study. The INTOXICATE study aims to collect data on admissions to ICUs after acute poisoning in Europe and other continents to determine the rate of eventful admissions among acutely intoxicated adult ICU patients and to provide information on the prognosis of these patients.

Methods

Study design, registration and approval

This prospective multicenter observational cohort study was prospectively registered in an Open Science Framework (OSF) (OSF registration ID: osf.io/7e5uy). The accredited Medical Research Ethics Committee of the University Medical Centre Utrecht (UMCU) did not consider the Dutch Medical Research Involving Human Subjects Act to be applicable to this study (ethics reference number: 20-495/C). The original name was the "TOXIC-Europe study", but the name was changed to the INTOXICATE-study in October 2021 at the request of researchers involved in the Toxicology Investigators Consortium (ToxIC) run by the American College of Medical Toxicology, to avoid confusion.

Setting

ICU physicians in Europe and other continents were invited to participate in the INTOXICATE study through the European Association of Poison Centres and Clinical Toxicologists (EAPCCT) [10] and the European Society of Intensive Care Medicine (ESICM) [11]. The eligibility criteria for ICUs were that they were university affiliated-, community teaching-, and community non-teaching hospitals in Europe or other continents. The ICU could be medical, surgical, specialized in toxicology or any other specialty, or mixed. An ICU was defined as a unit where a patient can be endotracheally intubated and mechanically ventilated. Therefore, high-dependency units (HDUs) or high-care units (HCUs) that can mechanically ventilate patients, were considered an ICU in this study. Ethics approval and signing a data sharing contract were mandatory.

Data were collected between 1st November 2020 and 30th June 2023. A list of collaborators is provided in the Acknowledgments section. The study was managed by a central coordinating team supervising the national coordinators. All participating units provided either local research ethics committee approval or a waiver of consent. A data-sharing contract had to be signed between the participating unit and the coordinating center.

Patients

The patient inclusion criteria were adult patients (aged 18 years or older); patients admitted to the ICU from an emergency department, ambulance, or ward; intoxication (due to poisoning) as the main reason for ICU admission; and patients who stayed in the ICU for four hours or more. Patients were excluded if they were younger than 18 years; admitted to the ICU because of another serious comorbidity (e.g., trauma from a car accident while intoxicated as the management, outcome and duration of admission were likely to be dictated by the comorbidity rather than the intoxication); and an ICU stay of less than four hours. Toxicity was defined as the occurrence of any toxic effect to humans following a single or repeated exposure to a mixture of natural or synthetic substances available on the market or present in the environment. Pure ethanol intoxication was covered by the exposure definition. Informed consent of the participating patients was either required or not, depending on the country and/or the unit.

Variables

The primary outcome, ‘eventful ICU admission’, was a composite outcome defined as the rate of patients who received any of the following treatments in the first 24 h after the ICU admission: oxygen supplementation with a FiO2 > 40%, mechanical ventilation, vasopressors, renal replacement therapy (RRT), cardiopulmonary resuscitation, antidotes, active cooling, fluid resuscitation (> 1.5 L of intravenous fluid of any kind), sedation, or who died in the hospital.

Exposure variables were the exposure exact name, category, dose, and route. Human medications were categorized according to the underlying pharmacological group based on [4]: alcohol (ethanol, other alcohols); analgesics; antidepressants (cyclic antidepressants, lithium); street drugs (opiates, cocaine, amphetamine); sedatives (hypnotics, antipsychotic, benzodiazepines); ‘other poisons’ (carbon monoxide, arsenic, cyanides); other toxins; and mixed intoxications (combination of two or more sub-types of intoxication).

Data were collected on potential predictors (type of units, unit size, country, patient’s age, sex, comorbidities, possible second reason for ICU admission, vital signs, investigations (including ECG).

Data sources

The data entered by local investigators included only information that would have been collected as part of routine clinical care. Local investigators reported only pseudoanonymized data. The data were entered into two study-specific databases (one for units and one for patients) developed in Castor EDC (Electronic Data Capture) [12]. Local investigators could access Castor through an account that required two-factor authentication (2FA). Local investigators entered their data into the database for patients identified as eligible, usually after hospital discharge. Data from Denmark were imported into Castor EDC in a single block (all patients from all Danish units at the same time) because Danish investigators collected patient data using the Redcap system; this made it easier for the Danish investigators to obtain the necessary institutional approval.

Bias

Standard definitions of the variables were provided on the study website. To ensure complete case ascertainment, any missing or inconsistent data were identified at the end of the overall study data collection period and the local investigator was contacted to update/provide the data. We predicted that there would be enough eligible patients in one year. However, the COVID-19 pandemic hit almost immediately after the start of the study, forcing us to extend the study by a further year and eight months.

Study size

Before the study began, we calculated the sample size, based on the hypothesized proportion of outcome in the population of interest, using the following formula:

$$n = \left( {z\frac{{\sqrt {p\left( {1 - p} \right)} }}{d}} \right)^{2} DE$$

Taking the values p = 6.5% (based on [4]); z = 1.96 for a 95% level of confidence; d = 0.065/2 (the allowable error); a correction factor DE = 7.65 for 20 clusters (the 20 countries where the study would be conducted, we get n = 1691. We applied a non-response rate of 10%, which resulted in n = 1691/0.9 = 1879 patients.

Quantitative variables

Quantitative variables concerning the ICUs were the number of ICU beds, the total number of ICU admissions in the last year, and the number of ICU admissions related to poisoning in the last year. All were grouped in categories.

Patient’s age, Body Mass Index (BMI), time elapsed between exposure and hospital admission, number of exposures, systolic blood pressure, heart rate, body temperature, SaO2, arterial pH, potassium, lactate, leucocytes, serum creatinine were considered as quantitative variables. None of them were categorized. The Glasgow Coma Score was categorized in four categories (GCS ≥ 14; GCS > 9 and < 14; GCS > 6 and ≤ 9; GCS < 6).

Statistical methods

Quantitative variables related to ICUs are expressed as numbers and percentages by category. Continuous patient data are expressed as median ± interquartile range. Patient categorical data are expressed as numbers (percentages). Rates were calculated as the number of outcomes divided by the total number of included patients, with the corresponding 95% confidence interval (CI). Rates were calculated before and after exclusion of patients who received mechanical ventilation, vasopressors or cardiopulmonary resuscitation before ICU admission.

Patients with missing data were identified. Percentages were calculated for those with available data, and the denominator with missing data removed is reported throughout. When patients were transferred to another ICU, this second ICU was contacted by the local investigator to obtain the patient data.

In a post-hoc analysis, we used two alternative definitions of ‘eventful ICU admission’ in order to compare our findings with previous studies [4, 13]. The alternative definitions of “eventful ICU admission” included fewer ICU treatments. In alternative definition 1, according to [13], an eventful ICU admission was defined as having received mechanical ventilation and/or vasopressors and/or renal replacement therapy (RRT) and/or cardiopulmonary resuscitation in the first 24 h in ICU or who died in-hospital. In alternative definition 2, according to [4], eventful ICU admission was defined as having received mechanical ventilation and/or vasopressors and/or cardiopulmonary resuscitation in the first 24 h in the ICU, or having died in hospital (similar to alternative definition 1 except that RRT was not considered).

We hypothesized that the rate of eventful admission (with alternative definition 1 and definition 2) was greater than previously reported in the two studies [4, 13]. A one-sample z test was used to test the difference in outcome in this study compared with the rate reported in each study (15.4% in [13] and 6.5% in [4]).

We also performed two sensitivity analyses. First, to mitigate the potential influence of the heterogeneity in enrollment rates between ICUS, we repeated the main analysis after including only units that included at least 80% of the patients admitted to their unit in the study. Second, to minimize the potential influence of a mandatory informed consent, we repeated the main analysis to units where an informed consent was not mandatory.

Statistical analyses were carried out in SPSS Statistics 29.0 and/or R studio version 2023.06.2 for Windows (R version 4.2.2.). The STROBE checklist was used in the preparation of this manuscript following the EQUATOR guidelines.

Results

During the study period, 237 units contacted us (Fig. 1), yet only 78 ICUs that met the inclusion criteria contributed data to the INTOXICATE study (Table 1). Data collection was complicated by various technical challenges, including the COVID-19 pandemic, the data sharing agreement, and the different application of the General Data Protection Regulation (GDPR) in European countries, which may or may not require patient informed consent (Additional file 1: Table S2).

Fig. 1
figure 1

Study Flow Diagram for units. Data entered by local investigators (black outlines), reasons for exclusion (purple), and study phase (blue) are represented

Full size image
Table 1 Location of the INTOXICATE intensive care units (ICUs) (N = 78)
Full size table

Characteristics of intensive care units

Table 2 shows the characteristics of the units. The majority (N = 49, 62%) were university-affiliated ICUs. The size of ICUs was also usually less than 30 beds, and the number of admissions was usually less than 2000 per year per unit. Most units had fewer than 60 poisoning-related admissions in the year before the study. In the majority of ICUs, the doctors who wrote the orders (laboratory tests and medication prescriptions) were specialists in intensive care medicine (62%), and 42% of the units had physicians certified in medical toxicology.

Table 2 Characteristics of intensive care units (ICUs) (N = 78 units)
Full size table

Patient characteristics

A total of 2,273 patients were enrolled during this extended data collection period. The Netherlands was the largest contributor, followed by Denmark and Spain. Table 3 shows the characteristics of the patients. The median patient age was 41 years (IQR 28–56) and slightly more women were affected (53.2% women). Most patients presented a comorbidity, either psychiatric and/or somatic; fifty-nine percent of patients had a coexisting psychiatric illness (other than addiction)diabetes was the most common somatic comorbidity. Most patients were admitted from the emergency department (92.5%), a second reason for ICU admission was recorded in 16.2% of the patients (Table 3).

Table 3 Collected patient data with missing values per variable (n = 2,273 patients)
Full size table

Exposure

The vast majority of the patients (72%) were exposed to intoxicating drugs, and almost a quarter were exposed to alcohol, most often mixed with another drug. The group of mixed intoxications accounted for 1131 cases (49.8%) in all patients (Table 3). When considering isolated intoxications, sedatives (10.6%) and street drugs (8.9%) were most frequently used.

Clinical features

Neurological signs and symptoms were the most common on admission, with the top three most common neurological symptoms being altered consciousness (N = 979 patients, 43% of admissions), coma (N = 747 patients, 32.9% of admissions), and agitation (N = 294 patients, 12.9% of admissions). Respiratory and gastro-intestinal signs were also frequently observed (at least one sign observed in 41.1% and 23.5% of admission, respectively). The most frequently observed cardiologic signs or symptoms were palpitations (4.8%) and hypotension (4.7%).

Primary outcome and in-hospital mortality

The observed rate of patients with an eventful ICU admission was 68% (95% CI: 64.6%; 71.4%) (n = 1546/2273 patients) in all patients. Six hundred and eighty-eight patients (n = 688/2273, 30.3%) received an ICU intervention (CPR or mechanical ventilation or vasopressors for at least one hour) prior to their ICU admission, and for 2 patients, it was unknown whether they had had a treatment before their ICU admission (n = 2/2273, 0.1%). The observed rate of patients with an eventful admission) was 56.5% (95% CI: 53%; 60%) (n = 895/1583 patients) after exclusion of patients who received an IC intervention before their ICU admission (Fig. 2). For 9 patients, the treatment received in ICU was missing (n = 9/1583, 0.6%).

Fig. 2
figure 2

Study Flow Diagram for patients in the main analysis. An “eventful ICU admission” was defined as receiving an ICU intervention within the first 24 h after ICU admission or in-hospital death. An ICU intervention was defined as receiving any of the following treatments: oxygen supplementation with a FiO2. 40%, mechanical ventilation, vasopressors, renal replacement therapy (RRT), cardiopulmonary resuscitation, antidote, active cooling, fluid resuscitation (> 1.5 L of intravenous fluid of any kind), and sedation. ICU intervention before ICU admission was defined as receiving mechanical ventilation or cardiopulmonary resuscitation or vasopressors (over at least 1 h) before ICU admission

Full size image

The majority of patients survived to hospital discharge, with 3.7% (n = 85/2273 patients) dying in the ICU and 0.8% (n = 18/2273 patients) dying in the ward following discharge from the ICU, resulting in an in-hospital mortality of 4.5% (n = 103/2273) (95% CI: 3.7%; 5.4%).

With the post-hoc analysis, the rate of ICU eventful admission after exclusion of the patients who received an ICU intervention before their ICU admission was 21.1% (n = 335/1583) when alternative definition 1 was used (Fig. 3, left panel). This rate was significantly greater than 15.4%, the rate reported previously in [13] (z-statistic = 6.47, p < 0.001). When alternative definition 2 was used, the rate of ICU eventful admission after exclusion of the patients who received an ICU intervention before their ICU admission was 18.9% (n = 299/1583) (Fig. 3, right panel). This rate was significantly greater than 6.5%, the rate of ICU eventful admission reported in [4] (z-statistic = 20.1; p < 0.001).

Fig. 3
figure 3

Study Flow Diagram for patients in the post-hoc analysis (using alternative definitions 1 and 2 for “eventful ICU admission”). In alternative definition 1, an “eventful admission” was defined as receiving mechanical ventilation and/or vasopressors and/or renal replacement therapy and/or cardiopulmonary resuscitation in the first 24 h after ICU admission, or in-hospital death (as in [13]. In alternative definition 2, only receiving mechanical ventilation and/or vasopressors, or in-hospital death were included in the definition of an “eventful ICU admission” (as in [4]. ICU intervention before ICU admission was defined as receiving mechanical ventilation or cardiopulmonary resuscitation or vasopressors (over at least 1 h) before ICU admission

Full size image

When including only the units that included at least 80% of the patients admitted due to intoxication to their unit in the study, the rate of eventful ICU admission was 68.7% (n = 574/835) before exclusion of the patients who received an ICU treatment before their ICU admission (versus 68% [95% CI: 64.6%; 71.4%] in the 2273 patients. The in-hospital mortality rate was 4.2% (n = 35/835) versus 4.5% [95% CI: 3.7%; 5.4%] in the 2273 patients.

When including only the units where an informed consent was not mandatory, the rate of eventful ICU admission was 71.6% (n = 756/1056 patients) before exclusion of the patients who received an ICU treatment before their ICU admission (versus 68% [95% CI: 64.6%; 71.4%] in the 2273 patients). The in-hospital mortality rate was 4.5% (n = 48/1056) versus 4.5% [95% CI: 3.7%; 5.4%] in the 2273 patients.

Discussion

The primary findings of our study show that almost all patients presenting with acute intoxication had comorbidities, with psychiatric comorbidities being the most common. The majority of intoxications in our study involved human medications. The overall mortality rate was low (4.5%). About two thirds of the patients admitted to the ICU received ICU-specific treatments, but this percentage drops to about 56.5% when excluding patients who had already received an ICU intervention before admission.

Our study confirms several findings from previous research. Consistent with previous studies, we observed a slightly higher number of females than males among the intoxicated patients [8]. The predominance of mixed intoxication [7, 13,14,15,16,17] and intoxicating drugs as the cause of intoxication [7, 9, 16, 18,19,20] is also in line with the existing literature. The low mortality rates both in the ICU (3.7%) and in the hospital (4.5%) are in accordance with the mortality rates reported in the literature. The ICU mortality rate ranged from 0.4 to 5.9% when considering studies with more than 100 intoxicated patients (Table S1) [4, 7,8,9, 19,20,21,22], while the in-hospital mortality rate reported in the literature ranged from 0.7 to 6.7% [1, 4, 9, 13, 14, 18, 20, 22].

However, our study differs significantly in the proportion of intoxicated patients admitted to the ICU who required mechanical ventilation, vasopressors or died in hospital. We found this proportion to be significantly higher than that reported in a French study that also included renal replacement therapy (RRT) as part of ICU-specific treatment (15.4%)[13] and a Dutch study that did not include RRT (6.5%) [4]. This suggests that our cohort had a higher severity of intoxication, as our criteria for ICU treatment were similar to prior studies [4, 13].

The strengths of our study are many. We achieved a high level of data completeness and quality, with very few missing values. Our prospective study design, in contrast to the retrospective nature of most previous studies, increases the reliability of our findings. The international scope of our study, covering approximately 20 countries, increases the generalizability of our findings. With 2,273 admissions, the sample size of our study is robust and exceeds many previous single-center or single-country studies. In addition, our inclusive criteria, covering all types of poisoning rather than focusing solely on suicides or intoxicating drugs, provide a comprehensive overview of the problem.

However, our study has limitations. There was an imbalance in enrolment between countries, with six countries (the Netherlands, Denmark, Spain, Sweden, Turkey and Belgium) contributing more than 75% of the patients. This bias does not reflect the population size of these countries in Europe. However, we believe that the management of ICU patients after acute poisoning does not differ significantly between European countries or between Europe and Australia. Therefore, the 3.7% ICU mortality observed in the study seems to be a plausible estimate for European countries, although variations may be more pronounced in regions with different resources, inpatient care or patients’ exposure. The sensitivity analysis showed that the effect of the heterogeneity of enrollment between units represented a limited bias since the rates of eventful ICU admission and in-hospital mortality rate were comparable (68% eventful ICU admissions and 4.5% in-hospital mortality rate in the total study sample versus 68.7% eventful ICU admissions and 4.2% in-hospital mortality rate in the units that included at least 80% of the patients admitted to their unit).

In addition, the over-representation of university hospitals in our study may indicate a bias towards more complex cases due to the research focus and capabilities of the centers. Future analyses will need to investigate whether patients at university centers had more severe exposures or comorbidities. Finally, written informed consent was mandatory in many intensive care units, which meant that we could not know how many patients were excluded from the study. We have therefore missed a certain number of patients. This may cause a selection bias, because the prognosis of the patients may vary according to whether an informed consent was considered necessary or not. However, the sensitivity analysis showed that this effect was limited.

Conclusions

Our results show a higher rate of intoxicated patients being treated in the ICU than has been reported in previous studies. Comprehensive data has been successfully collected on a large cohort of patients admitted to the ICU after acute intoxication, predominantly from European ICUs, with some representation from other continents. Future research needs to look more closely at outcomes by type of intoxication, externally validate existing prediction models predicting the need for ICU admission, identify risk factors for complicated intoxications, perform competing risk analysis for likelihood of discharge, and assess the prognosis of patients after specific exposures, such as street drugs. However, this requires large and detailed databases. INTOXICATE is a first step towards such a granular database. The findings from this study will inform future research efforts, particularly in understanding prognosis and refining data collection methods for similar studies.