A one-year observational study of all hospitalized acute poisonings in Oslo: complications, treatment and sequelae

This study was part of a larger one-year observational cross-sectional multicenter study conducted at the outpatient Emergency Medical Agency (EMA, “Oslo Legevakt”), the Institute of Forensic Medicine and the five hospitals in Oslo treating poisoned patients. This paper presents data on their clinical course and treatment in hospitals. Data on the poisoning patterns in hospitals, deaths outside hospitals and on outpatient treatment of acute poisoning are presented separately. For comparative reasons, this study was designed similarly to the studies conducted in 1980 and 2003 [2, 5]. In July 2008, there were 466 423 inhabitants ≥16 years in Oslo, the capital of Norway. As Akershus University Hospital treats patients from two municipalities, patients with residency outside Oslo treated in this hospital were excluded.

All adults (≥16 years) exposed to a substance in toxic amounts and leading to hospital admission were included consecutively from April 15^th 2008 to April 14^th 2009. Chronic poisonings were not included. One fatal case of subacute methotrexate poisoning (medication error) and one fatal case of acute renal failure causing subacute metformin poisoning were excluded. In cases where the poisoning was an additional diagnosis, the poisoning had to warrant admission itself. In cases where ethanol and trauma were codiagnoses, or in doubt, a blood alcohol concentration of 54 mmol/L (250 mg/dL) was used as the cutoff.

Of the 912 individuals studied, 460 (50%) were males and the median age was 36 years (range 16–93), 37 in males and 34 in females. The 23 (2.1%) patients who refused to participate did not differ from the studied population in terms of age, gender, poisoning intention or main agents.

All physicians attending poisoned patients in the five hospitals in Oslo participated in the data collection. All evaluations were done by the treating physician who completed a standardized form that included information that satisfied all of the study objectives. Each hospital had a coordinator (medical doctor) monitoring the study. These coordinators manually cross-checked the forms against the electronic patient lists to ensure that all patients meeting the criteria were included. The electronic data journal was used as a supplement where variables were missing. Variables still missing were coded as unknown and excluded from that particular analysis.

Data were manually entered into an SPSS spreadsheet and systematically checked for errors by randomly cross-checking 10% of the variables with the original forms, showing 99.94% consistency.

In patients exposed to more than one substance, the main toxic agent was defined as the substance suspected to be the most toxic in the amount assumed taken. This was the treating physicians’ clinical evaluation, based on the information available, such as statements from the patient, companions, ambulance service, clinical observations and laboratory findings. Substances presumed to be less toxic were registered as coagents. In six cases, the main agent was unknown. The median time from exposure to hospital admission was recorded for all patients, based on the information available, except where ethanol was the main toxic agent. This was due to practical difficulties in defining the exact time when the ethanol was ingested. Consciousness was measured at the time of hospital admittance. Coma was defined as <8 and drowsiness as 8–14 on the Glasgow Coma Scale (GCS). Complications were classified according to standard definitions, as in the similar study from 2003 [3], and registered in relation to the treatment given. Hypothermia was defined as body temperature below 35°C (95°F), hypoglycemia as serum glucose level <3.5 mmol/L (63 mg/dL). Respiratory depression was defined as respiratory acidosis (pH < 7.3, pCO₂ > 7.0 kPa), hypoxia (pO₂ < 8.0 kPa) or clinical need for ventilation support. Hypotension was defined as systolic blood pressure below 85 mmHg in two subsequent measurements. Conduction disturbances were classified as arrhythmia, but sinus tachycardia was not. Other complications were defined according to general clinical criteria. Each complication was evaluated by the physician as either related or unrelated to the treatment given. Where antidotes had been administered, the exact time for each complication was noted in relation to the antidotes given. Complications arising within one hour after intravenous N-acetylcysteine (NAC) administration and within 10 min after flumazenil or naloxone administration were considered treatment related. Treatment-related complications included among others bronchospasm, hypotension, arrhythmia, convulsions and rashes if not typical of that specific type of poisoning. More unspecific symptoms, such as nausea, vomiting, dizziness, coughing and headache, were not included in this term. Note that neither NAC nor flumazenil was administered in a prehospital setting. The NAC regimen was identical in all hospitals. NAC was administered intravenously according to the Rumack-Matthew normogram in three doses over a period of 21 h [9].

The intention behind the act of poisoning was evaluated by the treating physician. Poisonings with substances of abuse, including heroin and ethanol for recreational purposes, were classified as accidental overdoses with substances of abuse (AOSAs). Suicidal motivation in the act of poisoning was classified as either a possible or definite suicide attempt. This distinction was based on whether the patient considered the toxic agent lethal and whether other measures had been taken to ensure a lethal outcome. If the motive was ambivalent (e.g., by seeking help shortly after ingestion), the poisoning was classified as a possible suicide attempt. Accidents included both external causes of poisoning and self-inflicted accidents (e.g., taking the wrong medication) where the agent had been taken for neither self-harm nor intoxication purposes.

Univariate logistic regression was used to analyze the association between complications and antidotes given. Univariate and multivariate logistic regression were used to identify factors associated with intensive care treatment. All variables included in the multivariate model were clinically relevant and had a p value <0.20 in the univariate analysis. The assumptions underlying multivariate logistic regression analyses were checked and found to be adequately fulfilled. Findings with p values <0.05 were considered statistically significant. SPSS 16.0 (SPSS Inc., Chicago, IL) was used to perform the statistical analyses.

This study was approved by the Norwegian Regional Ethics Committee and the National Data Inspectorate. Studying intoxicated patients is difficult, including acquiring written consent, because of the nature of their behavior and physical condition. The study participants were therefore informed about the aims of the study and were given a written information leaflet with the name and phone number of the study coordinator. They were also given the right to refuse participation or to withdraw consent at any time without reprisal. Thereafter, their verbal consent was confirmed by the independent interviewing physician in a specific question included in the standardized interview form.

The proportion of patients receiving active treatment and antidotes was similar to the 2003 study [3]. The proportion receiving naloxone increased slightly from 2003 (14% vs. 17% in 2008, p = 0.03) reflecting an increase in opioid poisonings (7% vs. 11% in 2008) (Lund C, Teige B, Drottning P, Stiksrud B, Rui TO, Lyngra M, Ekeberg O, Jacobsen D, Hovda KE: A one-year observational study of all hospitalized and fatal acute poisonings in Oslo: Epidemiology, intention and follow-up, submitted). The proportion receiving flumazenil declined slightly (23% vs. 19% in 2008, p = 0.04), although the fraction of poisonings by benzodiazepines was unchanged. Interestingly, the proportion of patients receiving flumazenil diagnostically increased from 23% to 42%, although the total use declined. Thus, patients who were not benzodiazepine intoxicated were more likely to receive flumazenil than before, whereas those who had taken benzodiazepines were less likely to be treated with the antidote, probably because those patients were circulatory and respiratory stable and had a better defined diagnosis. The liberal diagnostic use of flumazenil was addressed after the 2003 study, but obviously with little effect.

Although still high, the rate of gastric lavage was almost halved compared with the 2003 study (17% vs. 9% in 2008) [3]. In comparison, studies from the Netherlands and the UK showed lavage rates of 2% and zero [10, 11], reflecting local differences in treatment policies, and indicating overuse of this procedure in Oslo. According to international guidelines, there is no evidence for gastric decontamination more than 1–2 h after intake, except for intake of very toxic agents, or agents with anticholinergic effects delaying gastric emptying [12, 13]. As such, it is surprising that one-fifth of these procedures were performed after this time frame. In our opinion, there is clinical evidence for gastric decontamination in the most severely poisoned patients (based on a clinical evaluation of complications and overall clinical condition), also with a larger time window between apparent time of ingestion and hospitalization, and such practice is often associated with tablet content removed from the stomach. A probable explanation for this may be delayed gastric emptying, well known to be associated with heavy ethanol intake [14]. We have only found indirect support for this in the literature, as the most severely poisoned patients are often excluded from studies on gastric decontamination. As these patients accounted for a small proportion in this study, the high use of gastric decontamination should be assessed.

It may seem somewhat surprising that the majority of ethanol poisonings received active treatment. However, this consisted almost solely of supportive intravenous fluids. Only 31% of these patients were exposed to multiple agents.

The rate of treatment-related complications was low (2.4%) and the only significant correlation was for intravenous NAC, which induced allergic reactions in 5% of cases. Only one case of bronchospasm mandated treatment termination, most probably related to the initial bolus dose given over one hour. Intravenous NAC has previously been correlated to allergic reactions at rates of 3–6% [15, 16], similar to the present study (5%).

Neither naloxone nor flumazenil, alone or in combination, could be linked to an increased rate of complications, such as convulsions, although this has been shown in previous studies [6, 17]. Although the diagnostic use was liberal, it is likely that physicians did not administer any antidotes under circumstances where they were considered contraindicated. The 2003 study showed an increased rate of convulsions among patients receiving flumazenil and naloxone in combination, but there was no information about the onset of complications in relation to the time of antidote administration [3]. This makes it difficult to determine whether the complications were caused by the antidotes given or if they were merely correlated to the poisoning itself. Poisoning may cause convulsions through respiratory depression and hypoxia [18]. As an illustration, the four patients that developed convulsions prior to antidote administration were hypoxic. Consequently, hypoxia may seem to be the main cause of convulsions in this study. A randomized controlled trial of flumazenil versus a placebo among comatose overdose patients showed no difference in major complications, such as convulsions [7]. Although seemingly safe, the cost-effectiveness of diagnostic use has been questioned [19]. With no significant side effects associated with its diagnostic (over)use, a more restrictive use by doctors wanting a rapid diagnosis in drowsy patients may be difficult to obtain. It has also been argued that this liberal use of flumazenil may reduce the number of cerebral CT scans.

Less than half the patients were awake on admission, illustrating the challenge in determining the main toxic agent and hence the usefulness of knowing the current pattern of poisoning in the area. GHB and ethanol were the most common causes of coma on admission, as they were in the 2003 study [3]. Opioids have now replaced benzodiazepines as the third most common cause. The comatose cocaine-poisoned subjects had also taken sedatives. The total proportion of patients presenting in a coma was slightly lower compared with 2003 (23% vs. 19% in 2008, p = 0.02) [3].

Significantly more patients developed complications as compared with 2003 (18% vs. 30% in 2008, p < 0.001) [3]. This increase was mainly seen in terms of “heavy” complications such as respiratory depression (largely due to opioids), pneumonia, hypotension, hypothermia, convulsions and renal failure. This indicates that the present patient population was more severely poisoned than that studied in 2003. Both the observed increase in suicidal motivation and the increase in opioid poisoning may support this (Lund C, Teige B, Drottning P, Stiksrud B, Rui TO, Lyngra M, Ekeberg O, Jacobsen D, Hovda KE: A one-year observational study of all hospitalized and fatal acute poisonings in Oslo: Epidemiology, intention and follow-up, submitted). A possible triage confounder could be the large number (n = 2348) of low acuity patients treated at the outpatient clinic in Oslo (the EMA), which more than doubled from 2003 to 2008 [20, 21]. As such, the present hospitalized patients may lack some of the “lighter” poisoned patients, as reflected by the fact that those given observation only—and no active treatment—decreased from 25% to 17% [3]. Prehospital treatment of opioid overdoses is common in Oslo [21]; hence, assuming that the population presenting in hospital is more severely poisoned or is exposed to multiple agents seems logical.

Interestingly, half the patients were observed in an ICU, although only 19% were comatose on admission. In comparison, only 5% were observed in an ICU in a Dutch study and the inhospital mortality was zero [10]. One reason for the extensive use is the triage pattern in Oslo, with the majority of low acuity poisonings treated at the EMA. However, the overall proportion treated in an ICU (15%, 519/3413) was still three times as high as reported in the Dutch study. This indicates overuse of expensive resources. Presence of complications was not associated with ICU treatment, in support of this. Nevertheless, the three predictors for ICU treatment, coma, suicidal intention and intake of neuroleptics, seem relevant from a patient safety perspective. Most of these patients were only observed for a couple of hours in an ICU until they were clarified/stabilized and could be referred to an observation unit. Even so, these patients could probably be treated more cost-efficiently without impacting patient safety.

Antidepressants, both selective serotonin reuptake inhibitors and tricyclic (TCA), are known to cause QTc prolongation in overdose [22]. As such, the observed increase in prolonged QTc is interesting as the number of poisonings with antidepressants was unchanged from 2003. One explanation could be the focus on prolonged QTc as a marker for patients at special risk of developing “torsade de pointes” ventricular tachyarrhythmias. The mean number of complications per admission was highest for tricyclic antidepressants (mean 1.3), as in 2003 [3]. Although newer and “safer” antidepressants are available, the proportion of both prescriptions and acute poisonings with TCA has not declined since 2003 [2, 23, 24]. This may reflect that severe depressions are still treated with TCA and these patients are the most likely to attempt suicide.

Although one-third of the patients experienced complications, the mortality was relatively low and few developed sequelae. The inhospital mortality in Oslo was 0.8%, compared with 1.1% in 2003 and 0.5% in 1980 [3, 5]. Both the mortality rate and rate of sequelae are therefore at about the same level as in similar previous studies from Oslo and from other countries [25, 26]. However, the increasing number of poisonings treated at the EMA in Oslo with zero mortality indicates that the mortality actually is lower. Note that all patients dying from cardiac arrest had asystolia (Table 6), known to be associated with a poorer prognosis than ventricular fibrillation [27]. Also note that the only young patient who died, who was not a drug abuser, became an organ donor. The main cause of sequelae was substance abuse poisonings in males, mainly by opioids. Methadone poisoning caused one drug-related death, and permanent sequelae in a suicide attempt. This may reflect increasing rates of methadone-related deaths in the Nordic countries in recent years [28, 29]. Although increasing fatalities, the number of participants in methadone maintenance treatments in Oslo is increasing (n = 1171 in 2009), hence resulting in a relatively low mortality rate [30]. Of all deaths from opioids in Oslo in 2008, 13% (9/72) were due to methadone (Lund C, Teige B, Drottning P, Stiksrud B, Rui TO, Lyngra M, Ekeberg O, Jacobsen D, Hovda KE: A one-year observational study of all hospitalized and fatal acute poisonings in Oslo: Epidemiology, intention and follow-up, submitted).

A major strength of this study is that it was conducted prospectively within a defined geographical area. It included unselected data of all acute poisonings throughout a whole year, eliminating bias from seasonal variations. Since similar studies were performed in the same area in 1980 and 2003, comparisons are more reliable than comparing retrospective studies from different—and often undefined—locations and populations.

In terms of limitations, many physicians assisted in the inclusion of patients. This was an advantage when ensuring complete inclusion, but it may have resulted in high interrater variability. A few patients may also have been missed, despite consecutive inclusion, cross-checking of forms against hospital patient lists, and thorough follow-ups.

It is debatable whether a routine verification of the main toxic agents should have been performed to link the complications to the right toxic agents. Moreover, the presence of other agents may have affected these correlations. However, as treatment is mainly symptomatic and not based on laboratory detection of toxic agents, this study was mainly based on the clinical findings and blood/urine tests used in a routine clinical setting. Furthermore, the limited value of laboratory testing in the clinical setting has been demonstrated [31]. The evaluation of treatment-related complications was limited by a small sample size. All available information was used to estimate the time from exposure to hospital admission, as this information would otherwise be difficult to evaluate.