a b s t r a c t

Context: routine biological tests are frequently ordered in self-poisoning patients, but their clinical relevance is poorly studied.

Materials and methods: This is a prospective multicentric observational study conducted in the emergency de- partments and intensive care units of 5 university and nonuniversity French hospitals. Adult self-poisoning pa- tients without severely altered vital status on admission were prospectively included.

Results: Routine biological test (Serum electrolytes and creatinine, Liver enzymes, bilirubin, blood cell count, pro- thrombin time) ordering and results were analyzed. A total of 1027 patients were enrolled (age, 40.2 +- 14 years; women, 61.5%); no patient died during the hospital stay. Benzodiazepine was suspected in more than 70% of cases; 65% (range, 48%-80%) of patients had at least 1 routine biological test performed. At least 1 abnormal test was registered in 23% of these patients. Three factors were associated with abnormal test results: age older than 40 years, male sex, and poisoning with a drug known to alter routine tests (ie, acetaminophen, NSAIDs, metformine, lithium). Depending on these factors, abnormal results ranged from 14% to 48%. Unexpected severe life-threatening conditions were recorded in 6 patients. Only 3 patients were referred to the intensive care unit solely because of abnormal test results.

Conclusion: Routine biological tests are commonly prescribed in nonsevere self-poisoning patients. Abnormal re- sults are frequent but their relevance at bedside remains limited.

(C) 2016

Introduction

Self-poisoning is a frequent cause of admission both in the emergency department (ED) and in the intensive care unit (ICU) (2%-4%) and is asso- ciated with important cost [1]. Most patients have an uneventful “somat- ic” course and mainly require psychiatric care [1,2]. However, depending

? This work was financially supported by institutional funding: Angers University, An- gers and Angers University Hospital, Angers, France.

?? All authors declare no conflict of interest.

? The work was performed in 5 French centers: Angers University Hospital, Brest University

Hospital, Cochin University Hospital (Paris), Le Mans Hospital, and Saint Malo Hospital, France.

* Corresponding author at: Departement de Reanimation Medicale et de Medecine Hyperbare, Centre Hospitalier Universitaire, 4 rue Larrey, F-49 933 Angers Cedex 9, France.

E-mail address: [email protected] (N. Lerolle).

on the dose and the composition of the drugs involved, severe drug poi- soning with organ dysfunction and potentially death may occur. Mortality of Poisoned patients remains generally less than 1% but may reach 10% and more for some cardiotropic drugs [3-7]. In the ED, among the vast amount of patients with low severity poisoning, tracking those at risk for organ dysfunction who require specific management is crucial.

It has been observed that repeated evaluation of vital signs allows detection of most case of Severe poisoning [2,8]. A consensus has been reached that toxicological analysis should be performed only for select- ed drugs (ie, lithium, paracetamol, digoxin…) with obvious concentration-toxicity correlation [9]. In this context, ordering of rou- tine biological tests (ie, serum electrolytes and creatinine, liver en- zymes, blood cell count) to assess organ damage before clinical onset makes sense, although it has been seldom evaluated [10-12].

http://dx.doi.org/10.1016/j.ajem.2016.04.002

0735-6757/(C) 2016

We set up a prospective observational multicenter study to assess cur- rent medical practice for ordering of routine biological tests in self- poisoning patients and to describe tests results. In accordance with the aim of detecting subclinical organ dysfunction, we restricted our study pop- ulation to self-poisoning patients presenting to the ED without obvious or imminent organ failure. The main objective was to describe the relevance of these tests to detect life-threatening condition requiring emergent care or ICU admission. A secondary objective was to search for criteria selecting a population with a low likelihood of abnormal test results.

Patients and methods

This prospective Multicenter observational study was conducted from December 2011 to June 2012 at the ED of 3 tertiary university hospitals (Angers, Brest, Cochin APHP) and 2 general hospitals (Saint Malo, Le Mans). We aimed to enroll at least 200 patients at each center. All partic- ipating sites routinely manage self-poisoning patients. Approval for this study was obtained from the ethics committee of Angers University Hos- pital, France, which waived the need for patient written consent.

Patient eligibility

All adult (>=18 years of age) patients with a suspected self-poisoning, evaluated at the ED of the participating hospitals, were considered. We enrolled patients with intentional self-inflicted drug overdose, irrespec- tive of the motivation (isolated alcohol intoxications were excluded). Involuntary intoxications and self-harm episodes without drug use were not considered.

Patients could not be included if they met at hospital admission one of the following criteria: Glasgow Coma Scale (GCS) less than 8; systolic blood pressure less than 90 mm Hg on 2 consecutive measurements; SpO₂ less than 90% on room air; respiratory rate less than 10 and greater than 25 per minute.

Data collection

The attending emergency physician prospectively collected the fol- lowing items: physician status (fellow/resident or attending physician [board certified in emergency or intensive care medicine]), patients de- mographic and anamnestic data (age, sex, medical history, current treat- ment, class(es) of suspected drugs ingested, presence of concomitant suspected alcohol ingestion), admission Vital parameters (GCS, heart and respiratory rates, systolic and diastolic blood pressures, and pulse ox- ygen saturation on room air), routine biological tests performed in the ED and abnormal results (serum electrolytes measurement [Na, K, Cl, bicar- bonate, glucose, together with urea and creatinine], arterial blood gases, blood cell count, lactate, prothrombin time, liver function tests [any of al- anine aminotransferase {ALT}, aspartate aminotransferase {ALT}, bilirubinemia], and others; electrocardiogram and chest x-ray were re- corded), toxicological analyses (ordering and results), therapy undertak- en in the ED (tracheal intubation, renal replacement therapy, activated charcoal, antidote [flumazemil, N-acetyl cysteine, naloxone, or other], va- sopressor, antibiotic, gastric lavage, or other treatment), referral after ED (ICU, high dependency unit, medical short term care unit, conventional Inpatient hospitalization, psychiatric ward, outpatient management, run- away or death), and outcome (length of hospital stay, secondary referral to ICU or high dependency unit, hospital mortality).

Physicians in charge of the patients received a specific training for data collection. No specific instructions were given regarding test order- ing and patient care: physicians were instructed to act “as usual.”

Definitions for classifications and thresholds

The following patients classifications were established a priori: med- ical history (absence of any ongoing disease or presence of any ongoing disease requiring current drug treatment or not); current treatment (no

treatment, any treatment); and type of ingested drugs according to poi- son center pharmacist investigator (GLR): drug group 1: all of the suspected ingested drugs being either benzodiazepines, opioids, nontriCyclic antidepressants as Selective serotonin reuptake inhibitors, antihistaminics, neuroleptics, or sleep inducers (this group thus includ- ed patients for whom the suspected ingested drugs were not expected to modify routine biological test in case of overdose except blood gases analyses in case of hypoventilation); drug group 2: any of the suspected ingested drug being cardiotropic drugs, Antiepileptic drugs, tricyclics, or carbamates (these drugs were not expected to modify rou- tine biological tests directly but secondarily due to acute organ failure such as shock) without any suspected drug qualifying for group 3; drug group 3: any other drug including notably paracetamol, nonsteroi- dal anti-inflammatory drug (NSAID), lithium, metformine (these drugs were suspected to potentially modify directly routine biological tests in case of overdose). Unknown ingestions were assigned to group 3.

Thresholds qualifying for presence of relevant vital parameter ab- normalities were also defined before study onset: GCS less than or equal to 14, heart rate less than 50 or greater than 110 per minute, sys- tolic blood pressure less than 95 or greater than 160 mm Hg, oxygen sat- uration on room air less than 92%.

For routine biological tests, as reference ranges for normal values varied across center laboratories, we decided to set up clinically relevant thresholds. These thresholds were obtained by consensus between cli- nician and poison center investigators for abnormalities that should be actually taken into consideration: serum sodium less than 133 or great- er than 147 mmol/L, serum potassium less than 3.3 or greater than 5 mmol/L, bicarbonate content of plasma less than or equal to 20 or great- er than or equal to 30 mmol/L, serum creatinine greater than or equal to 100 umol/L, prothrombin time less than or equal to 70%, hemoglobin concentration less than or equal to 12 g/dL, platelet count less than or equal to 150 000/mm³, AST greater than 50 IU/L, ALT greater than 70 IU/L, bilirubinemia greater than 20 umol/L.

Abnormalities of routine biological tests were reviewed independent- ly by 3 investigators (TR, GLR, and NL) for qualification as life-threatening abnormalities requiring emergency treatment and/or ICU admission. In case of disagreement, consensus was sought by discussion; in case of persisting disagreement, the abnormality was retained as life threatening. Purpose for transfer to ICU or high-dependency unit was evaluated by the independent investigators by reviewing the drugs involved, presence of clinical abnormalities, and routine test results.

Statistical analyses

All eligible patients with required data at baseline were included in the analyses. Categorical variables were expressed as frequency and percentage; continuous data were presented as mean and SD. Multivar- iate analyses were performed using logistic models. The significance of covariates was tested by the use of Wald tests. All tests were performed with a type I error set at 0.05. Analyses were performed using Stata 11.0 (StataCorp LP, College Station, TX).

Results

Patient characteristics and outcome

One thousand twenty-seven patients were recruited during the study period. Mean age was 40.2 +- 14 years old, and 61.5% were women. A junior physician provided care in 40% of cases (range among centers, 15%-97%). Number of drug classes suspected per patient was 1.6 +- 0.95 (range, 1-6). Drugs suspected ingested and ethanol are displayed on Fig. 1; 764 patients (74%) were drug group 1; 83 (8%),

drug group 2; and 180 (18%), drug group 3. Distribution of baseline vital parameters is displayed on supplementary online material. Pres- ence of at least 1 relevant vital parameter abnormality was observed in 386 patients (37.6%). Based on these data, 970 patients (94.4%)

Fig. 1. Classes of toxic suspected ingested.

were classified as “clinical” PSS 0 or 1, and 57 patients (5.6%) were clas- sified as “clinical” PSS 2 due to GCS less than 12 in all of them. Because of exclusion criteria, no patient was “clinical” PSS 3 at admission.

Referral after the ED is shown in Fig. 2. Length of stay in the ED was not measured but was less than 6 hours for all patients before referral. Eight patients (0.8%) were secondarily transferred to an ICU or high- dependency unit because of worsening clinical condition in the first 72 hours. Finally, all patients were discharged alive from the hospital,

30.5 +- 55 hours after ED admission (range, 0.3-798 hours).

Toxicological analyses performed in the ED

Acetaminophen blood content was assayed in 307 patients (with subsequent requirement for N-acetyl cysteine in 49). Ethanol Blood concentration was assayed in 557 patients and detected in 311 (mean 1.76 +- 0.97 g/L; range, 0.1-5.1 g/L). Sixteen patients had a blood or urine “screening panel,” and 84 patients had specific serum quantitative

measurements of toxics other than paracetamol (eg, lithium, Valproic acid, digoxin, other).

Routine biological analyses in the ED

A total of 671 patients (65%) were prescribed at least 1 routine biolog- ical test. In these, the mean number of tests prescribed was 1.75 +- 1.6 (range, 1-6) (among serum electrolytes measurement including serum creatinine and urea nitrogen, arterial blood gases, blood cell count, lactate, prothrombin time, liver function tests and others, each of these counting for one). Detail of tests prescribed is presented in Table 1; in addition, 21 patients had arterial blood gases obtained, and 12 had a lactate measure- ment. Thirty-seven (3.6%) and 578 (56%) patients had a chest x-ray and an electrocardiogram performed, respectively.

Ordering of routine biological tests varied from 48% to 80% of pa- tients across the 5 centers. Identically, ordering varied across drug groups (group 1, 58%; group 2, 82%; group 3, 89%). In multivariate

Fig. 2. Referral after initial evaluation.

Table 1

Ordering of routine biological tests and frequency of abnormal results according to clini- cally relevant thresholds

considered as relevant by the clinicians of this study (Table 1), referred subsequently as abnormal results. Frequency of abnormal results varied from 21% in group 1 to 39% in group 3. Distribution of the values outside

Routine analysis and thresholds No. of patients with

analysis prescribed (% refers to 1027 patients)

Results outside threshold range (% refers to n prescribed)

of threshold ranges is showed in Fig. 3. Thirteen patients (1.3% of the whole population) were considered to have potentially life- threatening blood test abnormalities (see online supplementary for de- tails). Such abnormalities were expected when considering the nature

Any routine test 671 (65.3%) 152 (22.6%)

Serum electrolytes and serum creatinine 657 (64%) 83 (13%) Na (b133 or N 147 mmol/L) 19 (2.9%)

K (b 3.3 or N 5 mmol/L) 28 (4.3%)

of the drug ingested and/or presence of vital status impairment in 5 pa- tients (ie, low prothrombin time in a patient with vitamin K antagonist in- toxication) and unexpected at admission in 8 (hypokalemia b 3 mmol/L of

Bicarbonate

(<= 20 mmol/L or >= 30 mmol/L)

29 (4.4%)

unknown origin in 6 patients, 2 patients with severe renal failure eventu- ally attributed to chronic kidney diseases). Three patients (0.3%) were

Serum creatinine (N 100 umol/L) 11 (1.7%)

Liver enzymes and bilirubin	332 (32%)	48 (14.4%)
ALT (N 70 IU/L)		20 (6.0%)
AST (N 50 IU/L)		38 (11.4%)
Bilirubin (N 20 umol/L)		8 (2.4%)
Prothrombin time (b 80%)	365 (36%)	14 (3.8%)

Blood cell count 412 (40%) 26 (6.3%)

Hemoglobin (b 12 g/dL) 21 (5.1%)

Platelet count (b 150 G/L) 5 (1.2%)

analysis (see supplementary online material), center, drug supposed ingested group 2 and 3, and GCS score less than 15 were significantly as- sociated with increased ordering.

Results of routine biological tests performed in the ED

Among the 671 patients who were prescribed routine tests, 152 (23%) had at least 1 biological test result outside the threshold ranges

transferred to the ICU solely based on test results, in the absence of altered vital status (displayed on supplementary online material). None of the pa- tients with no routine test prescribed on initial evaluation was transferred primarily or secondarily to an ICU or high-dependency unit.

Table 2 shows admission parameters associated with the presence of at least 1 abnormal test result (potentially life-threatening or not) among patients with at least 1 test carried out. Three factors were associated with abnormal routine biologic tests age older than 40 years, male sex, drug group 3, at the limit of statistical significance for the2 former. Comparison of patients having 0 to 3 of these criteria showed a significant difference for the frequency of abnormal routine tests from 14% to 48% (Fig. 4). How- ever, 2 patients with unexpected potentially life-threatening blood test abnormalities (K b 3 mmol/L) had 0 criteria.

Of note, modifying the thresholds by strictly applying each hospital laboratory’s reference ranges increased the percentage of patients with at least 1 abnormal test result up to 36%, but without affecting sig- nificantly the results of the analyses (data no shown).

Fig. 3. Distribution routine test values outside the thresholds defined as clinically relevant by the physicians of this study. Patients with results within thresholds are not displayed. The percentage of patients displayed on the figure over all patients with the test prescribed is indicated.

Table 2

Parameters independently associated with presence of at least 1 abnormal test result among patients with at least 1 realized

OR 95% CI P

Factors associated with abnormal routine test results Age

<= 40 years old	1.0
N 40 years old	1.5	1-2.2	.08
Sex Female	1.0
Male	1.5	1-2.3	.07
Toxic
Group 1	1.0
Group 2	1.5	0.7-3.0	.26
Group 3	1.7	1.1-2.7	.03

Parameters entered in the multivariate analyses were age, sex, type of drug ingested, suspected alcohol ingestion, vital parameters, medical history, and current treatment. Only parameters with Pb .1 are displayed

Abbreviations: OR, odds ratio; CI, confidence interval.

Discussion

This prospective multicenter practice survey evaluated 1027 self- induced drug overdose patients without severely altered vital status on admission. The frequency of routine biological tests was high (65%), most notably serum electrolytes and creatinine, liver enzymes, prothrombin time, and blood cell count. Class of suspected ingested drug and altered vital status impacted the frequency of ordering, but wide variations were observed between centers. Percentage of patients with relevant abnormal test results using clinician-based thresholds was as high as 23% and up to 36% when strictly considering laboratory reference ranges. Tests results rarely showed vital abnormalities (1.3%) and impacted exceptionally upon ICU referral (0.3%). As no death was observed in this study, performing or not a routine biological could not be tested for an association with vital outcome. Frequency of abnormal test results could be partly predicted by a combination of 3 factors: age older than 40 years, male sex, and toxic group 3 (ie, paracet- amol, lithium, NSAIDs, metformine…). A low likelihood of abnormal re- sult could be predicted in absence of any of these criteria, but it does not rule out presence of vital abnormalities, although exceptional.

Our study included 5 French centers of different sizes, urban and

rural, which allows to describe adequately both the heterogeneity of pa- tients with drug overdose and the practice of care in France. General purpose of the study was presented to the physicians before study onset; specifically, equipoise as to whether routine tests are useful or not was put forward. Therefore, physicians were asked to order biolog- ical routine and toxicological tests “as usual.” Main characteristics of our population selected on the absence of severely altered vital status on ad- mission was in accordance with previous publications regarding sex, age, outcome, and type of drugs in France but also in many other

Fig. 4. Percentage of abnormal routine test result according to the number of risk factors among age older than 40 years old, toxic group 2, female.

countries [13-15]. Benzodiazepines are indeed frequently reported as the leading class of drugs in drug overdose, its frequency ranging from 20% to 67% with the exception of Great Britain where paracetamol is the most frequent toxic encountered [16-18]. Considering the differ- ences of reference ranges between laboratories of the participating cen- ters, we chose to analyze our data based on relevant clinician-based thresholds. These are indeed debatable, but as are any thresholds re- garding laboratory data [15].

The very low frequency of severely altered routine tests contrasts, however, with the high frequency of abnormal results, even when choos- ing thresholds considered as clinically significant by emergency physi- cians. The usefulness of knowing abnormal but low severity results for the care of overdosed patients is very difficult to determine. These may have resulted in various therapeutic short-term interventions that this study was not designed to analyze. However, the wide variation in prac- tice between centers, without any apparent impact on outcome, does not argue for a major beneficial impact of such interventions at least in the short term. Our data showed that female patients who had ingested a toxic of the first group (ie, benzodiazepine, non cyclic antidepressant…) and who were younger than 40 years had a low likelihood of abnormal routine tests results. These criteria may be proxies for the severity of the intoxication and baseline health status, which both impact blood abnor- malities. As abnormal test result frequency in this population was still greater than 10%, these criteria can only be considered as a guide for thoughtful ordering rather than a rule, notably as these criteria do not rule out exceptional life-threatening abnormalities.

Based on our results, it is difficult to recommend whether routine bi- ological tests should be performed systematically in case of drug poison- ing or could be avoided in some patients. Variation in practice in the literature regarding care of poisoned patients, from systematic inpatient comprehensive assessment to early discharge after ambulance team eval- uation, already reflects different pattern of care for similar patients [19]. According to such already existing pattern of care, it is likely that our re- sults may be interpreted differently: centers performing outpatient care may retain that routine biological test very rarely detects vital abnormal- ities; centers with comprehensive assessments may hold that a few life- threatening abnormalities were detected by blood analysis. Our results highlight the challenge of clinical toxicology that has to tackle simulta- neously a huge amount of patients intoxicated with a limited number of drugs that do not represent any Diagnostic difficulty (ie, benzodiazepine) and the infinite possibilities of challenging substances in a few patients.

Many studies have showed that patients who self-harm have a

protracted higher risk of death over years, both from suicide attempts and from natural causes [14,20-25]. Notably, Bergen et al [14] showed re- cently that diseases of the circulatory and digestive systems accounted for a large proportion of early death, which could be related to sociodemographic factors, unhealthy lifestyle, and neglected medical care both from the patients and from health workers. Whether abnormal test results at the time of a self-harm episode may help to detect patients with ongoing chronic premorbid or morbid state can be hypothesized. Unfortunately, we have no data about the long-term outcome of our pa- tients. This might be of interest to focus interventions to group at higher risk for future adverse outcome. However, it would require a complete re- thinking on how we care for patients who self-harm, integrating a com- prehensive somatic care, from the acute to the chronic phase.

Several limitations of this study should be taken into consideration. Some have been already cited: notably, impact of test results on medical prescription was not registered. Only French centers were studied, and this may prevent applicability of the results to other centers with differ- ent epidemiology of drug poisoning, notably in Developing countries [26]. Delay from drug ingestion to ED admission and dose supposed ingested were not registered, as often considered unreliable. Variation in delay from ingestion to blood sampling may impact the proportion of patients with abnormal test result. However, taking into consider- ation these data may have help to refine our analyses. Although more than 1000 may yet represent an important number of patients, it may

be still be too small to assess the vast amount of different drug-induced situations. Physicians may have changed their practice during the study period, being aware of being observed. To prevent this bias, we paid much attention not to put forward any hypothesis or advice regarding test ordering. Interestingly, the wide difference observed between cen- ters for routine test ordering likely indicates that conducting the study did not significantly modify variations in practice. Finally, no data on outcome after hospital discharge were reported, and event occurring after discharge may have been missed. Of note, investigators at every center were asked to check for readmission in the following week after patients discharge.

Conclusion

In conclusion, routine biological tests, although very frequently pre- scribed in drug overdose patients, very infrequently detected infraclinical vital abnormalities requiring emergent care. Rather, abnormal test results were frequent, but their relevance for acute care is difficult to establish. Age younger than 40 years, female sex, and type of drug ingested may se- lect a population with a low likelihood of abnormal test result. However, using these criteria may miss exceptional patients with potentially life- threatening abnormal results, precluding their straightforward use as a rule to not prescribe routine blood test analysis.

Appendix A. Supplementary data

Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.ajem.2016.04.002.

References

1. Spiller HA, Appana S, Brock GN. Epidemiological trends of suicide and attempted sui- cide by poisoning in the US: 2000-2008. Leg Med (Tokyo) 2010;12:177-83. http:// dx.doi.org/10.1016/j.legalmed.2010.04.005.
2. Lund C, Teige B, Drottning P, Stiksrud B, Rui T, Lyngra M, et al. A one-year observa- tional study of all hospitalized and fatal acute poisonings in Oslo: epidemiology, in- tention and follow-up. BMC Public Health 2012;12:858. http://dx.doi.org/10.1186/ 1471-2458-12-858.
3. Boelle PY, Flahault A. Suicide trends in France and UK. Lancet 1999;353:1364. http:// dx.doi.org/10.1016/S0140-6736(05)74363-4.
4. Watson WA, Litovitz TL, Rodgers Jr GC, Klein-Schwartz W, Reid N, Youniss J, et al. 2004 annual report of the American Association of Poison control centers Toxic Ex- posure Surveillance System. Am J Emerg Med 2005;23:589-666. http://dx.doi.org/ 10.1016/j.ajem.2005.05.001.
5. Henry JA, Cassidy SL. Membrane stabilising activity: a major cause of Fatal poisoning.

   Lancet 1986;1:1414-7.

   Love JN, Litovitz TL, Howell JM, Clancy C. Characterization of fatal beta blocker inges- tion: a review of the American Association of Poison Control Centers data from 1985 to 1995. J Toxicol Clin Toxicol 1997;35:353-9.

6. Love JN, Howell JM, Litovitz TL, Klein-Schwartz W. Acute beta blocker overdose: fac- tors associated with the development of cardiovascular morbidity. J Toxicol Clin Toxicol 2000;38:275-81.
7. Christenson J, Etherington J, Grafstein E, Innes G, Pennington S, Wanger K, et al. Early discharge of patients with presumed opioid overdose: development of a clinical pre- diction rule. Acad Emerg Med 2000;7:1110-8.
8. Bartoli M, Berny C, Danel V, Delahaye A, Desch G, Guitton J, et al. Recommendations for the prescription, implementation and interpretation of medical examinations in biology in the context of severe poisoning. Ann Biol Clin (Paris) 2012;70:431-50. http://dx.doi.org/10.1684/abc.2012.0726.
9. Nelson L, Lewin N, Howland MA. Principles of managing the acutely poisoned or overdosed patient. Goldfrank’s toxicologic emergencies. 9th ed. New York: Mc Graw Hill; 2010. p. 37-44.
10. Manini AF, Kumar A, Olsen D, Vlahov D, Hoffman RS. Utility of serum lactate to pre- dict drug-overdose fatality. Clin Toxicol (Phila) 2010;48:730-6. http://dx.doi.org/10. 3109/15563650.2010.504187.
11. Megarbane B, Deye N, Malissin I, Baud FJ. Usefulness of the serum lactate concentra- tion for predicting mortality in acute beta-blocker poisoning. Clin Toxicol (Phila) 2010;48:974-8. http://dx.doi.org/10.3109/15563650.2010.534483.
12. Chen VCH, Tan HKL, Cheng ATA, Chen C-Y, Liao L-R, Stewart R, et al. Non-fatal rep- etition of self-harm: population-based prospective cohort study in Taiwan. Br J Psy- chiatry 2010;196:31-5. http://dx.doi.org/10.1192/bjp.bp.109.067009.
13. Bergen H, Hawton K, Waters K, Ness J, Cooper J, Steeg S, et al. Premature death after self-harm: a multicentre cohort study. Lancet 2012;380:1568-74. http://dx.doi.org/ 10.1016/S0140-6736(12)61141-6.
14. Siest G, Henny J, Grasbeck R, Wilding P, Petitclerc C, Queralto JM, et al. The theory of reference values: an unfinished symphony. Clin Chem Lab Med 2013;51:47-64. http://dx.doi.org/10.1515/cclm-2012-0682.
15. Hawton K, Simkin S, Deeks J. Co-proxamol and suicide: a study of national mortality statistics and local non-fatal self poisonings. BMJ 2003;326:1006-8. http://dx.doi. org/10.1136/bmj.326.7397.1006.
16. Staikowsky F, Theil F, Mercadier P, Candella S, Benais JP. Change in profile of acute self drug-poisonings over a 10-year period. Hum Exp Toxicol 2004;23:507-11.
17. Hawton K, Bergen H, Simkin S, Wells C, Kapur N, Gunnell D. Six-year follow-up of impact of co-proxamol withdrawal in England and Wales on prescribing and deaths: time-series study. PLoS Med 2012;9, e1001213. http://dx.doi.org/10.1371/journal. pmed.1001213.
18. Heyerdahl F, Hovda KE, Bjornaas MA, Nore AK, Figueiredo JCP, Ekeberg O, et al. Pre- hospital treatment of acute poisonings in Oslo. BMC Emerg Med 2008;8:15. http:// dx.doi.org/10.1186/1471-227X-8-15.
19. Neeleman J. A continuum of premature death. Meta-analysis of competing mortality in the psychosocially vulnerable. Int J Epidemiol 2001;30:154-62.
20. Hawton K, Harriss L, Zahl D. Deaths from all causes in a long-term follow-up study of 11,583 deliberate self-harm patients. Psychol Med 2006;36:397-405. http://dx.doi. org/10.1017/S0033291705006914.
21. Karasouli E, Owens D, Abbott RL, Hurst KM, Dennis M. All-cause mortality after non- fatal self-poisoning: a cohort study. Soc Psychiatry Psychiatr Epidemiol 2011;46: 455-62. http://dx.doi.org/10.1007/s00127-010-0213-3.
22. Wahlbeck K, Westman J, Nordentoft M, Gissler M, Laursen TM. Outcomes of Nordic mental health systems: Life expectancy of patients with Mental disorders. Br J Psy- chiatry 2011;199:453-8. http://dx.doi.org/10.1192/bjp.bp.110.085100.
23. Reith DM, Whyte I, Carter G, McPherson M, Carter N. Risk factors for suicide and other deaths following hospital treated self-poisoning in Australia. Aust N Z J Psychi- atry 2004;38:520-5. http://dx.doi.org/10.1111/j.1440-1614.2004.01405.x.
24. Nordentoft M, Breum L, Munck LK, Nordestgaard AG, Hunding A, Laursen Bjaeldager PA. High mortality by natural and unnatural causes: a 10 year follow up study of patients ad- mitted to a poisoning treatment centre after suicide attempts. BMJ 1993;306:1637-41.
25. Eddleston M, Juszczak E, Buckley NA, Senarathna L, Mohamed F, Dissanayake W, et al. Multiple-dose activated charcoal in acute self-poisoning: a randomised controlled trial. Lancet 2008;371:579-87. http://dx.doi.org/10.1016/S0140-6736(08)60270-6.