Dominique Vodovar

Drug Fever: a descriptive cohort study from the French national pharmacovigilance database.

BACKGROUND Although known as a rare adverse drug reaction (ADR), drug fever (DF) remains an important issue in medicine, with the risk of leading to inappropriate and potentially harmful diagnostic and therapeutic interventions. Only sparse data regarding DF have been published. OBJECTIVE The aim of the study was to investigate which drugs were associated with DF, and report outcomes. METHODS Cases of DF without skin reactions were selected from all ADRs reported from 1986 to 2007 in the French National Pharmacovigilance Database. Drugs potentially responsible for DF were assessed using a qualitative case-by-case analysis (Naranjos criteria) and quantitative measurement (proportional reporting ratio [PRR]). A drug was implicated as the cause of DF when the following criteria were validated: three or more cases and PRR of at least two with a Chi-squared value of at least four. RESULTS A total of 167 DF cases involving 115 drugs were eligible. Based on the PRR, 22 drugs were significantly associated with DF. Antibacterials represented the most frequently reported drugs, including amikacin (PRR 39.6 [95% CI 23.6, 69.0], oxacillin (9.1 [3.6, 23.4]), cefotaxime (5.5 [2.0, 15.3]), ceftriaxone (5.4 [2.6, 11.3]), rifampicin (4.0 [1.8, 9.2]), vancomycin (4.0 [1.4, 11.5]), ciprofloxacin (3.1 [1.2, 8.0]), isoniazid (3.9 [1.4, 11.4]), pristinamycin (3.1 [1.0, 9.1]) and cotrimoxazole (2.6 [1.2, 5.8]). Median time [interquartile range] from drug administration to fever onset was 2 days [1.0-10.5]. A diagnosis of DF was made following cessation of the suspected drugs (3 days [1.0-11.5] after fever onset. Drug rechallenge was performed (38.0%), resulting in recurrence of DF in all cases. DF resulted in life-threatening events (0.6%), hospitalization or prolonged hospital stay (24.5%) and persistent disability (0.6%). Final outcome was favourable in 96.9% of cases after drug discontinuation. CONCLUSION Diagnosing DF is challenging. Based on this large series, antibacterials remain the major class of drugs responsible for DF.

Lithium poisoning in the intensive care unit: predictive factors of severity and indications for extracorporeal toxin removal to improve outcome

Abstract Context: Lithium is responsible for life-threatening poisoning, not consistently improved by extracorporeal toxin removal (ECTR). Objective: Our aim was to identify predictive factors on admission of poisoning severity and based on an evaluation of practice, report indications for ECTR susceptible to improve outcome Methods: We performed a retrospective cohort study including all lithium-poisoned patients admitted to the ICU in a university hospital. The usual clinical, biological and toxicological variables were collected. Poisoning severity was defined by seizures, catecholamine infusion, mechanical ventilation >48 h and/or fatality. Univariate followed by multiple logistic regression analyses were performed to identify prognosticators of poisoning severity and ECTR use. Results: From 1992 to 2013, 128 lithium-poisoned patients including acutely (10%), acute-on-chronically (63%) and chronically poisoned patients (27%) were included. The presumed ingested dose of lithium was 17.0 g [8.0–24.5] (median [25th–75th percentiles]). Serum lithium concentrations were 2.6 mmol/l [1.5–4.6], 2.8 mmol/l [1.8–4.5] and 2.8 mmol/l [2.1–3.0] on admission, peaking at 3.6 mmol/l [2.6; 6.2], 4.3 mmol/l [2.4; 6.2] and 2.8 mmol/l [2.1; 3.1] in the three groups, respectively. Severe poisoning occurred in 48 patients (38%) including four fatalities. Using the regression analysis, predictive factors of poisoning severity were Glasgow coma score ≤10 (Odds ratio (OR), 11.1; 95% confidence interval (CI), [4.1–33.3], p < 0.0001) and lithium concentration ≥5.2 mmol/l (OR, 6.0; CI, [1.7–25.5], p = 0.005). Ninety-eight patients (77%) developed acute kidney injury according to KDIGO criteria and 22 (17%) were treated with ECTR. Peak lithium concentration ≥5.2 mmol/l (OR, 22.4; CI, [6.4–96.4]; p < 0.0001) and peak creatinine concentration ≥200 μmol/l (OR, 5.0; CI, [1.4–19.2]; p = 0.01) were associated with ECTR use. Only 21/46 patients who presented one of these two criteria were actually treated with ECTR. More significant neurological impairment persisted on discharge in patients not treated with ECTR (p = 0.0007) despite not significantly shorter length of ICU stay. Conclusions: Lithium poisoning is responsible for severe impairments but rare fatalities. Severity can be predicted on admission using Glasgow coma score and lithium concentration. Our results suggest that ECTR could be indicated if serum lithium ≥5.2 mmol/l or creatinine ≥200 μmol/l.

The optic nerve sheath diameter as a useful tool for early prediction of outcome after cardiac arrest: A prospective pilot study

INTRODUCTION Optic nerve sheath diameter (ONSD) measurement could detect increased intracranial pressure, and might predict outcome in post-cardiac arrest (CA) patients. We assessed the ability of bedside ONSD ultrasonographic measurement performed within day 1 after CA occurrence to predict in-hospital survival in patients treated with therapeutic hypothermia (TH). METHODS In two French ICUs, a prospective study included all consecutive patients with CA without traumatic or neurological etiology, successfully resuscitated and TH-treated. ONSD measurements were performed on day 1, 2, and 3 (ONSD1, 2, 3 respectively) after return of spontaneous circulation. All records were registered according to Utstein style. RESULTS ONSD1, 2, 3 were assessed in 36, 21, and 14 patients respectively. 19/36 patients (53%) were discharged alive from hospital, including 14/36 (39%) with favorable neurological outcome (Cerebral Performance Category [CPC] score 1-2). Survivors and non-survivors were similar regarding age, sex, cardiovascular risk factors, location and etiology of CA, simplified acute physiology score II, occurrence of post-CA shock, and clinical parameters collected during ONSD measurements. Median ONSD1 was significantly larger in non-survivors versus survivors (7.2mm [interquartile: 6.8-7.4] versus 6.5mm [interquartile: 6.0-6.8]; p=0.008). After adjustment on predictive factors, ONSD1 was significantly associated with in-hospital mortality (OR 6.3; 95%CI [1.05-40] per mm of ONSD1 above 5.5mm; p=0.03), and CPC score (OR for 1 point increase in CPC score: 3.2; 95%CI [1.2-9.4] per mm of ONSD1 above 5.5mm; p=0.03). ONSD1 was significantly correlated with brain edema assessed by the cerebrum gray matter attenuation to white matter attenuation ratio, measured by the brain computed tomography scan performed on admission in 20 patients (Spearman rho=-0.5, p=0.04). CONCLUSIONS ONSD seems a promising tool to early assess outcome in post-CA patients treated with TH.

Chloroquine poisoning-associated inverted Tako-tsubo cardiomyopathy.

Tako-tsubo cardiomyopathy, also called left apical ventricular ballooning syndrome , is a transient left ventricular (LV) dysfunction usually attributed to an acute stress situation, mimicking an acute coronary syndrome despite normal epicardial arteries. 1 LV shape in Tako-tsubo cardiomyopathy is described as looking like a Japanese fi shing pot used for trapping octopuses, with a round bottom and a narrow neck. However, variants have been reported with right ventricular, mid-ventricular, or basal involvement, which are referred to as inverted Tako-tsubo cardiomyopathy. To our knowledge, inverted in contrast to typical Tako-tsubo cardiomyopathy echocardiography patterns 2 were rarely reported in poisonings. 3

Extended-spectrum beta-lactamase producing Enterobacteriaceae in the intensive care unit: persistent issues to understand the transition from colonization to infection

We thank Denkel et al. for their interesting comments on our study [1]. The gut is the commensal site of Enterobacteriaceae, and as stated, rectal swabs are commonly used as indicators of patient colonization by extendedspectrum beta-lactamase producing Enterobacteriaceae (ESBL-PE). However, in vulnerable critically ill patients with invasive devices, colonization of sites by non-commensal bacteria is possible, as shown with vancomycinresistant Enterococcus [2]. In agreement with our findings, patients may develop ESBL-PE infections in the absence of previously documented ESBL-PE rectal colonization, as reported by Razazi et al. [3] in 1/110 patients and Goulenok et al. [4] in 162/671 patients. During our study period, the screening policy included systematic nasal and rectal swabs, as well as tracheal aspirations in intubated patients on ICU admission and weekly thereafter. ESBL-PE identification in any sample was considered as a marker of patient colonization as routinely done in the real-world and thus resulted in the maintenance of contact isolation preventive procedures until ICU discharge. The median time between the identification of an ESBL-PE-positive sample and the onset of ESBL-PE-related infection (7.5 days) clearly supported a prior colonization rather than an infection in its incubation period. Interestingly, only 2/5 asymptomatic patients with previous urine ESBL-PE colonization developed urinary tract infections. Denkel et al. underlined the elevated rate of ESBL-PE infections among our previously ESBL-PE-colonized patients in comparison to other published studies [3, 4] and to the ITS-KISS database of the German surveillance system of nosocomial infections on intensive care units [5] (Table 1). In our opinion, such a comparison should be cautious. Goulenok’s study [4] was conducted in a nonICU ward, in a pediatric and adult hospital. They excluded 162 patients without prior documented ESBL-PE colonization and only considered ESBL-PE Escherichia coli, while ESBL-PE Klebsiella pneumoniae remains a major cause of hospital-acquired infections. These two particularities of their study probably cause them to underestimate the exact prevalence of ESBL-PE infections in prior ESBLPE-colonized patients (7.8 %). In comparison to our findings, Razazi et al. also reported a significantly lower rate of ESBL-PE infections (15 %) among their previously ESBLPE-colonized ICU patients [3]. The significantly more marked severity of our patients in comparison to Razazi’s patients (median Simplified Physiology Score (SAPS) II: 58 vs. 35, respectively; SAPS II being a more adequate score to evaluate the severity of the acute disease than the MacCabe score which is aimed at evaluation of the patients’ underlying diseases) may have at least in part contributed to explain the observed difference in the ESBL-PE infection rate. Finally, extrapolation from the ITS-KISS database [5] should also be interpreted with caution, since this surveillance network included not only medical like our study but also surgical ICUs and both exact ESBL-PE colonization rate and patient severity scores are undetermined. In conclusion, we agree with Denkel et al. that internationally accepted definitions of the terms ‘‘colonization’’ D. Vodovar B. Megarbane (&) Department of Medical and Toxicological Critical Care, Lariboisiere Hospital, Paris, France e-mail: bruno-megarbane@wanadoo.fr

Do not Forget Gastrointestinal Decontamination in the Early Management of Lithium Poisoning

Dear Editor, Zamani et al. [1] recently reported an interesting case of lithium poisoning in a 25-year-old pregnant woman requiring emergent caesarean section due to foetal distress, however, resulting in the newborn’s death shortly after delivery. We agree with the authors that the gynaecologist’s recommendation to prematurely terminate the pregnancy at a gestational age of 28 weeks was probably not the optimal strategy to manage this severely lithium-poisoned patient presenting marked neurotoxicity with concomitant elevated serum lithium concentration. However, we would like to comment on the probable benefits of gastrointestinal decontamination to reduce lithium-induced maternal and foetal neurotoxicity that was not considered in this observation. On hospital presentation, the patient’s toxidrome combining deep consciousness impairment and muscle relaxant effects strongly suggested that neurological features were related to the sedative co-ingested drugs rather than to lithium. Interestingly, flumazenil administration did not induce seizures as expected if the observed neurological impairment was related to lithium. Moreover, its non-effectiveness to improve consciousness suggested a significant contribution of chlorpromazine in addition to clonazepam to the patient’s coma. Significant increase in serum lithium concentration from 2.1 to 4.95 mmol/L was observed during the first 6 hr after admission, strongly supporting persistent prolonged gastrointestinal absorption. In massive lithium ingestions, like in Zamani’s patient, especially if sustained-release tablets are involved, delayed absorption occurs resulting even sometimes in successive plasma lithium peaks [2]. Charcoal is ineffective to reduce lithium bioavailability [3], and when hospital admission is delayed over 1 hr after the ingestion, gastric lavage is not more recommended [4], while whole bowel irrigation (WBI) may still be useful as demonstrated in healthy volunteers [5] and suggested in lithium-poisoned patients with an observed consequent reduction in toxicity [6]. Despite the absence of controlled data, the current international guidelines recommend considering WBI for potentially toxic ingestions of sustained-release or enteric-coated drugs and drugs not adsorbed by activated charcoal like lithium [7]. Large amounts of an osmotically balanced polyethylene glycol (PEG) electrolyte solution should be administered in the absence of any contra-indications at a dosing schedule of 1500–2000 mL/h through a nasogastric tube in a half-seated position until rectal effluent is clear [7]. Interestingly, PEG, a FDA pregnancy category C drug, is widely prescribed as laxative in pregnant women and any potential risk during the third trimester seems unlikely despite the absence of adequate controlled studies [8]. Therefore, although usually being difficult to administer in lithium-poisoned patients experiencing nausea, vomiting and neurological impairments at risk of pulmonary aspiration, WBI was indicated and probably simple to consider in Zamani’s patient who was sedated and intubated. Another major issue highlighted in this observation is the absence of neurological improvement following haemodialysis, despite significant lithium elimination during the 1.5-hr session as supported by the decrease in serum lithium concentration from 4.95 to 1.2 mmol/L. While haemodialysis effectively eliminates lithium from the blood, its ability to remove lithium from other compartments including the brain and its contribution to improve the final outcome remains controversial, because only based on retrospective case series [2,9]. Thus, as previously recommended [10], determining the exact phase of lithium kinetics remains mandatory to allow implementing the best adapted management. Like in Zamani’s patient, increase in serum lithium concentration

Cortico-Amygdala-Striatal Activation by Modafinil/Flecainide Combination

Abstract Background Modafinil, a nonamphetaminic wake-promoting compound, is prescribed as first line therapy in narcolepsy, an invalidating disorder characterized by excessive daytime sleepiness and cataplexy. Although its mode of action remains incompletely known, recent studies indicated that modafinil modulates astroglial connexin-based gap junctional communication as administration of a low dose of flecainide, an astroglial connexin inhibitor, enhanced the wake-promoting and procognitive activity of modafinil in rodents and healthy volunteers. The aim of this study is to investigate changes in glucose cerebral metabolism in rodents, induced by the combination of modafinil+flecainide low dose (called THN102). Methods The impact of THN102 on brain glucose metabolism was noninvasively investigated using 18F-2-fluoro-2-deoxy-D-glucose Positron Emission Tomography imaging in Sprague-Dawley male rats. Animals were injected with vehicle, flecainide, modafinil, or THN102 and further injected with 18F-2-fluoro-2-deoxy-D-glucose followed by 60-minute Positron Emission Tomography acquisition. 18F-2-fluoro-2-deoxy-D-glucose Positron Emission Tomography images were coregistered to a rat brain template and normalized from the total brain Positron Emission Tomography signal. Voxel-to-voxel analysis was performed using SPM8 software. Comparison of brain glucose metabolism between groups was then performed. Results THN102 significantly increased regional brain glucose metabolism as it resulted in large clusters of 18F-2-fluoro-2-deoxy-D-glucose uptake localized in the cortex, striatum, and amygdala compared with control or drugs administered alone. These regions, highly involved in the regulation of sleep-wake cycle, emotions, and cognitive functions were hence quantitatively modulated by THN102. Conclusion Data presented here provide the first evidence of a regional brain activation induced by THN102, currently being tested in a phase II clinical trial in narcoleptic patients.

Prognosis and outcome of severe lithium poisoning

In the last issue of the journal, Ott et al. reported the clinical course and renal outcome of lithium intoxication in a cohort of chronically lithium-treated patients (Ott et al., 2016). They classified poisoning as moderate or severe based on serum lithium concentration using a 2.5 mmol/L cut-off point, while it is now well-established that severity cannot be exclusively related to lithium concentration on admission (Bailey and McGuigan, 2000). Neurotoxicity is mainly related to the brain lithium content that strongly depends on the pattern of lithium poisoning (acute vs. acute-on-therapeutic vs. chronic) as suggested by animal studies (Hanak et al., 2015). However, brain lithium concentration is accurately estimated by neither the serum nor the red blood cell lithium concentration (El Balkhi et al., 2009). Consistently, lithium-induced severe encephalopathy has been reported at therapeutic serum lithium levels (Bell et al., 1993). Therefore, we strongly believe that severity should be evaluated based on the onset of clinical complications, including repeated seizures, coma and cardiovascular failure requiring prolonged mechanical ventilation or resulting in fatality. Surprisingly, patients in this series developed none of these usual complications of lithium poisoning, despite a marked intensive care unit (ICU) admission rate (34.1%). Such complications occurred in 38% of the 128 lithiumpoisoned patients admitted to the ICU that we recently reported (Vodovar et al., 2016). Interestingly, we found that a serum lithium level ⩾5.2 mmol/L and a Glasgow Coma Score ⩽10 independently predicted severity (sensitivity of 0.64, specificity of 0.86, positive predictive value of 0.74 and negative predictive value of 0.80). The more elevated cut-off value to identify severity in our study in comparison to Ott et al.’s series (5.2 vs. 2.5 mmol/L) corresponded to the inclusion of a higher proportion of acute-on-therapeutically poisoned patients (63% vs. 31%) with higher lithium concentrations (percentage of patients with serum lithium level ⩾2.5 mmol/L: 73% vs. 42%). Using a 2.5 mmol/L threshold of serum lithium concentration to predict severity in our patients would have led to lower sensitivity (0.62), specificity (0.51), positive (0.43) and negative predictive values (0.69). Fatality is rare in lithium poisoning, while prolonged or persistent neurological symptoms represent a challenging issue. In our series of closely monitored and safely managed lithium poisonings, four (3%) patients died, and significant neurological disabilities persisted on ICU discharge (19%; Vodovar et al., 2016). Whereas the absence of fatalities in Ott et al.’s series may be consistent with previous data, the lack of a report on neurological status upon hospital discharge represents a clear limitation to the conclusions drawn on lithium-poisoning outcome. Although, similarly to Ott et al., we reported more frequent tremor, drowsiness, stupor and extrapyramidal symptoms in the acute-on-therapeutically versus the chronically lithium-poisoned patients, we demonstrated that chronic poisoning more frequently resulted in neurological impairments on ICU discharge, despite a longer stay and lower serum lithium peaks. Such differences in recovery could be explained in the chronic poisoning by the markedly delayed diagnosis, the higher rate of older patients with underlying brain vulnerability and the more prolonged lithium elimination half-life leading probably to enhanced brain exposure. Thus, given the high rate of chronic poisonings in Ott et al.’s study (69%), the lack of neurological recovery investigation on patients’ discharge does not allow outcomes to be assessed properly. To explain the discrepancy between the suggested poisoning severity and the reported outcome, biases could be hypothesised in relation to data collected from a retrospective cohort study not designed specifically to investigate lithium poisoning and thus clearly underestimating its actual consequences. Furthermore, despite an initial increase in serum creatinine, Ott et al. stated that a lithium overdose has no final impact on renal function one month after intoxication (Ott et al., 2016). Similar to these authors, we found that acute kidney injury (AKI) occurred in 77% of our lithium-poisoned patients, more frequently in the chronically poisoned patients (Vodovar et al., 2016). AKI appeared reversible, although the long-term impact of lithium poisoning on kidney function remained unknown. Interestingly, Ott et al. stated that the decline in glomerular filtration rate was inversely correlated with the maximal lithium concentration. Although mathematically exact, this observation should be limited to the chronically poisoned patients with AKI representing the main cause of poisoning, often related to sepsis or prescription drugs interacting with the renin-aldosterone-angiotensin system. In contrast, in acute-on-therapeutic poisoned patients, a decline in glomerular filtration rate is expected to be parallel to the peak lithium concentration, since AKI is the consequence of lithium toxicity, mainly in relation to water loss induced by diarrhoea, vomiting and cardiovascular failure, although not clearly described in the present study. Additionally, as acknowledged by the authors, lower thresholds were used to indicate dialysis than those recommended by the Extracorporeal Treatments in Poisoning Workgroup, that is, a serum lithium level >5 mmol/L or >4 mmol/L in the presence of renal impairment, as long as lifethreatening symptoms are absent (Decker et al., 2015). In our ICU lithium-poisoned patients, we showed that peak serum lithium Prognosis and outcome of severe lithium poisoning

Defining predictive factors of severity and indications for extracorporeal toxin removal in lithium poisoning: not an easy objective!

We thank Villeneuve and colleagues for their comments on the methodological limitations of our study. We enrolled intensive care unit (ICU) patients whose admission was decided by the physicians in charge.[1] We acknowledged that our patients did not represent the larger lithium overdose spectrum. However, since close monitoring during the first hours is recommended,[2] ICU admission at least in acute ingestions of significant lithium doses is usual. Interestingly, 38% of our patients developed severe complications, supporting the high risk of worsening in lithium poisoning. We agree that using Hansen & Amdisen Scale to categorize all ingestions is questionable and therefore defined severity based on unquestionable complications, i.e. seizure onset, catecholamine infusion, mechanical ventilation lasting >48h and fatality. Based on a logistic regression analysis, Glasgow Coma Score 10 and lithium concentration 5.2mmol/L predicted poisoning severity on admission.[1] Decision for extracorporeal toxin removal (ECTR) was based on clinical severity and plasma lithium concentrations. Although imperfectly reflecting the glomerular filtration rate, serum creatinine appears to be the most relevant parameter in practice to quantify the decrease in lithium elimination. Kidney Disease Improving Global Outcomes (KDIGO) criteria are the standards used to define acute kidney injury[3]; however, they require (1) to know the patient’s past creatinine concentration, which is rarely available on admission and (2) to monitor urine output which may be difficult in the presence of encephalopathy. Therefore, recognizing that KDIGO stages are inappropriate at the bedside, the EXTRIP workgroup preferred to use theoretical values of serum creatinine to define the patients corresponding to KDIGO stages 2 and 3.[4] We agree that our threshold of 5.2mmol/L indicating ECTR is only valid for acute and acute-on-chronic poisonings since the peak lithium concentrations in our chronic poisonings were consistently <4mmol/L.[1] Due to the low occurrence of ECTR use (22 patients), we were unable to compare the impact on outcome of the time of ECTR initiation and the choice of ECTR technique (i.e., intermittent hemodialysis vs. continuous veno-venous hemodiafiltration or both techniques successively used). Neurological recovery on discharge is the most interesting parameter to assess lithium-poisoned patient’s outcome. In our series, 24 lithium-poisoned patients presented persistent neurologic impairment on ICU discharge including confusion, dysarthria, hypertonia, myoclonus and ataxia despite undetectable serum lithium.[1] Their previous neurological status was determined based on their past medical history collected from family, psychiatrist and general practitioner. Since no follow-up after ICU discharge was performed, the final neurological outcome and the proportion of patients who developed the syndrome of irreversible lithium-effectuated neurotoxicity were undetermined. Interestingly, chronically lithium-poisoned patients more frequently developed persistent neurological impairment on ICU discharge (p< 0.02), supporting the hypothesis that prolonged supratherapeutic lithium concentrations due to limited lithium elimination may lead to enhanced brain exposure and thus poorer final outcome.[5] In conclusion, we agree that our retrospective observational study has several limitations. Only randomized controlled trials can definitively clarify whether ECTR is useful in preventing and reversing lithium-related toxicity, which patients should benefit from ECTR, when ECTR should be started and which ECTR technique is the most relevant. However, due to the rarity of lithium poisoning, an international collaboration-based registry may represent the first step to confirm our study findings.