Prolonged sedative and analgesic use in children gives rise to tolerance and withdrawal phenomena, which in turn have been related to a prolongation of mechanical ventilation and hospital stay.1 The literature offers sedation and analgesia guidelines applicable to both pediatric2 and adult Intensive Care,3 though no specific protocols are available in reference to the management of withdrawal syndrome (WS) in children. Most of the existing information refers to the use of opiates (OP) and benzodiazepines (BZD), as these are the most widely used drugs in the Pediatric Intensive Care Unit (PICU),4,5 and almost all the studies have been made in neonates or nursing infants–with very few studies in older children. Moreover, the published studies use different inclusion criteria and diagnostic methods, and the results are analyzed after applying different prevention or treatment protocols. As a result, very few data can be extrapolated to the general population. Based on the subjective impression that we had a high incidence of WS in our PICU, we decided to carry out a study with the primary objective of determining the true incidence of the problem in our Unit, and with the secondary objective of identifying risk factors for the development of WS in our patients.

A retrospective or historical, single cohort study was made, including all patients admitted to our PICU between January 2004 and July 2007, requiring mechanical ventilation and who received sedoanalgesia in continuous perfusion exclusively with midazolam (MDZ) and fentanyl (FENT) for at least 48h. Each patient was followed-up on throughout the duration of stay in the Unit. In these patients, on starting weaning from mechanical ventilation, the perfusion of FENT and MDZ was either abruptly suspended or decreased on a gradual basis. The appearance of WS was monitored based on clinical criteria, defining the syndrome as a clinical condition characterized by alterations of the central and/or autonomic nervous system, or gastrointestinal disorders6 (fundamentally tremor, restlessness, insomnia, tachypnea, fever, arterial hypertension, vomiting or diarrhea), correlated in time to the suspension of sedoanalgesia. Furthermore, we used the Finnegan scale7 as supporting diagnostic tool (Table 1) in all the patients, since it was the only such instrument available in pediatric practice during the period of the study. In this context, a score of 8 or more and of less than 12 was regarded as consistent with mild WS, while a score of 12–16 corresponded to moderate WS, and a score of over 16 indicated severe WS. The score obtained with this scale is always recorded in the case of patients in whom weaning from mechanical ventilation is started and prolonged perfusion of sedoanalgesia is suspended or reduced. In order to homogenize the results, we excluded all those patients administered some other sedative or analgesic in continuous perfusion additional to MDZ and FENT, as well as those administered replacement drug therapy for the prevention of WS. On the other hand, we also excluded patients who died or were moved to another PICU while receiving sedoanalgesia, along with those presenting disorders that could be confused with WS (since the diagnosis of WS was established on an exclusion basis): severe hemodynamic instability or respiratory failure, or serious neurological damage.

Clinical and demographic parameters were recorded such as age, gender, body weight, reason for admission and the Pediatric Risk of Mortality Score (PRISM), which expresses the probability of mortality as a percentage. We likewise registered the cumulative FENT and MDZ dose; maximum perfusion and the mean duration of continuous perfusion of both drugs up to the time of suspension or gradual reduction of perfusion; the joint administration or not of a muscle relaxant; the method of MDZ and FENT treatment withdrawal and the duration of the latter; the presence of WS and its severity as determined by the Finnegan scale7; and the treatment started and its duration. Since this was a historical cohort with no intervention of any kind, informed consent was not requested from the parents.

The SPSS version 19.0 statistical package was used for analysis of the results. The Student t-test, Mann–Whitney U-test and chi-squared test were applied to assess differences in the study variables between the two groups. Statistical significance was accepted for p<0.05. Logistic regression analysis was used to analyze the variables that best explained the appearance of WS. To this effect a model was developed in which the dependent variable was withdrawal syndrome (yes/no), and the independent variables were those parameters yielding p<0.15 in the univariate analysis and which did not constitute confounding factors. Logistic regression with the forward stepwise Wald method was carried out. Receiver operating characteristic (ROC) curves were plotted for this explanatory model to assess the predictive capacity of the variables referred to the development of the WS.

During the 42 months of the study a total of 620 patients were admitted to our Unit and were subjected to follow-up for the full duration of admission; of these subjects, 73 required sedoanalgesia in continuous perfusion during more than 48h. We excluded three patients transferred to another PICU, four who died while receiving sedoanalgesia, 7 administered other sedatives or analgesics in continuous perfusion (propofol, pentothal) in addition to MDZ and FENT, 6 in which replacement therapy was started with methadone and/or benzodiazepines, and 5 patients with severe neurological damage after weaning from ventilation. A total of 48 patients thus met all the inclusion criteria. The characteristics of these 48 patients are summarized in Table 2. The indications of sedoanalgesia were respiratory disease in 50% of the cases, traumatisms in 23%, septic shock in 10.4%, and others in 16.6%. Fifty percent developed WS, though this incidence increased to 80% on selecting the patients with perfusion for more than 5days. There were no significant differences between the patients who developed withdrawal (WS group) and those who did not (non-WS group) for any of the demographic variables (age, gender, weight). The data relating to the treatment received and the development of WS are summarized in Table 3. The mean duration of treatment and the cumulative dose of both FENT and MDZ, together with muscle relaxant utilization, were significantly higher in the WS group than in the patients without WS. However, after stratifying the data, muscle relaxant use was identified as a confounding factor, since the patients receiving such medication had also received higher cumulative doses of FENT and MDZ. However, no differences were observed between the two groups in terms of the maximum perfusion dose of both drugs. Of all the variables showing significant differences between the two groups, logistic regression analysis with the forward stepwise Wald method showed the duration of MDZ perfusion to offer the best explanation of the development of WS, with an odds ratio (OR) of 1.061 (95%CI 1.021–1.103). Analysis of the ROC curves (Table 4) of the explanatory model identified the values offering the best sensitivity and specificity, and which most likely would be found in the case of a diagnosis of WS, namely: a cumulative FENT dose of 0.48mg/kg, a duration of FENT or MDZ perfusion of 5.75days, and a cumulative MDZ dose of 40mg/kg. Brusque perfusion suspension was carried out in two patients with WS and in 7 without WS (no significant differences), while gradual suspension was carried out in 22 patients with WS and in 17 without WS. This gradual reduction in the non-WS group was always carried out simultaneously for the two drugs (mean duration of reduction 26.2±11.7h), while in the WS group the decrease in administration was carried out independently for the two drugs in most cases (decrease in MDZ: median 4days, range 2–12; decrease in FENT: median 5days, range 2–14). None of the patients showed discrepancies between the diagnosis based on clinical criteria and the diagnosis established on the basis of the Finnegan score. In this context, all the patients selected according to clinical criteria had a Finnegan score of 8 or higher. The majority developed mild (37.5%) or moderate WS (50%), and the syndrome proved severe in only 12.5%. Practically all of the patients developed clinical manifestations between 12 and 24h after suspending sedoanalgesia. Treatment consisted of methadone in 87.5% of the cases (median 8days, range 2–21), at a starting dose of 0.1mg/kg every 6h, though in some patients we had to increase the dose to 0.2–0.3mg/kg. Other drugs used were diazepam in 58.3% of the cases (median 7days, range 2–14), dipotassium clorazepate in 25%, and clonidine in 16.6%.

The present study confirms the high incidence of WS in children following the prolonged use of MDZ and FENT in continuous perfusion. The analysis was limited to patients receiving only MDZ and FENT in continuous perfusion, and who had not received previous preventive or replacement therapy for withdrawal syndrome (WS); accordingly, the only intervention in the study was the gradual reduction of perfusion. We consider this to be the best approach for determining the true incidence of WS with hardly no intervention, since it would not be ethical to brusquely suspend perfusion in all the cases, knowing that the use of OP and BZD in prolonged perfusion produces secondary WS beyond certain cumulative doses or perfusion times.1 In any case, the incidence found in our patients is high, but consistent with the limited data published to date. Bicudo et al.8 collected information on 36 children under two years of age administered MDZ and FENT for more than 24h, with a WS incidence of 50%, coinciding with our own data. Furthermore, in the same way as in our study, on selecting the patients subjected to perfusion for more than 5days, the incidence approached 100%, in concordance with the 86% incidence reported by Franck et al.9 in another study involving 15 children under 28 months of age receiving OP and BZD perfusion for over 5days. In the other published study offering data in this respect,10 a global incidence of 34% was reported–this percentage reaching 49% in the subgroup administered the highest drug doses. A point to be taken into account in the aforementioned studies is the fact that the age of the patients did not exceed two years in most cases, while in our series 24 children were under 22 months of age and the remaining 24 were between 3 and 13 years of age. This implies that our series is one of the few to contribute data on the incidence of WS due to MDZ and FENT in the PICU in pediatric patients not limited to neonates or nursing infants.

The symptoms of our patients were similar to those described in the literature, and although quantification was not possible due to the retrospective cohort design of the study, most of the patients suffered tremor, arterial hypertension, tachypnea, fever, perspiration, vomiting, diarrhea and mydriasis. In the study published by Bicudo et al.,8 100% of the subjects presented insomnia, tremor and tachypnea; 83% suffered fever; and a lesser percentage presented symptoms more characteristic of small children, such as excessive suctioning, nasal flutter or an exaggerated Moro reflex. All this coincides with the literature analysis published by Birchley,6 who concluded that the most frequent symptoms of WS due to OP and BZD in children are tremor, restlessness, irritability, insomnia, tachypnea, hypertension, fever, vomiting and diarrhea.

The mean cumulative FENT dose in the WS group was 0.98mg/kg, which is similar to that reported by other authors,8 while the cumulative MDZ dose was 70.57mg/kg in the WS group–this being higher than the dosage generally described in the few studies that have examined WS due to BZD on an isolated basis.11–13 There were no differences between the two groups in terms of the maximum perfusion of one drug or the other; this parameter therefore did not constitute a risk factor, in coincidence with the findings of most other investigators. The mean duration of perfusion was 13days in the WS group, versus 3–4days in the non-WS group. This likewise coincides with the data found in the literature. Only one patient who developed WS received perfusion for less than 5days, while 7 of the 24 patients who did not develop WS received perfusion for more than 5days. This observation is useful for establishing the ideal time for monitoring or preventing WS: if a preventive protocol is adopted for patients with more than 5days of perfusion, few will develop the syndrome without first having started replacement therapy–though this implies starting such therapy in some individuals who would not develop WS. On the other hand, if the appearance of WS is monitored in children with more than three days of perfusion, even when replacement therapy is not introduced from the start, we will be able to identify those patients who would develop WS despite having received perfusion for less than 5days–thereby allowing us to start symptomatic treatment as soon as possible. In fact, this is consistent with the data obtained from the analysis of the ROC curves, whereby a perfusion time of over 5.75days for both drugs significantly increases the risk of developing WS. To date, other authors have found that both the cumulative dose and the duration of perfusion of OP constitute risk factors for the development of WS. Specifically, in studies involving neonates,14–17 cumulative FENT doses of between 0.4 and 1.6mg/kg have been identified as risk factors. In older children, only one study18 has analyzed the possible risk factors, though it was limited to patients under two years of age. This study found a total FENT dose of 1.5mg/kg or a duration of perfusion in excess of 5days to imply a risk for the development of WS. Regarding BZD, there is only one previous study involving few patients,19 which has found a cumulative MDZ dose of 60mg/kg to constitute a risk factor, while in our series the MDZ dose correlated to increased risk was 40mg/kg. Of all the variables, logistic regression analysis found the MDZ perfusion time to be the parameter that best predicts the appearance of WS. In the analysis of the ROC curves, this variable yielded the largest area under the curve (AUC) and the best sensitivity and specificity performance. However, this result is not evaluable, due to the colinearity of the independent variables analyzed in the explanatory model. Accordingly, on considering the WS predictors separately, they lose predictive capacity in the multivariate model. Although there were differences between the WS group and the non-WS group in terms of the concomitant perfusion of a muscle relaxant, there were also significant differences in terms of the cumulative dosage of FENT and MDZ between those who had received muscle relaxant treatment and those who did not. This was also observed in a study of adult patients20 and in another study involving children,4 where for the same reason as mentioned above the authors were unable to establish risk for the development of WS.

A gradual decrease in perfusion was carried out in a larger number of patients who developed WS than in patients who did not develop the syndrome. Moreover, while in the non-WS group we always lowered MDZ and FENT at the same rate, in the WS group the decrease in each drug was carried out on an independent basis–thereby precluding statistical comparisons. Dose reduction was rapid in most cases: within 24–36h among the patients who did not develop WS, and in 4–5days in those who presented the syndrome. This protocol has also been used in most other studies.14,17,18,21 Although this practice is widely used for the prevention of WS,22 there is still no universally accepted protocol for the reduction of sedoanalgesia in children. In fact, no major differences have been found between the results of studies in which rapid dose reduction was carried out and those in which a slower regimen of up to 5–10days was preferred.9 Likewise, no gradual dose reduction protocol has been able to fully eradicate the development of WS.

The Finnegan scale7 was used to support the clinical diagnosis and to monitor WS, since it was the only instrument available during the period of the study. Most cases were mild or moderate. Although the scale was originally developed for application to the newborn offspring of women addicted to opiates, and has not been validated in older children, most authors have used the Finnegan scale for diagnostic and follow-up purposes. In recent years other tests specifically designed for older children have been proposed9,23,24–some of them still pending validation25–and have been adapted to the clinical characteristics of WS in non-neonate children. Such instruments therefore should be used for diagnosing and monitoring WS in children (Table 5), reserving the Finnegan scale7 for neonates and small nursing infants.

The treatment of the patients in our series was not preventive but symptomatic, once the syndrome had already developed. Methadone was used in most cases, with a median duration of 8days. In this context, the literature describes protocols lasting between 5 and 10days26–28 and several weeks.1 Most patients also received BZD via the oral route as symptomatic treatment. Although in recent years most protocols include clonidine for treatment in the context of WS, this drug was little used in our series–being reserved particularly for children with associated arterial hypertension. Recently, some studies have analyzed the role of clonidine in the management of WS, with good results.29–31

Our study has a number of limitations. On the one hand, the limited sample size can condition the validity of the results obtained. Moreover, the results can only be extrapolated to those patients who have exclusively received MDZ and FENT in continuous perfusion. In part because of this, and in view of the colinearity of the analyzed independent variables, the results obtained were invalidated following the logistic regression analysis. Thus, we were unable to analyze the effect of each drug separately upon the development of WS. Retrospective cohort studies allow us to control many of the limitations of case–control series (pure retrospective studies), since the information is obtained on a prospective basis, even though the posterior analysis has effect–cause directionality. Furthermore, another limitation of the study is the fact that several observers diagnosed and assessed the severity of WS–though it is true that use was made of the Finnegan scale,7 which shows agreement of up to 82% between different evaluators.32 Another limitation precisely corresponds to the fact of having used the Finnegan scale in support of the diagnosis, since this instrument has not been validated in children over 2–3 months of age, even though the diagnosis was fundamentally established on a clinical basis. The score obtained was always recorded in the clinical history of those patients in which the prolonged sedoanalgesia regimen was gradually withdrawn or suspended–a fact that eliminates possible selection bias referred to patients in the WS group or non-WS group. The variability in the decrease in perfusion may have influenced the number of patients who did or did not develop WS, or at least may have influenced the severity of the syndrome, though it must be emphasized that this was the only interventional measure adopted in the study. On the other hand, the fact that the diagnosis of WS was established on an exclusion basis may imply that the true incidence of the syndrome is higher or lower than the reported incidence. In this sense, we always considered the clinical manifestations in relation to reduction of the medication.

After analyzing the results and reviewing the literature, our Unit adopted a protocol for the prevention and treatment of WS following the perfusion of OP and BZD (Table 6). This protocol was subsequently modified and adopted by the Spanish Society of Pediatric Intensive Care (SECIP), and we hope to evaluate its usefulness in future studies.

In conclusion, it can be affirmed that the prolonged administration of OP and BZD is associated with the development of withdrawal syndrome. The appearance of this syndrome is related to high cumulative doses of both drugs, and to prolonged administration times. In patients who meet these criteria, monitoring is required in order to detect the syndrome, using an appropriate clinical scale and adopting preventive measures to avoid development of the syndrome or at least lessen its severity.

The authors declare that they have no conflicts of interest.