Among the main findings of the study is the estimate of the incidence of DRPs in NICU of 6.8 per 100 patient-days, a result not previously described in the literature. The main cause of DRPs involved the prescription of inappropriate doses, which translated into potential problems of therapeutic effectiveness and drug toxicity. Anti-infectives, especially gentamicin, were the drugs most involved in DRPs. Another important fact was the significant occurrence of DRPs with clinical relevance, with more than one third of newborns at great risk of significant damage. Also noteworthy was the high acceptability of interventions proposed by the pharmacist to the NICU physicians and nurses. Several methodology features give strength to our results, namely the cohort design, the prospective data collection, the large number of patients and the adoption of a standard DRP classification system. The PCNE classification system was chosen because of its clear hierarchical structure of problems and causes, as well as of its wide application in DRP research studies .
A very small number of studies have evaluated DRP in neonates, and none has been specifically designed for this population. In general pediatrics, we have found only three studies evaluating the frequency and nature of DRPs in hospitalized children. Two prospective cohort studies involving less than 120 neonates, one in Hong Kong and another in the United Kingdom and Saudi Arabia, found an overall prevalence of DRP of less than 50% [7, 8]. Both studies had a duration of 3 months and adopted the PCNE definition of DRPs. In these studies, DRPs were identified by review of medical charts and physician orders, but DRPs occurring during weekends were not considered. With the same DRP identification method, but using another classification of DRPs, Birarra et al.  found an overall prevalence of 30% in pediatric wards at a hospital in Ethiopia. The study was a cross-sectional study involving 285 children, but only 21 neonates, for 3 months. Although our work has adopted a method similar to the above studies, our prevalence was higher, with almost 60% of patients experiencing at least one DRP. Importantly, our study had a greater number of neonates as well as a longer recruitment period (3 years).
It should be noticed that in our study the DRPs occurred even though the NICU has an institutional clinical practice guideline that includes dosing guidelines for all drugs. Thus, one of the main reasons for the high occurrence of DRPs in NICUs is the physiological immaturity of neonates. Neonates have characteristics that change the pharmacokinetics of many drugs, with a significant impact on the pharmacotherapy. This population, unlike adults, has a low plasma protein concentration and a higher percentage of body water, in addition to decreased liver metabolism and renal clearance . These characteristics vary constantly along the growth and maturation of the neonate making it difficult to establish the adequate dose for each case. Consequently, the risk of either drug ineffectiveness or toxicity is always present in neonates [15, 20]. Another aspect of neonatal drug therapy is that medicinal formulations appropriate for this population are rare and, therefore, the dilution of medications for adult use is a common and necessary practice, a process that may also lead to subdoses as well as overdoses .
Accordingly, several studies in the pediatric population [7,8,9,10], as well as our study, have shown a predominance of DRPs with the potential for therapeutic ineffectiveness, mainly due to inappropriate dose selection. The main medicine that illustrates the difficulty in establishing optimal dosage schedules is gentamicin. This medicine is preferably distributed in aqueous compartments and is excreted unchanged almost exclusively by the kidneys [19, 20]. Because of these characteristics, neonates tend to have lower serum concentrations due to the progressive renal maturation and to the large body water volume during the first days of life. Hence, it is recommended that gentamicin dose be adjusted frequently as a function of postnatal life . In addition, the existence of multiple recommended dose regimens makes it difficult to prescribe gentamicin in neonatal practice [22, 23]. Such aspects explain why gentamicin was often involved in DRPs in our study, a finding consistent with other studies [8, 24,25,26].
Although less frequent than the problems of effectiveness, adverse reactions are also significant. In the first 72 h of life, the neonate may present a body weight reduction of more than 20% and the absence of dose adjustment of medicines contributes to a greater risk of toxicity . However, most adverse reactions in our cohort were only potential, with only 22 actual adverse drug reactions affecting 4% of newborns.
In our study, we estimate that nearly nine out of ten DRPs were preventable. Most pharmaceutical interventions were related to drug prescription and almost all were accepted by the NICU team, a result similar to that reported in other papers [9, 27]. Before proposing an intervention, the pharmacist should always consider the condition of the patient as well as the resources offered by the hospital and the health professionals. Thus, for an intervention to be adequate, the severity of each DRP must be equated. A study has shown that, compared to adults, pediatric patients are at a higher risk of severe DRPs, but published information on the actual risk of DRPs is limited . In our study, the safety-relevance analysis of DRPs showed that more than one third of neonates are exposed to DRPs with a considerable risk of causing moderate to severe injury, representing almost 40% of all identified DRPs. Using a different tool, Ibrahim et al.  and Rashed et al. [7, 8] observed that 30 to 50% of the DRPs were of moderate severity, but no severe DRPs were identified.
In addition to the performance of clinical pharmacists in the NICU, there are other strategies to reduce DRP as computerized physician order entry integrated with clinical decision support, barcode dispensing and administration system, reporting system of errors and adverse events and programs of training and continuing education . All those tools are in use at our NICU, except that the computerized physician order entry does not yet automatically check doses.
This study has some limitations. Firstly, the study was conducted in a single NICU, which may limit generalization of the results. However, the large majority of published studies on this and related topics were also single center studies. Secondly, the data were collected from secondary sources, including pharmacotherapy records, clinical charts, nursing records, physician orders and pharmacovigilance notifications, which might decrease data quality, but this was likely to have little impact on the study results because patient data was examined and evaluated prospectively and as was being recorded. Thirdly, the evaluation of the safety-relevance of the DRPs was made only by pharmacists with an unvalidated tool and supported by Neofax® 2011, which in part may have compromised this analysis. However, in the context of DRPs, we consider the relevance-safety analysis more adequate than just severity, because it combines the severity of a potential adverse event with its likelihood, offering a better measure of the potential risk to which the patient was exposed. Lastly, the therapeutic drug monitoring service is not a usual practice in our NICU, so it is possible that some DRPs have been underestimated.
Future research in this topic should preferably make efforts to include the evaluation of clinical outcomes related to DRP and to analyze the actual risk of DRP instead of the potential risk, as was the case in this study and in several other published studies. Studies on risk factors for DRP are also needed.