Next Article in Journal
DoubleSG-DTA: Deep Learning for Drug Discovery: Case Study on the Non-Small Cell Lung Cancer with EGFRT790M Mutation
Next Article in Special Issue
Overview of Side-Effects of Antibacterial Fluoroquinolones: New Drugs versus Old Drugs, a Step Forward in the Safety Profile?
Previous Article in Journal
Inhibition of Hepatitis B Virus (HBV) by Tachyplesin, a Marine Antimicrobial Cell-Penetrating Peptide
Previous Article in Special Issue
A Volumetric Absorptive Microsampling UPLC-MS/MS Method for Simultaneous Quantification of Tacrolimus, Mycophenolic Acid and Creatinine in Whole Blood of Renal Transplant Recipients
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Pharmaceutics
Volume 15
Issue 2
10.3390/pharmaceutics15020673
pharmaceutics-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Automated Interlaboratory Comparison of Therapeutic Drug Monitoring Data and Its Use for Evaluation of Published Therapeutic Reference Ranges

by
Jens Borggaard Larsen
1,*,
Elke Hoffmann-Lücke
2,3,*,
Per Hersom Aaslo
4,
Niklas Rye Jørgensen
5,6 and
Eva Greibe
2,3,*
1
The Datasupport Centre for Personalised Medicine, Region South, DK 5000 Odense, Denmark
2
Department of Clinical Biochemistry, Aarhus University Hospital, DK 8200 Aarhus, Denmark
3
Institute for Clinical Medicine, Health, Aarhus University, DK 8000 Aarhus, Denmark
4
The Laboratory of the Danish Epilepsy Centre–Filadelfia, DK 4293 Dianalund, Denmark
5
Department of Clinical Biochemistry, Rigshospitalet, DK 2600 Glostrup, Denmark
6
Department of Clinical Medicine, Faculty of Health and Medical Sciences, University of Copenhagen, DK 1165 Copenhagen, Denmark
*
Authors to whom correspondence should be addressed.
Pharmaceutics 2023, 15(2), 673; https://doi.org/10.3390/pharmaceutics15020673
Submission received: 11 January 2023 / Revised: 14 February 2023 / Accepted: 14 February 2023 / Published: 16 February 2023
(This article belongs to the Special Issue Therapeutic Drug Monitoring)
Browse Figures
Review Reports Versions Notes

Abstract

:
Therapeutic drug monitoring is a tool for optimising the pharmacological treatment of diseases where the therapeutic effect is difficult to measure or monitor. Therapeutic reference ranges and dose-effect relation are the main requirements for this drug titration tool. Defining and updating therapeutic reference ranges are difficult, and there is no standardised method for the calculation and clinical qualification of these. The study presents a basic model for validating and selecting routine laboratory data. The programmed algorithm was applied on data sets of antidepressants and antipsychotics from three public hospitals in Denmark. Therapeutic analytical ranges were compared with the published therapeutic reference ranges by the Arbeitsgemeinschaft für Neuropsychopharmakologie und Pharmakopsychiatrie (AGNP) and in additional literature. For most of the drugs, the calculated therapeutic analytical ranges showed good concordance between the laboratories and to published therapeutic reference ranges. The exceptions were flupentixol, haloperidol, paroxetine, perphenazine, and venlafaxine + o-desmethyl-venlafaxine (total plasma concentration), where the range was considerably higher for the laboratory data, while the calculated range of desipramine, sertraline, ziprasidone, and zuclopenthixol was considerably lower. In most cases, we identified additional literature supporting our data, highlighting the need of a critical re-examination of current therapeutic reference ranges in Denmark. An automated approach can aid in the evaluation of current and future therapeutic reference ranges by providing additional information based on big data from multiple laboratories.

1. Introduction

Therapeutic drug monitoring (TDM) is a clinical tool where the concentration of a pharmaceutical drug in a biological matrix is used to optimise and individualise the treatment of a patient. TDM is applicable for drugs having a narrow therapeutic range; a correlation between the blood concentration and the therapeutic effect or adverse reactions; and where the symptoms of the disease are difficult to monitor or quantify [1,2,3]. This is particularly true for the treatment of many neurological and psychiatric disorders, but TDM is also used for monitoring immunosuppressive drugs and antibiotics [4,5].
An important parameter for the successful clinical application of TDM is the availability of an accurate therapeutic reference range. Therapeutic drug monitoring can be applied for optimising the dosage and hereby treatment with a drug, using two strategies [2,4,5,6]. If a well-defined therapeutic reference range is available, dose titration can be performed until the patient is within this range. During treatment, any relapse of the disease or acquired adverse reactions should trigger a new measurement and adjustment of dosage accordingly, within the range [3].
As an alternative to the use of a common therapeutic reference range, a patient can be used as his or her own reference [6]. In this individual approach, the dosing of the patient is titrated until the clinician judges the treatment to be optimal, with a minimum of adverse effects. A measurement of the plasma concentration taken at this point in time will serve as the patient’s future reference and used as a target for dose adjustments should changes in the treatment arise. Although, a therapeutic reference range using this method is not strictly necessary, it still holds considerable value as a guidance of the therapy and allows the optimal dosing of the patient to be related to a toxic alarm limit [6].
The therapeutic reference range is defined by a lower concentration of the drug where the treatment is starting to take effect and an upper where the increase in the therapeutic effect declines, and there is an increased chance of the patients’ acquiring adverse drug reactions. The reference range should therefore mirror the dosage span, containing the optimal therapeutic effect for most of the patients. As in the case of endogen substances, the plasma concentration of a drug varies considerably, even when adjusting for dosage. This is primarily due to environmental, physiological, and genetic factors. However, compared with an endogen substance, there are important differences when defining the reference range for a therapeutic drug [7]:
  • The plasma level of a TDM drug is dosage dependent. The lowest measurement is zero (or the limit of quantification of the laboratory analysis), while the upper depends on the prescribed dosage.
  • Defining a reference range for an endogenous substance is carried out to identify the variance in the healthy population. In contrast, an optimal therapeutic reference range should be based on the variation in a sick but, in regard to the therapeutic effect of the drug, well-treated cohort.
  • The same drug may be used for the treatment of different diseases, having different biochemical origins, symptoms, and therapeutic ranges. This is particularly true for psychopharmacologic treatments.
The calculation of a therapeutic reference range can further be made challenging because of a limited population size [7]. This is especially true when it comes to the treatment of rare diseases or to not commonly prescribed drugs. Because of this, the published therapeutic reference ranges of psychoactive drugs are often based on a limited number of specifically selected patients, which does not necessarily reflect the variations in the real population [6]. Small population sizes and large individual variation may also explain a lack of correlation between the blood concentration and therapeutic effect found in some studies [3].
An alternative to a controlled population, when defining the therapeutic reference range of a drug, is the use of retrospective data from laboratory information management systems (LIMS) [8,9,10]. The benefit of this is a larger data set and the capability of being able to relate drug plasma concentration with adverse effects, e.g., neutropenia or liver function. However, normally, it comes with no information on dosage, the clinical reasoning for ordering the measurement, or the therapeutic effect of the treatment. Laboratory data sets also commonly hold numerous results from patients that are excluded in a controlled setup.
The Arbeitsgemeinschaft für Neuropsychopharmakologie und Pharmakopsychiatrie (AGNP) published the “Consensus Guidelines for Therapeutic Drug Monitoring in Neuropsychopharmacology” [3], which has become the primary reference within the field of TDM. This work evaluates indications for the use of drug monitoring for psychiatric drugs and presents therapeutic reference ranges that are based on the current literature. The guideline is an impressive and important contribution that lists more than 1300 references, and it is currently in its third edition. However, one point of concern regarding this work is the validity of some of the older citations used to support the ranges herein and the transparency of the calculation of those that have been concatenated from several studies [11]. Although the variation in the precision between the laboratories has significantly improved over the past 20 years, most published ranges date back to a few early studies performed when the drug was first introduced to the market. Commonly, these studies relied on a single laboratory test developed in-house, and verification using modern techniques is therefore necessary. The continual re-evaluation of therapeutic reference ranges is also important as the treatment paradigm of a drug may change over time [12]. For these reasons, a ratification is warranted before a published range can be used in the clinic, and the agreement with current laboratory methods needs to be ensured.
In this study, we present a model for sorting and selecting routine TDM data from laboratory LIMS systems. The algorithm of the model has been integrated into the software RefIT, which facilitates the fast and easy batch calculation of analytical ranges that are based on percentiles of the sorted data sets. The software was applied on data sets from a TDM analysis of antidepressants and antipsychotics, from three public hospitals in Denmark. The results allow a comparison between the actual laboratory findings and the therapeutic reference ranges published in the AGNP consensus guideline [3].

2. Materials and Methods

2.1. Model Description

To avoid confusion between the therapeutic reference ranges, which have been clinically validated, and the ones calculated on the basis of retrospective laboratory data in this study, throughout the manuscript, we use the term “therapeutic analytical range” for the latter. This, because we find it a more correct term than “concentration in blood”, used by Hiemke et al. [3].
Laboratory LIMS data may hold several subpopulations that are otherwise omitted when calculating therapeutic reference ranges on the basis of controlled experimental setups. These subpopulations are not outliers in a laboratory sense. Rather, they are samples or patients that would normally be excluded before performing the calculation. The individual subpopulations may consist of the following:
  • Patients that have not been medication fasting prior to sampling and therefore are not at their minimum at steady state.
  • Patients that have not reached steady state at the time of the sampling.
  • Samples taken during adjustment of dosage and where the patient is not optimally treated.
  • Samples from patients not taking the medicine as prescribed (noncompliance).
  • Patients that are abusing the medicine.
  • Samples from patients with a high degree of comedication or who are taking the medicine in combination with general drug abuse.
  • Patients receiving a standard dosage but who are pharmacogenetically poor or ultrafast metabolisers of the drug.
  • Patients that are misdiagnosed and thus cannot be optimally treated.
As there is no information on the health status or therapeutic effect of a patient at the sampling time in the LIMS system, there is no direct way of validating results for calculating a therapeutic analytical range. One option is to include all samples. However, this would include all the subpopulations in the data set. In contrast, selection between the results from a patient can be performed by either taking the first or last of the measurement. These methods all weight each of the patients equally: both patients that never obtain optimal treatment and stable patients in long-term treatment.
The presented model for evaluating and excluding such subpopulations from retrospective laboratory data sets relies on the premise that the TDM measurements are requested by trained clinicians, following common practice for ordering this type of analysis (Figure 1). Although it provides a very simplified version of the clinical decision-making, it facilitates the development of an algorithm capable of objectively sorting and selecting large numbers of data. A flowchart of the model is shown in Figure 1. With no information in the LIMS data on why a sample has been requested, the selective parameter of the model is set as the time interval between sequential measurements. The main assumption is that TDM is requested to check the patient for compliance or for optimising the therapeutic treatment using the common or individual range approach, described in the introduction of this paper. For all these three scenarios, it is anticipated that any discrepancy in the result will be followed by clinical intervention and an additional measurement from the patient. A single measurement will in this regard either suggest a check for compliance, a sample taken as reference, or fit within a published therapeutic reference range. If there is no immediate additional sequential sample from the same patient, this would suggest that the result was approved and thus can be included in the data set for calculating the therapeutic analytical range. However, if the result is followed shortly after by a second measurement, this would indicate that the first is obsolete and should be excluded from the analysis (Figure 1).
Alternatively, if the time between two samples is long (months), this suggests that the patient was optimally treated during the interval. The second measurement may be requested either on the background of a routine consultancy or because of a change in the treatment. In the first case, this second result is included to ensure that optimally treated patients are weighted higher in the data set when calculating the therapeutic analytical range. The latter would be considered as a new treatment and be included depending on the presence of further sequential samples (Figure 1).

2.2. Data Collection

Retrospective results from TDM analysis of antidepressants and antipsychotics were extracted from the LIMS system of the three participating Danish laboratories. These were the Department of Clinical Biochemistry, Aarhus University Hospital, Aarhus (AUH); the Laboratory of the Danish Epilepsy Centre–Filadelfia, Dianalund (EHL); and the Department of Clinical Biochemistry, Rigshospitalet (RH), Copenhagen. Approvals were obtained from the respective hospital boards, prior to downloading the data.
Therapeutic drug monitoring at AUH is carried out by using LC-MS/MS technology and assays that have been developed in-house. The analyses are all accredited according to ISO:15189:2013, and the quality is externally monitored by proficiency testing. Results were extracted for the following drugs covering a period from 1 January 2014 to 31 December 2018: amitriptyline/nortriptyline (metabolite), aripiprazole/dehydroaripiprazole (metabolite), citalopram/escitalopram, clomipramine/desmethylclomipramine (metabolite), clozapine, duloxetine, fluoxetine/norfluoxetine (metabolite), imipramine/desipramine (metabolite), mirtazapine, olanzapine, perphenazine, quetiapine, risperidone/paliperidone (metabolite), sertraline, venlafaxine/o-desmethyl-venlafaxine (metabolite), ziprasidone, and zuclopenthixol.
All the assays performed at EHL have been developed in-house by using LC-MS/MS technology. These include the same drugs as are analysed at AUH and in addition flupentixol, haloperidol, and paroxetine. Although part of the laboratory production, fluoxetine/norfluoxetine, mianserine, and ziprasidone were excluded, owing to a low number of samples (<100). The assays (n = 26) are accredited according to ISO15189:2013, with the exception of aripiprazole/dehydroaripiprazole, and mirtazapine. The quality is externally monitored by proficiency-testing programmes, covering all analytes. For this study, data were collected from the laboratory LIMS system spanning a period from 1 January 2012 to 30 March 2022.
The analyses for therapeutic drugs at RH is performed by HPLC using UV-detection. The following drugs are included in the laboratory repertoire: amitriptyline/nortriptyline, clomipramine/desmethylclomipramine, clozapine, dosulipine/northiaden, and imipramine/desipramine, of which amitriptyline, nortriptyline, clozapine, and imipramine/desipramine are accredited according to ISO 15189:2013. The quality of the assays is monitored by external proficiency-testing schemes. For the calculations, data were collected covering the period from 9 May 2011 to 26 April 2022.

2.3. Data Analysis and Evaluation of the Model

The model for sorting and selecting the retrospective data sets has been incorporated into the software RefIT (available from GitHub under a general public license (GPL)—download address: https://github.com/JensLarsen/RefIT/releases/tag/V1.0). For evaluation purposes, RefIT has additional functions for calculating a therapeutic analytical range using one data point per patient (either the first or the last entry) and by including all samples in the data set. Percentiles are set in the software, which also has the selection of the different mathematical methods for calculating them. Other built-in features are the selection of sex and age interval, the removal of outliers on the basis of Tukey’s fences, and setting the minimum time for including two sequential samples in the TDM model. The software takes data input from Excel files and provides data export for the results to the same file format for easy validation. In addition, percentile and normal distribution graphs can be exported as image files.
The TDM model for selecting data and calculating therapeutic analytical ranges was evaluated by comparing the four models of data selection in the software. This was carried out on a single data set from AUH, using data from the drugs clozapine, perphenazine, and imipramine, covering high (>2000), mid (>500), and low (<200) numbers of data points in the data set.

2.4. Calculation of Therapeutic Analytical Ranges

The RefIT software was used on data sets obtained from the LIMS systems of the participating laboratories. These are situated in three of the five regions of Denmark, and uses two different LIMS software. At AUH and RH, Labka II is used (CSC Denmark A/S, Copenhagen, Denmark), while EHL uses LIMS software BCC (CGI Inc., Ballerup, Denmark). Both LIMSs facilitate the export of data as Excel files that can be imported directly into RefIT.
Data were selected by setting the period for routine consultancy in the model to 7 months. All calculations were performed while not removing outliers (Tukey’s fences deselected in the RefIT software). To increase the robustness of the ranges, a limit of >100 samples was set for calculating them. Therapeutic analytical ranges were calculated using the selected data from each of the three laboratories, as 10–90 and 25–75 percentiles. General statistics, including the total number of samples and samples included in the analysis, are provided as Supplementary Data files to this paper.
For each analyte, a combined and a sex-specific therapeutic analytical range was calculated. To investigate the effect of age, the intervals 20–64 years and 65–100 years were selected. Age- and sex-specific differences were calculated on the basis of the data set supplied by AUH.
The lower and upper percentile limits from each compound were compared between the individual laboratories. This was done by allowing a maximum bias of 30% (+/−15%) compared with the average value between the laboratories.

3. Results

In total, 199,964 measurements of 31 antidepressant and antipsychotic drugs were collected from the LIMSs of the three participating laboratories. Applying the TDM model for the selection of data led to the inclusion of 111,300 data points from 79,119 patients (see Supplementary Data files). As the repertoire differs between the laboratories, not all the analytes are represented in each data set.

3.1. Evaluation of the TDM Model

The TDM model for selecting data (Figure 1) was evaluated by comparing the calculated therapeutic analytical range with the ranges obtained by including all samples in the data set and by using one point per patient, defined as either the first or the last sample in the series. Figure 2 shows the complete data set of clozapine from AUH, and Figure 3 shows the four calculated percentile plots. Table 1 shows the therapeutic analytical ranges of clozapine, perphenazine, and imipramine obtained using all of the four methods. Including all the data points resulted in a much wider therapeutic analytical range for clozapine and imipramine, while using only the first data point from each patient resulted in a lower calculated range. Including only the last entry from each patient resulted in a calculated range closer to that of the TDM model, but this was based on a lower number of samples.

3.2. Interlaboratory Comparison

In general, there was good concordance between the calculated therapeutic analytical ranges of the participating laboratories (Table 2). The lower 10% percentile limit deviated more than was the case of the upper 90% level. For three of the analytes, there was disagreement (>30% bias) in regard to the upper 90% limit. These were citalopram/escitalopram (same LC-MS/MS analysis as escitalopram is the s-enantiomer of citalopram), quetiapine, and mirtazapine. Here, the data sets from EHL showed a higher calculated therapeutic analytical range than that of AUH.

3.3. Comparison to Published Therapeutic Reference Ranges and Ranges in the AGNP Consensus Guideline

The calculated therapeutic analytical ranges were compared with the reference ranges listed in the “Consensus Guidelines for Therapeutic Drug Monitoring in Neuropsychopharmacology” and the citations given here in [3,13,14,15,16,17,18]. As many of the ranges in the guideline are concatenated from multiple studies, with no additional information on the primary references and the calculation method, both 10–90 (Table 2) and 25–75 (Table 3) percentiles were calculated for each drug. For most of the ranges, there were a better correlation when the analytical range was calculated as a 10–90 percentile (Table 2). Although some variations existed between the laboratories, the 10% lower limit of the calculated therapeutic analytical ranges was, for most analytes, below those reported in the AGNP guideline. This was true for all except the drugs flupentixol, haloperidol, perphenazine, sertraline, and venlafaxine + o-desmethyl-venlafaxine (total) (Table 2). The upper 90% limit calculated for these compounds was also significantly higher than the published therapeutic reference ranges of the AGNP.
For perphenazine, the reported range of the AGNP consensus guideline is 1.5–6.0 nmol/L, with the main reference being a review by Putten et al. [19]. This review cites two studies from the early 1980s, defining a lower limit of therapeutic effect at 1.5 nmol/L, with adverse effects observed at a plasma concentration of >3 nmol/L. The therapeutic analytical ranges calculated from multiple laboratories in our study indicate a higher range, similar to that published by Kistrup et al. at 1.8–18 nmol/L (Table 2) [20].
The published therapeutic reference range for the antipsychotic drug haloperidol is from 2.7 to 26.6 nmol/L [3]. However, several studies not cited in the AGNP guideline report a larger range and a higher upper limit: Morselli et al. 2.66–39.9 nmol/L [21]; 5.3–63.8 nmol/L [22]; 26.6–66.5 nmol/L [23,24]; and Palao et al. 14.6–38 nmol/L [25]; 14.9–44.9 nmol/L [26]; 13.3–45 nmol/L [27]. In addition, Putten et al. cites four other studies [19]. The therapeutic analytical range from 5 to 48 nmol/L that was calculated in our study supports these other findings (Table 2).
The AGNP cites four references for the presented therapeutic reference range for flupentixol, denoting it to be from 1.2 to 11.5 nmol/L [3]. The main citation is a study from 1985 by Balant-Gorgia et al., who calculated a threshold for antipsychotic response to 4.7 nmol/L [28]. A more recent study of those cited was by Roman et al. 2008, who presented a range from 2.3 to 34.5 nmol/L [29]. This was similar to the one defined by Kistrup et al. from 1.2 to 37 nmol/L [20]. Together with the therapeutic analytical range from 2 to 23 nmol/L calculated here (Table 2), these studies indicate that the upper limit of flupentixol should be higher than that reported by AGNP.
The published therapeutic reference range for paroxetine is reported to be from 61 to 198 nmol/L [3]. This range seems to be concatenated mainly from two studies: one by Gex Fabry et al., indicating an optimal plasma level from 21 to 198 nmol/L, and one by Tomita et al., reporting a range from 61 to 182 nmol/L [30,31]. The therapeutic analytical range calculated in the current study is considerably higher, ranging from 61 to 537 nmol/L (Table 2). However, this result stems from a single laboratory and has been calculated on the basis of a relatively low number of samples (n = 324). However, Reis et al., in a similar population-based study, reported a comparably high range for paroxetine: 29–433 nmol/L [8].
The AGNP guideline cites 12 studies for the therapeutic reference range for venlafaxine and its active metabolite o-desmethyl-venlafaxine [3]. The total range for the combined substances is given as 361–1520 nmol/L. Although the exact origin for this range is not indicated, this is very close to the upper limit reported by Shams et al., 685–1534 nmol/L, and Veefkind et al., 722–1482 nmol/L, which are two of the cited papers [32,33]. However, Shams et al. used 25–75 percentiles, which may be the primary reason why the upper limit is considerably lower than the therapeutic analytical ranges in our study, which is on average from 550 to 2555 nmol/L when applying 10–90 percentiles (Table 2). A higher upper limit than the one presented in the guideline was suggested by Reis et al. [8], and Scherf-Clavel et al. also argued for a higher therapeutic reference range for the total moiety of venlafaxine and the active metabolite o-desmethyl-venlafaxine [12].
Opposite to the drugs mentioned above, four others were identified who had calculated therapeutic analytical ranges that were considerably lower than those reported in the AGNP consensus guideline. These were that of the antidepressants desipramine and sertraline and the antipsychotics ziprasidone and zuclopenthixol (Table 2).
The AGNP cites a single reference as the source of the range for desipramine, defined as being from 375 to 1125 nmol/L [3]. However, this study by Pedersen et al. focused on self-intoxicated patients, thus defining the upper limit of the range as the same as the toxic alarm level [3,34]. Desipramine has nonlinear kinetics, and although two studies by Nelson established a level of response at 431 nmol/L, to our knowledge, there is no study defining a complete therapeutic reference range [35,36,37]. According to the calculated therapeutic analytical range presented here, from 47 to 460 nmol/L (average from all participating laboratories), most patients would be below the responsive level if treated with desipramine alone (Table 2).
The range presented by the AGNP for sertraline is from 33 to 491 nmol/L [3]. In contrast, the two laboratory data sets included in our study both gave a calculated therapeutic analytical range <300 nmol/L (average 35–240 nmol/L; see Table 2). In a comparative study by Reis et al., also based on naturalistic data, a therapeutic analytical range was calculated as 19–182 nmol/L [8].
The therapeutic reference range defined in the AGNP consensus guideline for ziprasidone is from 128 to 510 nmol/L [3]. This seems to be based primarily on the two references by Cherma et al. and Vogel et al., who defined a range from 209 to 479 nmol/L and from 128 to 332 nmol/L, respectively [9,38]. Both of these reported results were based on naturalistic data and calculated as 25–75 percentiles. The calculated therapeutic analytical range in this study is considerably smaller, from 77–237 nmol/L, when performed using similar 25–75% fractions (Table 3—38–347 nmol/L with 10–90 percentile range). It should in this regard be noted that data only were available from one laboratory and at a relatively low number (n = 375). Regenthal et al., however, listed an even-lower therapeutic range for ziprasidone: from 51 to 153 nmol/L [39].
The therapeutic range reported by the AGNP for zuclopenthixol is from 10 to 125 nmol/L [3]. In contrast, the calculated therapeutic analytical range in our study, obtained from two separate laboratory data sets, gave a maximum range from 6 to 62 nmol/L (Table 2). In a comparable study by Jonsson et al., also based on routine laboratory data, they calculated a range of 5 to 65 nmol/L, while Kjølbye et al. suggested a range of 5 to 15 nmol/L [40,41]. All of these are considerably below that of the AGNP.
Although, in general, the closest correlation to the published therapeutic reference ranges seemed to be achieved by comparison to 10–90 percentiles, a nearly perfect match was obtained for amitriptyline + nortriptyline (total), aripiprazole, dehydroaripiprazole, aripiprazole + dehydroaripiprazole (total), and dosulipine when the same calculation was performed using 25–75 percentiles (Table 3). For amitriptyline + nortriptyline (total), the main references for the range in the AGNP guideline seems to be that of Perry et al. and that of Vandel et al. [42,43]. For aripiprazole, dehydroaripiprazole, and aripiprazole + dehydroaripiprazole (total), several of the listed references herein are population-based studies where the therapeutic reference range is calculated as 25–75 percentiles [44,45,46]. In contrast, the citations given for the therapeutic reference range of dosulipine are all older studies (>25 years) and not population based [3].

3.4. Investigation of Age and Sex Differences

Sex difference in the calculated therapeutic analytical ranges was investigated by using the built-in feature of the RefIT software. The combined and sex-specific therapeutic analytical ranges are shown in Table 4. In general, ranges calculated on the basis of data from women alone were higher than those from men. The only exception was for the antidepressant escitalopram.
To investigate the effect of age, therapeutic analytical ranges were calculated on the basis of two intervals: 20–64 years and 65–100 years (Table 5). Only minor differences could be observed for most of the therapeutic drugs, and no consistent trend was identified between the age groups.

4. Discussion

The purpose of the presented work was to compare routine laboratory data sets from multiple laboratories and evaluate their correspondence to published therapeutic reference ranges. As the data for calculating a therapeutic reference range rely on the study design, population type and size, dosage, and precision of the analytical method, conducting a direct comparison is challenging [7]. Additionally, in contrast to endogenous substances, there is no consensus method for the inclusion of patients or for calculating therapeutic reference ranges for drugs [3,7,47]. While a 10–90% range is commonly applied for population-based studies, many report a 25–75 percentile [8,41]. The AGNP consensus guideline does not clearly distinguish between them, and in addition, many of the listed ranges seem to have been concatenated from several studies (both population-based and controlled setups) [11]. In our experience, the use of a lower fraction can lead to too-narrow ranges. As measurements from the same patient may vary considerably even when all samples are taken at the minimum drug concentration at steady state, this can cause reason for untimely adjustment of dosage. Because of this, we opt for the use of 10–90% ranges, while for comparison, we have included values for the 25–75 percentiles (Table 2 and Table 3).
Although the developed TDM model allows the objective selection of data when calculating therapeutic analytical ranges, it does not provide the complete removal of all unwanted subpopulations from a data set. As the last sample is always included, the data set is presumed to contain results from patients who have had the treatment terminated for various reasons. By allowing the inclusion of more than one sample per patient, the model tries to correct for the effect of these, by enriching the data set for patients in long-term treatment. The model also increases the total number of included samples, thus obtaining better statistic support from a small data set compared with using only a single result per patient (Table 1).
The calculation of therapeutic analytical ranges from different data sets facilitated a comparison of the analytical methods used by the participating laboratories. In general, there was good concordance between them, except for three of the analytes. These were citalopram/escitalopram, mirtazapine, and quetiapine. The analysis of mirtazapine was discontinued in 2018 at the EHL owing to interference issues that could not be resolved on the standard equipment at that time. This discrepancy was therefore anticipated (Table 2). The deviations observed for citalopram/escitalopram may be attributed to variations from using a small data set. The low number of patients here may also suggest that the results have been requested as part of compliance or intoxication test rather than routine TDM monitoring. However, the comparison of the larger data sets shows the strength of monitoring the trend in patient data and using these for interlaboratory comparison. This because in contrast to regular internal quality controls and proficiency testing, it monitors over time the trend in real patient samples and matrices and provides a test of the whole environment of the analysis, from blood sampling to analytical instrument and calibration, through middleware, and to the final report in LIMS. An automated approach as presented here may thus provide a fast comparison between laboratories without exchanging samples.
Although the retrospective laboratory data from a TDM analysis come without any prior knowledge of dosage, comedication, reason why a test has been ordered, anamnesis of the patient, or therapeutic outcome, the data have the benefit of representing a naturalistic population rather than a specifically selected study cohort. Therefore, this range reflects all the factors which may influence the result of the measurement. While the upper limit of the therapeutic analytical range is influenced primarily by dosage and interpersonal variations, the lower 10 percentile may be influenced more by additional factors. These factors include the limit of the quantification of the analysis, the number of patients not in compliance, the prolonged fasting period prior to sampling, and the placebo effect where a clinical response is observed, at what would normally be at an insufficient dosage. Because of this, the deviation of the calculated lower limit of the therapeutic analytical range is expected to be larger than that for the upper limit, and correlates less to that of the therapeutic reference range. This may explain why the lower limits of the therapeutic analytical ranges were in general found to be below those of the published therapeutic reference ranges (Table 2). Although this held true for most of the drugs in this study, exceptions were observed for flupentixol, haloperidol, paroxetine, perphenazine, and total venlafaxine + o-desmethyl-venlafaxine. Here, both the calculated 10% and 90% limit of the percentiles were higher than the ranges published in the AGNP consensus guideline. In contrast, sertraline, ziprasidone, and zuclopenthixol each had an upper 90% range that was markedly lower than these. It is interesting to note that the drugs flupentixol, haloperidol, perphenazine, and zuclopenthixol all are quantified at the lower (<100) nmol/L level in plasma. As quantification at this scale even with modern equipment can be troublesome. This may explain some of the discrepancy as many of the therapeutic reference ranges dates back to studies performed in the 1980s, and are based on a single analytical method. However, the difference may also be due to the naturalistic setting of the data sets in this study, compared with a controlled population. Both perphenazine and zuclopenthixol have been found to be strongly influenced by the metabolic status of the CYP2D6 enzyme [48,49,50,51,52,53,54,55,56]. For these drugs, patients prescribed a standard dosage and who are either poor metabolisers or subject to drug interactions are expected to harbour a higher serum concentration. The higher upper fraction of the therapeutic analytical ranges in this study may thus reflect such patients in the data sets that are treated with a standard dosage.
Our data on paroxetine, which are supported by the findings of Reis et al., showed a higher upper therapeutic analytical range than the reference provided in the AGNP [3,8]. This is interesting as a negative effect on the treatment at higher dosage has been shown for this drug [57]. The higher upper limit of the calculated therapeutic analytical range may suggest that a relatively large number of the patients receiving paroxetine may be above the limit where there is negative response to the treatment [57,58]. It is therefore important that this difference is resolved.
Using the built-in features of the RefIT software, we investigated the effect of sex and age on the calculated therapeutic analytical ranges. Consistent with previous reports, women in general had higher blood concentrations than did men. This was true for all the investigated drugs, except for the antidepressant escitalopram (Table 4). Escitalopram is the stereo enantiomer of citalopram, and both are metabolised by the same CYP p450 enzymes, namely CYP2C19, CYP3A4, and to a lesser extent CYP2D6. Although previous studies have indicated that women have higher concentrations of some metabolites of escitalopram and higher activity of CYP3A4 than men do, studies conducted for the labelling of the drug did not show any sex difference [59,60]. However, as there are more results for women in the data set than for men, the difference could also be due to statistical uncertainty when applying the model.
Age is a second well-known factor influencing pharmacokinetics, with the primary reason being an acquired lower blood flow through the liver with ageing [61,62,63,64]. Comparing the calculated therapeutic analytical ranges for two age groups, 20–64 and 65–100 years, did not show any consistent differences between the results for the analysed drugs (Table 5). A possible explanation for this might be that the data used for the calculation were not corrected for dosage, and most patients in the data set were titrated to diminish the effect of age.
Although the calculated therapeutic analytical ranges do not provide any information on the effect of the treatment, a certain correlation with the therapeutic reference range is anticipated. The current study therefore shows that literature data cannot be directly transferred into a routine laboratory setting without validation. Furthermore, it highlights the importance of periodically re-evaluating laboratory therapeutic reference ranges as treatment paradigms may change over time [12]. Developing a consensus method for the calculation, as well as software capable of automatically handling large data sets, can help establish more reliable and thoroughly validated therapeutic reference ranges.

5. Conclusions

In this study, we designed a model for selecting and cleaning routine laboratory data. We then applied the algorithm for calculating therapeutic analytical ranges from the TDM data sets of antidepressants and antipsychotics from three laboratories. The analysis supported the therapeutic reference ranges published in the AGNP consensus guideline for the drugs amitriptyline + nortriptyline (total), nortriptyline, aripiprazole, aripiprazole + dehydroaripiprazole, citalopram, escitalopram, clomipramine + desclomipramine (total), clozapine, duloxetine, fluoxetine + norfluoxetine (total), imipramine + desipramine (total), mirtazapine, olanzapine, paliperidone, quetiapine, risperidone + paliperidone (total), and sertindole. In contrast, major deviations were observed when comparing laboratory data with the therapeutic reference ranges for the drugs desipramine, flupentixol, haloperidol, paroxetine, perphenazine, venlafaxine + o-desmethyl-venlafaxine (total), sertraline, ziprasidone, and zuclopenthixol. Support for our findings was found in other studies than those cited by the AGNP, and we therefore opt for a re-examination of these. Our model and our algorithm provide the first step of automation and present a tool to support the clinical validation of therapeutic reference ranges based on big data [10].

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/pharmaceutics15020673/s1, Table S1: Table 1—Supplementary data, Table S2: Table 2—Supplementary data, Table S3: Table 3—Supplementary data, Table S4: Table 4—Supplementary data, Table S5: Table 5—Supplementary data. The RefIT software v1 can be downloaded at https://github.com/JensLarsen/RefIT/releases/tag/V1.0.

Author Contributions

Conceptualisation, J.B.L., E.H.-L. and E.G.; methodology, J.B.L., E.H.-L. and E.G.; software, J.B.L.; validation, J.B.L., E.H.-L., P.H.A., N.R.J. and E.G.; formal analysis, J.B.L., E.H.-L., P.H.A., N.R.J. and E.G.; investigation, J.B.L., E.H.-L., P.H.A., N.R.J. and E.G.; data curation, J.B.L., E.H.-L., P.H.A., N.R.J. and E.G.; writing—original draft preparation, J.B.L.; writing—review and editing, J.B.L., E.H.-L., P.H.A., N.R.J. and E.G.; visualisation, J.B.L., E.H.-L. and E.G. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no funding.

Data Availability Statement

Data are contained within the article or supplementary material.

Conflicts of Interest

The authors declare no conflict of interests.

References

  1. Schwarz, M.J.; Hiemke, C.; Baumann, P.; AGNP TDM Expert Group. Clinical relevance of TDM of SSRIs. Ther. Drug Monit. 2006, 28, 716–717. [Google Scholar] [CrossRef] [PubMed]
  2. Urban, A.E.; Cubala, W.J. Therapeutic drug monitoring of atypical antipsychotics. Psychiatr. Pol. 2017, 51, 1059–1077. [Google Scholar] [CrossRef] [PubMed]
  3. Hiemke, C.; Bergemann, N.; Clement, H.W.; Conca, A.; Deckert, J.; Domschke, K.; Eckermann, G.; Egberts, K.; Gerlach, M.; Greiner, C.; et al. Consensus Guidelines for Therapeutic Drug Monitoring in Neuropsychopharmacology: Update 2017. Pharmacopsychiatry 2018, 51, 9–62. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  4. Bertino, J.S., Jr. Therapeutic drug monitoring of antibiotics. Lancet Infect. Dis. 2014, 14, 1180–1181. [Google Scholar] [CrossRef]
  5. Mohammadpour, N.; Elyasi, S.; Vahdati, N.; Mohammadpour, A.H.; Shamsara, J. A review on therapeutic drug monitoring of immunosuppressant drugs. Iran J. Basic Med. Sci. 2011, 14, 485–498. [Google Scholar]
  6. Bengtsson, F. Therapeutic drug monitoring of psychotropic drugs. TDM "nouveau". Ther. Drug Monit. 2004, 26, 145–151. [Google Scholar] [CrossRef]
  7. CLSI. Defining, Establishing, and Verifying Reference Intervals in the Clinical Laboratory. In Approved Guideline, 3rd ed.; CLSI document C28-A3; Clinical and Laboratory Standards Institute: Wayne, PA, USA, 2008. [Google Scholar]
  8. Reis, M.; Aamo, T.; Spigset, O.; Ahlner, J. Serum concentrations of antidepressant drugs in a naturalistic setting: Compilation based on a large therapeutic drug monitoring database. Ther. Drug Monit. 2009, 31, 42–56. [Google Scholar] [CrossRef]
  9. Chermá, M.D.; Reis, M.; Hägg, S.; Ahlner, J.; Bengtsson, F. Therapeutic drug monitoring of ziprasidone in a clinical treatment setting. Ther. Drug Monit. 2008, 30, 682–688. [Google Scholar] [CrossRef]
  10. Ma, C.; Wang, X.; Wu, J.; Cheng, X.; Xia, L.; Xue, F.; Qiu, L. Real-world big-data studies in laboratory medicine: Current status, application, and future considerations. Clin. Biochem. 2020, 84, 21–30. [Google Scholar] [CrossRef]
  11. Noel, C. A review of a recently published guidelines’ “strong recommendation” for therapeutic drug monitoring of olanzapine, haloperidol, perphenazine, and fluphenazine. Ment. Health Clin. 2019, 9, 287–293. [Google Scholar] [CrossRef]
  12. Scherf-Clavel, M.; Hommers, L.; Wurst, C.; Stonawski, S.; Deckert, J.; Domschke, K.; Unterecker, S.; Menke, A. Higher venlafaxine serum concentrations necessary for clinical improvement? Time to re-evaluate the therapeutic reference range of venlafaxine. J. Psychopharmacol. 2020, 34, 1105–1111. [Google Scholar] [CrossRef] [PubMed]
  13. Baumann, P.; Hiemke, C.; Ulrich, S.; Eckermann, G.; Gaertner, I.; Gerlach, M.; Kuss, H.J.; Laux, G.; Muller-Oerlinghausen, B.; Rao, M.L.; et al. The AGNP-TDM expert group consensus guidelines: Therapeutic drug monitoring in psychiatry. Pharmacopsychiatry 2004, 37, 243–265. [Google Scholar] [CrossRef] [PubMed]
  14. Baumann, P.; Hiemke, C.; Ulrich, S.; Eckermann, G.; Kuss, H.L.; Laux, G.; Muller-Oerlingenhausen, B.; Rao, M.L.; Riederer, P.; Zernig, G.; et al. Therapeutic drug monitoring (TDM) of psychotropic drugs: A consensus guideline of the AGNP-TDM group. Rev. Med. Suisse. 2006, 2, 1413–1418, 1420–1422, 1424–1426. [Google Scholar] [PubMed]
  15. Baumann, P.; Hiemke, C.; Ulrich, S.; Gaertner, I.; Rao, M.L.; Eckermann, G.; Gerlach, M.; Kuss, H.J.; Laux, G.; Muller-Oerlinghausen, B.; et al. Therapeutic monitoring of psychotropic drugs: An outline of the AGNP-TDM expert group consensus guideline. Ther. Drug Monit. 2004, 26, 167–170. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  16. Baumann, P.; Ulrich, S.; Eckermann, G.; Gerlach, M.; Kuss, H.J.; Laux, G.; Muller-Oerlinghausen, B.; Rao, M.L.; Riederer, P.; Zernig, G.; et al. The AGNP-TDM Expert Group Consensus Guidelines: Focus on therapeutic monitoring of antidepressants. Dialogues Clin. Neurosci. 2005, 7, 231–247. [Google Scholar] [CrossRef] [PubMed]
  17. Hiemke, C. Consensus Guideline Based Therapeutic Drug Monitoring (TDM) in Psychiatry and Neurology. Curr. Drug Deliv. 2016, 13, 353–361. [Google Scholar] [CrossRef]
  18. Unterecker, S.; Hefner, G.; Baumann, P.; Grunder, G.; Bergemann, N.; Clement, H.W.; Conca, A.; Deckert, J.; Domschke, K.; Eckermann, G.; et al. Therapeutic drug monitoring in neuropsychopharmacology: Summary of the consensus guidelines 2017 of the TDM task force of the AGNP. Nervenarzt 2019, 90, 463–471, Erratum in Nervenarzt 2019, 90, 958–960. [Google Scholar] [CrossRef] [Green Version]
  19. Van Putten, T.; Marder, S.R.; Wirshing, W.C.; Aravagiri, M.; Chabert, N. Neuroleptic plasma levels. Schizophr. Bull. 1991, 17, 197–216. [Google Scholar] [CrossRef] [Green Version]
  20. Kistrup, K.; Gerlach, J.; Aaes-Jorgensen, T.; Larsen, N.E. Perphenazine decanoate and cis(z)-flupentixol decanoate in maintenance treatment of schizophrenic outpatients. Serum levels at the minimum effective dose. Psychopharmacology 1991, 105, 42–48. [Google Scholar] [CrossRef]
  21. Morselli, P.L.; Bianchetti, G.; Dugas, M. Haloperidol plasma level monitoring in neuropsychiatric patients. Ther. Drug Monit. 1982, 4, 51–58. [Google Scholar] [CrossRef] [PubMed]
  22. Volavka, J.; Cooper, T.B.; Meisner, M.; Bitter, I.; Czobor, P.; Jaeger, J. Haloperidol blood levels and effects in schizophrenia and schizoaffective disorder: A progress report. Psychopharmacol. Bull. 1990, 26, 13–17. [Google Scholar] [PubMed]
  23. Stevens, A.; Stevens, I.; Mahal, A.; Gaertner, H.J. Haloperidol and lorazepam combined: Clinical effects and drug plasma levels in the treatment of acute schizophrenic psychosis. Pharmacopsychiatry 1992, 25, 273–277. [Google Scholar] [CrossRef] [PubMed]
  24. Stevens, A.; Mahal, A.; Gaertner, H.J. Haloperidol and reduced haloperidol serum levels: Correlation with psychopathology in acute schizophrenia. Pharmacopsychiatry 1992, 25, 218–223. [Google Scholar] [CrossRef] [PubMed]
  25. Palao, D.J.; Arauxo, A.; Brunet, M.; Bernardo, M.; Haro, J.M.; Ferrer, J.; Gonzalez-Monclus, E. Haloperidol: Therapeutic window in schizophrenia. J. Clin. Psychopharmacol. 1994, 14, 303–310. [Google Scholar] [CrossRef] [PubMed]
  26. Ulrich, S.; Wurthmann, C.; Brosz, M.; Meyer, F.P. The relationship between serum concentration and therapeutic effect of haloperidol in patients with acute schizophrenia. Clin. Pharmacokinet. 1998, 34, 227–263. [Google Scholar] [CrossRef] [PubMed]
  27. Ulrich, S.; Neuhof, S.; Braun, V.; Meyer, F.P. Therapeutic window of serum haloperidol concentration in acute schizophrenia and schizoaffective disorder. Pharmacopsychiatry 1998, 31, 163–169. [Google Scholar] [CrossRef]
  28. Balant-Gorgia, A.E.; Eisele, R.; Aeschlimann, J.M.; Balant, L.P.; Garrone, G. Plasma flupentixol concentrations and clinical response in acute schizophrenia. Ther. Drug Monit. 1985, 7, 411–414. [Google Scholar] [CrossRef]
  29. Roman, M.; Kronstrand, R.; Lindstedt, D.; Josefsson, M. Quantitation of seven low-dosage antipsychotic drugs in human postmortem blood using LC-MS-MS. J. Anal. Toxicol. 2008, 32, 147–155. [Google Scholar] [CrossRef] [Green Version]
  30. Gex-Fabry, M.; Gervasoni, N.; Eap, C.B.; Aubry, J.M.; Bondolfi, G.; Bertschy, G. Time course of response to paroxetine: Influence of plasma level. Prog. Neuropsychopharmacol. Biol. Psychiatry 2007, 31, 892–900. [Google Scholar] [CrossRef]
  31. Tomita, T.; Yasui-Furukori, N.; Nakagami, T.; Tsuchimine, S.; Ishioka, M.; Kaneda, A.; Nakamura, K.; Kaneko, S. Therapeutic reference range for plasma concentrations of paroxetine in patients with major depressive disorders. Ther. Drug Monit. 2014, 36, 480–485. [Google Scholar] [CrossRef] [PubMed]
  32. Shams, M.E.; Arneth, B.; Hiemke, C.; Dragicevic, A.; Muller, M.J.; Kaiser, R.; Lackner, K.; Hartter, S. CYP2D6 polymorphism and clinical effect of the antidepressant venlafaxine. J. Clin. Pharm. Ther. 2006, 31, 493–502. [Google Scholar] [CrossRef] [PubMed]
  33. Veefkind, A.H.; Haffmans, P.M.; Hoencamp, E. Venlafaxine serum levels and CYP2D6 genotype. Ther. Drug Monit. 2000, 22, 202–208. [Google Scholar] [CrossRef] [PubMed]
  34. Pedersen, O.L.; Gram, L.F.; Kristensen, C.B.; Moller, M.; Thayssen, P.; Bjerre, M.; Kragh-SoRensen, P.; Klitgaard, N.A.; Sindrup, E.; Hole, P.; et al. Overdosage of antidepressants: Clinical and pharmacokinetic aspects. Eur. J. Clin. Pharmacol. 1982, 23, 513–521. [Google Scholar] [CrossRef] [PubMed]
  35. Cooke, R.G.; Warsh, J.J.; Stancer, H.C.; Reed, K.L.; Persad, E. The nonlinear kinetics of desipramine and 2-hydroxydesipramine in plasma. Clin. Pharmacol. Ther. 1984, 36, 343–349. [Google Scholar] [CrossRef] [PubMed]
  36. Nelson, J.C.; Jatlow, P.; Quinlan, D.M.; Bowers, M.B., Jr. Desipramine plasma concentration and antidepressant response. Arch. Gen. Psychiatry 1982, 39, 1419–1422. [Google Scholar] [CrossRef]
  37. Nelson, J.C.; Jatlow, P.I.; Mazure, C. Desipramine plasma levels and response in elderly melancholic patients. J. Clin. Psychopharmacol. 1985, 5, 217–220. [Google Scholar] [CrossRef]
  38. Vogel, F.; Gansmuller, R.; Leiblein, T.; Dietmaier, O.; Wassmuth, H.; Grunder, G.; Hiemke, C. The use of ziprasidone in clinical practice: Analysis of pharmacokinetic and pharmacodynamic aspects from data of a drug monitoring survey. Eur. Psychiatry 2009, 24, 143–148. [Google Scholar] [CrossRef]
  39. Regenthal, R.; Krueger, M.; Koeppel, C.; Preiss, R. Drug levels: Therapeutic and toxic serum/plasma concentrations of common drugs. J. Clin. Monit. Comput. 1999, 15, 529–544. [Google Scholar] [CrossRef]
  40. Kjølbye, M.; Thomsen, K.; Rogne, T.; Rehfelt, E.; Olesen, O.V. Search for a therapeutic range for serum zuclopenthixol concentrations in schizophrenic patients. Ther. Drug Monit. 1994, 16, 541–547. [Google Scholar] [CrossRef]
  41. Jonsson, A.K.; Spigset, O.; Reis, M. A Compilation of Serum Concentrations of 12 Antipsychotic Drugs in a Therapeutic Drug Monitoring Setting. Ther. Drug Monit. 2019, 41, 348–356. [Google Scholar] [CrossRef] [PubMed]
  42. Perry, P.J.; Zeilmann, C.; Arndt, S. Tricyclic antidepressant concentrations in plasma: An estimate of their sensitivity and specificity as a predictor of response. J. Clin. Psychopharmacol. 1994, 14, 230–240. [Google Scholar] [CrossRef] [PubMed]
  43. Vandel, S.; Vandel, B.; Sandoz, M.; Allers, G.; Bechtel, P.; Volmat, R. Clinical response and plasma concentration of amitriptyline and its metabolite nortriptyline. Eur. J. Clin. Pharmacol. 1978, 14, 185–190. [Google Scholar] [CrossRef] [PubMed]
  44. Molden, E.; Lunde, H.; Lunder, N.; Refsum, H. Pharmacokinetic variability of aripiprazole and the active metabolite dehydroaripiprazole in psychiatric patients. Ther. Drug Monit. 2006, 28, 744–749. [Google Scholar] [CrossRef]
  45. Kirschbaum, K.M.; Muller, M.J.; Malevani, J.; Mobascher, A.; Burchardt, C.; Piel, M.; Hiemke, C. Serum levels of aripiprazole and dehydroaripiprazole, clinical response and side effects. World J. Biol. Psychiatry 2008, 9, 212–218. [Google Scholar] [CrossRef]
  46. Bachmann, C.J.; Rieger-Gies, A.; Heinzel-Gutenbrunner, M.; Hiemke, C.; Remschmidt, H.; Theisen, F.M. Large variability of aripiprazole and dehydroaripiprazole serum concentrations in adolescent patients with schizophrenia. Ther. Drug Monit. 2008, 30, 462–466. [Google Scholar] [CrossRef]
  47. Horn, P.S.; Pesce, A.J. Reference intervals: An update. Clinica Chimica Acta 2003, 334, 5–23. [Google Scholar] [CrossRef]
  48. Fang, J.; Baker, G.B.; Silverstone, P.H.; Coutts, R.T. Involvement of CYP3A4 and CYP2D6 in the metabolism of haloperidol. Cell Mol. Neurobiol. 1997, 17, 227–233. [Google Scholar] [CrossRef]
  49. Jerling, M.; Dahl, M.L.; Aberg-Wistedt, A.; Liljenberg, B.; Landell, N.E.; Bertilsson, L.; Sjoqvist, F. The CYP2D6 genotype predicts the oral clearance of the neuroleptic agents perphenazine and zuclopenthixol. Clin. Pharmacol. Ther. 1996, 59, 423–428. [Google Scholar] [CrossRef]
  50. Linnet, K.; Wiborg, O. Steady-state serum concentrations of the neuroleptic perphenazine in relation to CYP2D6 genetic polymorphism. Clin. Pharmacol. Ther. 1996, 60, 41–47. [Google Scholar] [CrossRef]
  51. Linnet, K.; Wiborg, O. Influence of Cyp2D6 genetic polymorphism on ratios of steady-state serum concentration to dose of the neuroleptic zuclopenthixol. Ther. Drug Monit. 1996, 18, 629–634. [Google Scholar] [CrossRef] [PubMed]
  52. Ohara, K.; Tanabu, S.; Ishibashi, K.; Ikemoto, K.; Yoshida, K.; Shibuya, H. Effects of age and the CYP2D6*10 allele on the plasma haloperidol concentration/dose ratio. Prog. Neuropsychopharmacol. Biol. Psychiatry 2003, 27, 347–350. [Google Scholar] [CrossRef] [PubMed]
  53. Ozdemir, V.; Bertilsson, L.; Miura, J.; Carpenter, E.; Reist, C.; Harper, P.; Widen, J.; Svensson, J.O.; Albers, L.J.; Kennedy, J.L.; et al. CYP2D6 genotype in relation to perphenazine concentration and pituitary pharmacodynamic tissue sensitivity in Asians: CYP2D6-serotonin-dopamine crosstalk revisited. Pharmacogenet. Genomics 2007, 17, 339–347. [Google Scholar] [CrossRef] [PubMed]
  54. Suzuki, A.; Otani, K.; Mihara, K.; Yasui, N.; Kaneko, S.; Inoue, Y.; Hayashi, K. Effects of the CYP2D6 genotype on the steady-state plasma concentrations of haloperidol and reduced haloperidol in Japanese schizophrenic patients. Pharmacogenetics 1997, 7, 415–418. [Google Scholar] [CrossRef]
  55. Tveito, M.; Smith, R.L.; Molden, E.; Hoiseth, G. Impact of age and CYP2D6 genotype on exposure of zuclopenthixol in patients using long-acting injectable versus oral formulation-an observational study including 2044 patients. Eur. J. Clin. Pharmacol. 2021, 77, 215–221. [Google Scholar] [CrossRef]
  56. Waade, R.B.; Solhaug, V.; Hoiseth, G. Impact of CYP2D6 on serum concentrations of flupentixol, haloperidol, perphenazine and zuclopenthixol. Br. J. Clin. Pharmacol. 2021, 87, 2228–2235. [Google Scholar] [CrossRef]
  57. Tomita, T.; Yasui-Furukori, N.; Nakagami, T.; Tsuchimine, S.; Ishioka, M.; Kaneda, A.; Nakamura, K. Effects of personality on the association between paroxetine plasma concentration and response. Neuropsychiatr. Dis. Treat. 2018, 14, 3299–3306. [Google Scholar] [CrossRef] [Green Version]
  58. Gilles, C.; Luthringer, R. Pharmacological models in healthy volunteers: Their use in the clinical development of psychotropic drugs. J. Psychopharmacol. 2007, 21, 272–282. [Google Scholar] [CrossRef]
  59. Ji, Y.; Schaid, D.J.; Desta, Z.; Kubo, M.; Batzler, A.J.; Snyder, K.; Mushiroda, T.; Kamatani, N.; Ogburn, E.; Hall-Flavin, D.; et al. Citalopram and escitalopram plasma drug and metabolite concentrations: Genome-wide associations. Br. J. Clin. Pharmacol. 2014, 78, 373–383. [Google Scholar] [CrossRef] [Green Version]
  60. Sramek, J.J.; Murphy, M.F.; Cutler, N.R. Sex differences in the psychopharmacological treatment of depression. Dialogues Clin. Neurosci. 2016, 18, 447–457. [Google Scholar] [CrossRef]
  61. Edelbroek, P.M.; Zitman, F.G.; Knoppert-van der Klein, E.A.; van Putten, P.M.; de Wolff, F.A. Therapeutic drug monitoring of amitriptyline: Impact of age, smoking and contraceptives on drug and metabolite levels in bulimic women. Clin. Chim. Acta 1987, 165, 177–187. [Google Scholar] [CrossRef] [PubMed]
  62. Gex-Fabry, M.; Balant-Gorgia, A.E.; Balant, L.P. Therapeutic drug monitoring of olanzapine: The combined effect of age, gender, smoking, and comedication. Ther. Drug Monit. 2003, 25, 46–53. [Google Scholar] [CrossRef] [PubMed]
  63. Hansen, M.R.; Kuhlmann, I.B.; Pottegard, A.; Damkier, P. Therapeutic Drug Monitoring of Venlafaxine in an Everyday Clinical Setting: Analysis of Age, Sex and Dose Concentration Relationships. Basic Clin. Pharmacol. Toxicol. 2017, 121, 298–302. [Google Scholar] [CrossRef] [PubMed] [Green Version]
  64. Unterecker, S.; Riederer, P.; Proft, F.; Maloney, J.; Deckert, J.; Pfuhlmann, B. Effects of gender and age on serum concentrations of antidepressants under naturalistic conditions. J. Neural. Transm. 2013, 120, 1237–1246. [Google Scholar] [CrossRef]
Pharmaceutics 15 00673 g001 550
Figure 1. Flowchart of the TDM model used for qualification of data for calculating therapeutic analytical ranges.
Figure 1. Flowchart of the TDM model used for qualification of data for calculating therapeutic analytical ranges.
Pharmaceutics 15 00673 g001
Pharmaceutics 15 00673 g002 550
Figure 2. Plot of data points from the data set of clozapine from Aarhus University Hospital. Red dots are the sample results excluded and green are the samples included by the TDM model. The median is shown as a blue line, while the 10 and 90 percentiles are shown in black. The percentiles and median were calculated on the basis of the included samples only (included n = 3224, excluded n = 10,871).
Figure 2. Plot of data points from the data set of clozapine from Aarhus University Hospital. Red dots are the sample results excluded and green are the samples included by the TDM model. The median is shown as a blue line, while the 10 and 90 percentiles are shown in black. The percentiles and median were calculated on the basis of the included samples only (included n = 3224, excluded n = 10,871).
Pharmaceutics 15 00673 g002
Pharmaceutics 15 00673 g003 550
Figure 3. Evaluation of the described TDM model for selecting results from retrospective laboratory data sets. Graphs shows percentile plots for clozapine as examples, where each line represents a sample result. (A) All samples in the data set, (B) samples selected using the TDM model, (C) only the first result from a patient is used, and (D) only the last result from a patient is used. The 10–90% percentile is marked in green. The graphs were generated and exported by using the RefIT software.
Figure 3. Evaluation of the described TDM model for selecting results from retrospective laboratory data sets. Graphs shows percentile plots for clozapine as examples, where each line represents a sample result. (A) All samples in the data set, (B) samples selected using the TDM model, (C) only the first result from a patient is used, and (D) only the last result from a patient is used. The 10–90% percentile is marked in green. The graphs were generated and exported by using the RefIT software.
Pharmaceutics 15 00673 g003
Table
Table 1. Therapeutic analytical ranges for clozapine, perphenazine, and imipramine calculated based on the data set from Aarhus University Hospital.—(A) all samples in the data set, (B) samples selected by using the TDM model, (C) only the first result from a patient, and (D) only the last result from a patient.
Table 1. Therapeutic analytical ranges for clozapine, perphenazine, and imipramine calculated based on the data set from Aarhus University Hospital.—(A) all samples in the data set, (B) samples selected by using the TDM model, (C) only the first result from a patient, and (D) only the last result from a patient.
Clozapine
Patients = 1398		Perphenazine
Patients = 405		Imipramine
Patients = 127
Incl. SamplesRange nmol/LIncl. SamplesRange nmol/LIncl. SamplesRange nmol/L
(A) All samples14,095409–214312241.5–14.328143–846
(B) TDM model3224400–19806211.8–13.814627–437
(C) First sample1398276–20344051.0–14.412725–331
(D) Last sample1398354–20914051.5–13.612726–390
Table
Table 2. Comparison between therapeutic reference ranges listed in the AGNP consensus guideline and calculated 10–90 percentiles from the three laboratories. The data were selected on the basis of the TDM model, and ranges were calculated by using the RefIT software. The number of patients is shown under p. Abbreviations: ND—no data, AUH—Aarhus University Hospital, EHL—the Danish Epilepsy Centre, and RH—Rigshospitalet. Total number of samples, number of included samples used for the calculation of percentiles, and additional statistics are supplied as Supplementary Data Table S1.
Table 2. Comparison between therapeutic reference ranges listed in the AGNP consensus guideline and calculated 10–90 percentiles from the three laboratories. The data were selected on the basis of the TDM model, and ranges were calculated by using the RefIT software. The number of patients is shown under p. Abbreviations: ND—no data, AUH—Aarhus University Hospital, EHL—the Danish Epilepsy Centre, and RH—Rigshospitalet. Total number of samples, number of included samples used for the calculation of percentiles, and additional statistics are supplied as Supplementary Data Table S1.
Therapeutic Drug AUHEHLRH
AGNPpRange nmol/LpRange nmol/LpRange nmol/L
AmitriptylineND103335–46461283–479149220–441
Nortriptyline266–6464436105–6232016104–630467169–661
Amitriptyline + metabolite288–720130879–809570201–9171025129–906
Aripiprazole223–7811695116–11481449154–1064ND-
DehydroaripiprazoleND166242–347136683–406ND-
Aripiprazole + metabolite 335–11151889217–13781365283–1465ND-
Citalopram154–339198169–37963280–499ND-
Escitalopram46–24679533–1979931–283ND-
Dosulipine153–339ND-ND-10720–505
NorthiadenNDND-ND-10520–356
Dosulipine + metaboliteNDND-ND-92119–888
ClomipramineND119296–610400122–62844871–652
DesmethylclomipramineND118582–777394136–89544545–950
Clomipramine + metabolite731–14941156285–1252394326–1445409272–1497
Clozapine1071–18361398400–19801355349–23892028130–2180
Duloxetine100–403146747–42817947–421ND-
FluoxetineND468118–1183ND-ND-
NorfluoxetineND472255–1169ND-ND-
Fluoxetine + metabolite388–1695326356–1486ND-ND-
Flupentixol1.2–11.5ND-1362–23ND-
Haloperidol2.7–26.6ND-4455–48ND-
ImipramineND12727–4379571–53017820–588
Desimipramine375–112512620–4098873–51017920–528
Imipramine + metabolite641–109812468–81279175–1032136141–1137
Mirtazapine113–302100037–25625375–398ND-
Olanzapine64–256185438–247266330–273ND-
Paroxetine61–198ND-26861–537ND-
Perphenazine1.5–64051.8–13.83602–18ND-
Quetiapine261–1305265333–851164948–1268ND-
RisperidoneND15643–588807–82ND-
Paliperidone47–141234911–10995917–121ND-
Risperidone + metabolite41–146209816–13985135–177ND-
Sertindole114–227ND-11448–236ND-
Sertraline33–491343833–23393436–289ND-
VenlafaxineND415457–966293896–1073ND-
O-Desmethyl-venlafaxineND4181309–17743107312–1694ND-
Venlafaxine + metabolite.361–15204134489–25222926615–2588ND-
Ziprasidone128–51024638–347ND-ND-
Zuclopenthixole10–1258856.5–42.810627–62.2ND-
Table
Table 3. Comparison of calculated 25–75 percentiles between the participating laboratories and to the therapeutic reference ranges published in the AGNP consensus guideline. For abbreviations, see Table 2. Additional statistics is in supplementary File Table S2.
Table 3. Comparison of calculated 25–75 percentiles between the participating laboratories and to the therapeutic reference ranges published in the AGNP consensus guideline. For abbreviations, see Table 2. Additional statistics is in supplementary File Table S2.
AUHEHLRH
Therapeutic DrugAGNPpRange nmol/LpRange nmol/LpRange nmol/L
AmitriptylineND103377–310 612140–332149220–281
Nortriptyline266–6464436221–5042016195–4764671195–502
Amitriptyline + metabolite288–7201308180–596570304–6841025245–676
Aripiprazole223–7811695266–7881449279–726ND-
DehydroaripiprazoleND166293–2551366123–294ND-
Aripiprazole + metabolite335–11151889392–10351365448–1040ND-
Citaloprame154–3391981113–271642133–336ND-
Escitaloprame46–24679550–1339954–134ND-
Dosulipine153–339ND-ND-107100–347
NorthiadenNDND-ND-10544–206
Dosulipine + metaboliteNDND-ND-92203–578
ClomipramineND1192188–444400191–439448167–444
DesmethylclomipramineND1185188–575394231–603445189–637
Clomipramine + metabolite731–14941156477–1011394477–990409474–1127
Clozapine1071–18361398677–15281355656–17502028478–1502
Duloxetine100–403146786–27817983–259ND-
FluoxetineND468249–771115363–989ND-
NorfluoxetineND472421–845112401–764ND-
Fluoxetine + metabolite388–1695326584–1100110838–1895ND-
Flupentixole1.2–11.5ND-1363–13ND-
Haloperidole2.7–26.6ND-4459–31ND-
ImipramineND12771–25395100–32417868–394
Desimipramine375–112512641–2498897–34717955–327
Imipramin + metabolite641–1098124128–53579262–632136254–857
Mirtazapine113–302100064–178253106–242ND-
Olanzapine64–256185465–167266358–178ND-
Paroxetine61–198ND-268114–321ND-
Perphenazine1.5–64053.2–8.43604–11ND-
Quetiapine261–1305265379–4901649125–709ND-
RisperidoneND15646–2888010–38ND-
Paliperidone47–141234923–7395930–82ND-
Risperidone + metabolite41–146209831–9585153–123ND-
Sertindole114–227ND-11478–160ND-
Sertraline33–491343858–15393462–179ND-
VenlafaxineND4154133–5752938165–645ND-
O-Desmethyl-venlafaxineND4181545–13283107530–1267ND-
Venlafaxine + metabolite.361–15204134815–19012926883–1921ND-
Ziprasidone128–51024677–237ND-ND-
Zuclopenthixole10–12588510.5–29.5106213–41ND-
Table
Table 4. Comparison between calculated 10–90 percentile therapeutic analytical ranges for men and women performed on the data set from Aarhus University Hospital. Here, p indicates the number of patients that the calculations are based on. For additional statistics, see Table S4.
Table 4. Comparison between calculated 10–90 percentile therapeutic analytical ranges for men and women performed on the data set from Aarhus University Hospital. Here, p indicates the number of patients that the calculations are based on. For additional statistics, see Table S4.
MenWomen
Therapeutic DrugpRange nmol/LpRange nmol/L
Amitriptyline36634–42366636–473
Nortriptyline161898–6132817111–629
Amitriptyline + metabolite44676–75386181–835
Aripiprazole79791–1122897138–1197
Dehydroaripiprazole78435–33487749–356
Aripiprazole + metabolite883201–13541005237–1400
Citalopram67166–331130970–392
Escitalopram26934–21352632–186
Clomipramine45781–606734110–613
Desmethylclomipramine45469–71473090–823
Clomipramine + metabolite441241–1183714307–1276
Clozapine818380–1930578428–2030
Duloxetine44638–373102050–438
Fluoxetine12483–836344138–1255
Norfluoxetine125206–1065347291–1223
Fluoxetine + metabolite94323–1215232380–1633
Mirtazapine43336–24356739–262
Olanzapine107738–23377638–272
Perphenazine1811.7–12.32241.8–15.3
Quetiapine109837–857155331–847
Risperidone8673–576964–59
Paliperidone135511–10399311–118
Risperidone + metabolite120216–12789517–157
Sertraline110831–223233034–236
Venlafaxine137351–840278061–1043
O-Desmethyl-venlafaxine1381286–16992799316–1798
Venlafaxine + metabolite.1369462–24142764517–2582
Ziprasidone8834–32215740–376
Zuclopenthixole4897.4–41.23966.1–44.6
Table
Table 5. Examination of the influence of age on the calculated 10–90 percentile therapeutic analytical range. Ranges of 20–64 and 65–100 years were calculated with the RefIT software by using the data set from Aarhus University Hospital. Here, p indicates the number of patients included in the data set of each compound. ND—no data; <100 samples. For additional statistics, see Supplementary Data File Table S5.
Table 5. Examination of the influence of age on the calculated 10–90 percentile therapeutic analytical range. Ranges of 20–64 and 65–100 years were calculated with the RefIT software by using the data set from Aarhus University Hospital. Here, p indicates the number of patients included in the data set of each compound. ND—no data; <100 samples. For additional statistics, see Supplementary Data File Table S5.
Age 20–64Age 65–100
Therapeutic DrugpRange nmol/LpRange nmol/L
Amitriptyline77234–45425039–471
Nortriptyline3510107–62594797–614
Amitriptyline + metabolite121477–80032887–836
Aripiprazole1424135–11541081–1196
Dehydroaripiprazole140546–3549612–380
Aripiprazole + metabolite1590211–1381102221–1409
Citalopram140865–36355478–409
Escitalopram60732–18818836–219
Clomipramine99699–62319683–560
Desmethyl-clomipramine98785–77519770–804
Clomipramine + metabolite964284–1252192314–1234
Clozapine1272411–1999137333–1793
Duloxetine115443–40229063–468
Fluoxetine269111–1210ND-
Norfluoxetine270246–1119ND-
Fluoxetine + metabolite215327–1650ND-
Mirtazapine61534–23537642–280
Olanzapine149739–25632335–204
Perphenazine3351.6–14692.1–12.7
Quetiapine215233–87735231–699
Risperidone12383–622703.3–42
Paliperidone188312–11135311–115
Risperidone + Paliperidone164718–14134118–142
Sertraline265032–23240232–231
Venlafaxine344554–93470176–1084
O-Desmethyl-venlafaxine3470295–1699703364–1973
Venlafaxine + metabolite3430468–2460697683–2774
Ziprasidone23040–361ND-
Zuclopenthixol7296.9–44.11645.5–35.5
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Larsen, J.B.; Hoffmann-Lücke, E.; Aaslo, P.H.; Jørgensen, N.R.; Greibe, E. Automated Interlaboratory Comparison of Therapeutic Drug Monitoring Data and Its Use for Evaluation of Published Therapeutic Reference Ranges. Pharmaceutics 2023, 15, 673. https://doi.org/10.3390/pharmaceutics15020673

AMA Style

Larsen JB, Hoffmann-Lücke E, Aaslo PH, Jørgensen NR, Greibe E. Automated Interlaboratory Comparison of Therapeutic Drug Monitoring Data and Its Use for Evaluation of Published Therapeutic Reference Ranges. Pharmaceutics. 2023; 15(2):673. https://doi.org/10.3390/pharmaceutics15020673

Chicago/Turabian Style

Larsen, Jens Borggaard, Elke Hoffmann-Lücke, Per Hersom Aaslo, Niklas Rye Jørgensen, and Eva Greibe. 2023. "Automated Interlaboratory Comparison of Therapeutic Drug Monitoring Data and Its Use for Evaluation of Published Therapeutic Reference Ranges" Pharmaceutics 15, no. 2: 673. https://doi.org/10.3390/pharmaceutics15020673

APA Style

Larsen, J. B., Hoffmann-Lücke, E., Aaslo, P. H., Jørgensen, N. R., & Greibe, E. (2023). Automated Interlaboratory Comparison of Therapeutic Drug Monitoring Data and Its Use for Evaluation of Published Therapeutic Reference Ranges. Pharmaceutics, 15(2), 673. https://doi.org/10.3390/pharmaceutics15020673

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop