How appropriate is therapeutic drug monitoring for lithium? Data from the Belgian external quality assessment scheme

Clinical Biochemistry 48 (2015) 617–621

Contents lists available at ScienceDirect

Clinical Biochemistry journal homepage: www.elsevier.com/locate/clinbiochem

How appropriate is therapeutic drug monitoring for lithium? Data from the Belgian external quality assessment scheme I.K. Delattre a,d, P. Van de Walle a,⁎, C. Van Campenhout a, H. Neels b, A.G. Verstraete c, P. Wallemacq d a

Quality of Medical Laboratories, Scientific Institute of Public Health, Belgium Laboratory of Toxicology and Therapeutic Drug Monitoring, Ziekenhuis Netwerk Antwerpen campus Stuivenberg, Belgium Laboratory of Clinical Biology/Toxicology, Ghent University Hospital, Belgium d Laboratory of Toxicology and Therapeutic Drug Monitoring, University Hospital St Luc, Belgium b c

a r t i c l e

i n f o

Article history: Received 18 December 2014 Received in revised form 9 March 2015 Accepted 11 March 2015 Available online 25 March 2015 Keywords: Bipolar disorder External quality assessment Lithium Therapeutic drug monitoring

a b s t r a c t Background: Lithium remains a mainstay in the management of mood disorders. As with many psychotropic drugs, lithium treatment requires continuous observation for adverse effects and strict monitoring of serum concentrations. The present study aimed to assess the appropriateness of lithium assays used by Belgian laboratories, and to evaluate acceptability of their clinical interpretations. Methods: Nine in-house serum samples spiked with predetermined concentrations of lithium were distributed to 114 participants in the Belgian external quality assessment scheme. Laboratories were requested to report the assay technique, lithium measurements and interpretations with regard to measured concentrations. Inter/intramethod imprecision and bias were reported and acceptability of clinical interpretations was assessed. The intramethod variability was evaluated by selecting methods used by 6 laboratories or more. Flame photometry (IL 943) was considered as the reference method. Results: Laboratories returned assay results using colorimetry (69.3%), ion selective electrode (15.8%), flame photometry (8.8%), atomic absorption spectroscopy (5.2%) or mass spectrometry (0.9%). Lithium concentrations were systematically higher when measured with the Vitros assay (median bias: 4.0%), and were associated with consecutive biased interpretations. In contrast, the Thermo Scientific Infinity assay showed a significant negative bias (median bias: 9.4%). 36.0% of laboratories reported numerical values below their manufacturer cut-off for the blank sample; 16.6% of these laboratories detected residual lithium concentrations. Conclusions: The present study revealed assay-related differences in lithium measurements and their interpretations. Overall, there appeared to be a need to continue EQA of therapeutic drug monitoring for lithium in Belgium. © 2015 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.

Introduction After more than 60 years of experience, lithium remains a keystone therapy in bipolar disorders. Most experts and recent guidelines still consider the drug as a first-choice mood stabilizer, protecting against both depression and mania and reducing the risk of suicide and shortterm mortality [1–6]. Lithium efficacy is clearly dose-dependent and reliably correlates with serum concentrations [7]. However, the therapeutic benefits of lithium are restricted by adverse effects (i.e. neurological, cardiovascular and renal) and a narrow therapeutic range (0.8–1.2 mmol/L) [8,9]. In addition, its toxicity can occur at therapeutic doses due to inter- and intraindividual pharmacokinetic variations, and may result ⁎ Corresponding author at: Quality of Medical Laboratories, Scientific Institute of Public Health, Rue Juliette Wytsman, 14, B-1050 Brussels, Belgium. Fax: +32 2 642 56 45. E-mail address: [email protected] (P. Van de Walle).

in mortality [10,11]. The combined use of lithium with other medications may moreover result in substantial increase in lithium levels and thereby precipitating intoxication [11,12]. Concentration monitoring is therefore required for optimal use and dosing of lithium. Although alternative methods of blood sampling have been suggested, serum measurements remains the mainstay of therapeutic drug monitoring (TDM) of lithium [12]. Lithium dosages should be adjusted in steady-state patients, using the serum trough level (i.e. usually measured 12 hours after the last twice-daily dose) [11,12]. Analytical methods available for routine determination of lithium in serum include colorimetry, flame photometry, atomic absorption spectroscopy and potentiometry with ion-selective electrode [13,14]. In Belgium, the External Quality Assessment Scheme (EQAS) for TDM has been monitoring the performance of lithium assays since 2008 and now has N 100 participant laboratories, including both hospital and private laboratories. EQAS is a unique opportunity to assess the appropriateness of TDM for lithium. The present study aimed to

http://dx.doi.org/10.1016/j.clinbiochem.2015.03.009 0009-9120/© 2015 The Canadian Society of Clinical Chemists. Published by Elsevier Inc. All rights reserved.

618

I.K. Delattre et al. / Clinical Biochemistry 48 (2015) 617–621

Table 1 Intra- and interassay results obtained in the Belgian EQAS for lithium. Methods

Flame photometry with internal standard (n = 8) Atomic absorption spectroscopy (n = 6) Ion selective electrode – Roche Integra (n = 13) Colorimetry – OCD Vitros (n = 22) Colorimetry – Siemens Dimension (n = 7) Colorimetry – Roche Cobas (n = 31) Colorimetry – Thermo Scientific Infinity (n = 11) All methods (n = 114)

Estimates

Med, mmol/L CV, % Med, mmol/L CV, % Med, mmol/L CV, % Med, mmol/L CV, % Med, mmol/L CV, % Med, mmol/L CV, % Med, mmol/L CV, % Med, mmol/L CV, %

Samples R/10038

R/10034

R/10036

R/10031

R/10039

R/10033

R/10037

R/10035

0.28 7.9 0.24 9.3 0.27 5.5 0.30 7.4 0.26 18.5 0.26 11.4 0.25 23.7 0.26 17.1

0.52 4.3 0.48 3.1 0.50 1.5 0.57 13.0 0.45 9.9 0.50 3.7 0.46 6.4 0.50 7.4

1.03 4.7 0.99 3.7 1.00 2.2 1.08 6.9 0.99 3.4 0.99 3.0 0.95 5.9 1.00 3.0

1.50 1.5 1.43 5.7 1.48 3.5 1.59 4.7 1.44 5.7 1.46 2.3 1.39 2.7 1.47 5.5

1.99 2.8 1.82 6.9 1.96 2.6 2.05 3.6 1.91 2.3 1.93 3.5 1.85 2.6 1.94 3.8

2.99 2.4 2.92 8.1 2.93 2.8 3.10 3.1 2.89 1.9 2.81 5.0 2.70 3.8 2.90 5.1

3.98 1.4 3.71 15.4 3.91 1.9 4.00 8.4 3.80 0.9 3.90 5.7 3.54 8.1 3.90 5.5

5.96 2.4 5.66 17.2 5.69 2.9 6.20 4.8 5.70 4.3 5.66 5.2 5.14 7.1 5.75 7.3

Med, median; CV, coefficient of variation. 16 laboratories were not evaluated because of the low number of method users (n b 6).

evaluate the analytical performance of lithium methods used by Belgian laboratories, and to investigate acceptability of their clinical interpretations. Materials and Methods Sample preparation and distribution Fresh blood donations were obtained from one patient, from the dialysis center of the Jules Bordet Institute (Brussels, Belgium). Blood donations were centrifuged at 4000 ×g for 10 minutes at 4 °C after blood clotting. Collected serum was assayed for lithium at the Medical Centre for Family Physicians (Leuven, Belgium) to ensure absence of the analyte, and thereby enable serum pooling. The standard solution of lithium (Merck, Darmstadt, Germany) was prepared by dissolving 77.58 mg of lithium carbonate in 350 mL of the final serum pool in order to reach a resultant concentration of 6 mmol/L. The standard solution was then successively diluted to obtain the following lithium concentrations: 4.0, 3.0, 2.0, 1.5, 1.0, 0.5 and 0.25 mmol/L. Each portion was aliquoted (500 μL) into 1.5 mL screwtopped vials with ‘o’ rings, and stored at −20 °C. Serum portions were labelled as samples: R/10032 (0.0 mmol/L), R/10038 (0.25 mmol/L), R/10034 (0.5 mmol/L), R/10036 (1.0 mmol/L), R/10031 (1.5 mmol/L), R/10039 (2.0 mmol/L), R/10033 (3.0 mmol/L), R/10037 (4.0 mmol/L), and R/10035 (6.0 mmol/L). The appropriate numbers of aliquots of each portion were packed as part of the standard Belgian EQAS distribution. They were sent to the participant laboratories and received within two days of dispatch.

were evaluated by selecting methods used by 6 laboratories or more [15]. The coefficient of variation (CV) around the median was calculated as a measure of imprecision. The relative difference between the measured and the reference value was taken as a measure of bias. For each sample, the reference value was calculated as the median of reported values by participant laboratories that used the reference method (i.e. flame photometry with internal standard (IL 943)), considering one reported value by each participant laboratory. The overall bias of each method was assessed using the median relative bias of the method from the reference value. Comparison and agreement between the reference method and the other evaluated methods were assessed secondly using the leastsquares linear regression, in order to investigate possible systematic errors [17]. Potential outliers were detected using the Grubbs test and excluded from the statistical parametric analysis. The estimates of slope and intercept and their standard errors were then computed. A Student t-test was used to test for intercept = 0 and slope = 1, i.e. for possible constant and proportional systematic errors [17,18]. A final lack-of-fit test for linear regression was performed over the reported concentrations in lithium in order to investigate linearity of data for each method. P-values b 0.05 were considered to be statistically

Analysis and data processing Belgian EQAS laboratories were invited to store samples at 2–8 °C until analysis and to analyse them as soon as possible on arrival. They were requested to report within 2 weeks the assay technique, the lithium measurements and the interpretations with regard to the measured concentrations. Participants returned results by post, fax or through a secure webpage. After the closing date, statistics were performed in order to derive descriptive statistics (i.e. inter- and intramethod medians with standard deviations) based on the Tukey’s robust approach used for Belgian EQAS [15]. Based on the assumption that 25th and 75th percentiles are generally not influenced by outliers, this nonparametric approach has the advantage to be more realistic in the interpretation of EQAS results (small number of method users, non-Gaussian distribution) [15,16]. Measures of central tendency and dispersion

Fig. 1. Relative bias of lithium assays used by Belgian EQAS participants, from the reference method (i.e. flame photometry with internal standard (IL 943)).

I.K. Delattre et al. / Clinical Biochemistry 48 (2015) 617–621

619

Table 2 Linear regression estimates from method comparison of lithium measurements: flame photometry with internal standard (IL 943) as reference method (x) and routine assays (y). Methods

Atomic absorption spectroscopy Ion selective electrode – Roche Integra Colorimetry – OCD Vitros Colorimetry – Siemens Dimension Colorimetry – Roche Cobas Colorimetry – Thermo Scientific Infinity

Least-squares regression Parameters

Coefficients

SD errors

95% CI

P-value

Intercept Slope Intercept Slope Intercept Slope Intercept Slope Intercept Slope Intercept Slope

0.0529 0.9017 0.0159 0.9705 −0.0006 1.0356 −0.0292 0.9749 0.0182 0.9460 0.0651 0.8569

0.0729 0.0250 0.0241 0.0087 0.0179 0.0062 0.0184 0.0065 0.0157 0.0057 0.0308 0.0105

−0.0938 to 0.1996 0.8514 to 0.9521 −0.0319 to 0.0637 0.9532 to 0.9877 −0.0359 to 0.0347 1.023 to 1.0479 −0.0660 to 0.0076 0.9619 to 0.9879 −0.0127 to 0.0492 0.9348 to 0.9571 0.0038 to 0.1264 0.8371 to 0.8787

0.4718 b0.0001 0.5112 b0.0001 0.9729 b0.0001 0.1175 b0.0001 0.2476 b0.0001 0.0376 b0.0001

SD errors, standard errors; 95% CI, 95% confidence intervals.

significant. Statistical analyses were performed using JMP Statistical DiscoveryTM 10.0.2 software (SAS Institute Inc., Cary, NC, USA). A consensus between laboratories was considered when N 70% of laboratories agreed in the result interpretation. Results 114 laboratories participated to the present Belgian EQAS for lithium. Overall, participants returned assay results using colorimetry (69.3%), ion selective electrode (15.8%), flame photometry (8.8%), atomic absorption spectroscopy (5.2%) and mass spectrometry (0.9%). 16 laboratories were not included in the main statistical analysis, due to the low number of method users (n b 6). In addition, no statistics were described for the blank sample R/10032 because of the large number of left-censored data (64.0%), i.e. values lower than the

quantification limit. Among 36.0% of laboratories that reported numerical values for that sample R/10032, 16.6% detected residual lithium concentrations, ranging from 0.01 to 0.21 mmol/L, with an abnormal value at 2.57 mmol/L. Medians and variations assessed for each assay category of the study are given in Table 1. The Roche Cobas assay and the Vitros assay were the most widely used techniques (46.5%) for lithium measurements. Intramethod imprecision for the low quality control varied from 5.5% to 23.7% and from 2.4% to 17.2% for the high quality control. Overall, the lowest imprecisions were reported for the reference method and the Roche Integra assay, averaging 2.6% and 2.7% respectively. Lithium concentrations were systematically higher when measured with the Vitros assay. As illustrated in Fig. 1, this method showed a marked positive bias, averaging 4.0%. In contrast, the Thermo Scientific Infinity assay showed a negative bias, averaging −9.4% (Fig. 1). Relative biases

Fig. 2. Regression plots between lithium concentrations: flame photometry with internal standard (IL 943) as reference method (x) and routine assays (y). (R2, coefficient of regression; p-value, p-value from the lack-of-fit test).

620

I.K. Delattre et al. / Clinical Biochemistry 48 (2015) 617–621

Table 3 Interpretations reported by evaluated Belgian laboratories in the EQAS for lithium. Samples [Reference values]

R/10032 [0.00 mmol/L]

R/10038 [0.28 mmol/L]

R/10034 [0.52 mmol/L]

R/10036 [1.03 mmol/L]

R/10031 [1.50 mmol/L]

R/10039 [1.99 mmol/L]

R/10033 [2.99 mmol/L]

R/10037 [3.98 mmol/L]

R/10035 [5.96 mmol/L]

Subtherapeutic Therapeutic Potentially toxic Severely toxic No interpretation

93 – – – 5

91 1 – – 6

56 37b – – 5

– 91 1 – 6

– 10 81 – 7

– – 69 23c 6

– 1 7 84 6

– – 1 92 5

2a – – 90 6

a b c

These interpretations were associated with abnormal low concentrations in lithium (0.1 and 0.3 mmol/L). 2 users of the Vitros assay (n = 22) reported the “therapeutic” diagnosis. 9 users of the Vitros assay (n = 22) reported the “severely toxic” diagnosis.

reported for the Roche Integra group mostly were the smallest, ranging from −1.0% to −4.1%. One outlying laboratory using the Roche Cobas assay was excluded from the parametric statistical analysis. Three values were also flagged as outliers and removed: 0.3 mmol/L (sample R/10035, Roche Cobas group), 0.7 mmol/L (sample R/10034, Vitros assay) and 0.77 mmol/L (sample R/10036, Thermo Scientific Infinity assay). Results from the least-squares linear regression are depicted in Table 2. Unlike other methods, the intercept of the Thermo Scientific Infinity assay was statistically tested different from 0 (p-value b 0.05). All methods showed a slope statistically different from 1. Correlation coefficients exceeded 0.99 for most methods. A significant linear trend were observed for all methods (pvalue N 0.05), except for the Roche Cobas group (p = 0.0304). Regression plots are given in Fig. 2. Interpretations of EQAS results are listed in Table 3. A clear consensus statement (N 89.0%) was reached by participants in the result interpretation, except for samples R/10034 and R/10039. For the sample R/10034 (reference value: 0.52 mmol/L), 60.2% of responding laboratories interpreted the lithium value as “subtherapeutic”, while 39.8% disagreed with this statement, interpreting the result as “therapeutic”. For that sample, more than half of the Vitros users (54.5%) reported “therapeutic”; their reported values averaged 0.6 mmol/L. For the sample R/10039 (reference value: 1.99 mmol/L), 75.0% of responding laboratories interpreted the lithium value as “potentially toxic”, and 25.0% as “severely toxic”. For that sample, 40.9% of the Vitros users gave the “severely toxic” diagnosis. Conclusions TDM is essential for optimal use and dosing of lithium. Levels of the drug have to be constantly monitored because of its potential toxicity. To ensure the reliability of the drug measurements, regular participation in an EQAS is recommended. Originally promoted as educational activity, this programme facilitates comparison of various analytical techniques and identification of problems in the assay performance. In Belgium, EQAS for lithium TDM has been developed over more than 5 years. To our knowledge, no study has been reported on the quality of lithium analysis based on EQAS data. The current study was designed to assess the performance of various commercially available lithium assays used by Belgian laboratories, with regard to precision, accuracy and linearity. In addition, the study sought to evaluate agreement in clinical interpretations of lithium measurements between Belgian participating laboratories in the lithium EQAS. In general, the Roche Integra assay showed the best performance in terms of variability (CV ranging from 1.5% to 5.5%) and bias (relative biases ranging from − 1.0% to − 4.1%) (Table 1, Figs. 1 and 2). Low intramethod variations (averaging 2.6%) were also reported for the reference method (i.e. flame photometry with internal standard (IL 943)). One routine method, the Vitros assay, consistently reported higher values and could be associated with consecutive biased interpretations (with regard to the consensus statement) for both samples around the therapeutic and toxic limits (Table 3).

Interference of the matrix (i.e. dialysed blood from one patient) may however affect one or various lithium assays, leading to the misinterpretation of values. Therefore, the present results need to be confirmed using patient samples without any dialysis. To assist in assessing the statistical significance of descriptive data, a regression analysis was applied. Precisely, a linear least-squares regression was performed to compare and evaluate agreement between methods. These statistics allow to assess the systematic error and to judge the method acceptability [17]. Based on that approach, the regression of the test method to the reference method should yield a straight line nonsignificantly different from the equality line (intercept = 0 and slope = 1). A deviation from the equality line indicates a lack of agreement between the two comparative methods [19]. Overall, estimates of the slope and the intercept were reliable (correlation coefficients N 0.99), except for the atomic absorption photometric assay [17]. This could be easily explained by the limited number of laboratories using this technique (n = 6). As depicted in Table 2, only the Thermo Scientific Infinity assay showed a significant lack of agreement with the reference method, displaying an intercept and a slope statistically different from 0 and 1 respectively. Such deviations actually reveal constant and proportional systematic errors between the comparative methods. While constant systematic errors could usually be explained by some type of inference in the assay, inadequate blanking or a misset zero calibration point, proportional systematic errors are often caused by poor standardization or calibration (rarely by a matrix effect) [17]. The relative bias of the Thermo Scientific Infinity assay averaged −9.4% (Fig. 1). Despite the lack of agreement between the comparative methods, the Thermo Scientific Infinity users reported adequate interpretations with regard to the consensus statement. Samples spiked with other lithium concentrations could obviously lead to biased interpretations for these method users, as observed for Vitros users. A major concern of the least-squares regression is the statistical assumption on independent variables X, which should be measured without error (i.e. the analytical standard deviation is zero) [19]. This assumption is necessary to obtain unbiased estimates of the intercept and the slope, but it is rarely fulfilled. Every method in clinical chemistry is subject to some random measurement errors. Alternative regression techniques have been suggested to overcome the limitation, such as the Deming method or the Passing-Bablok method [20]. Although advantageous, these regression techniques were not appropriate in the present study because of the unpaired number of method users. The present study is the first evaluation of the EQAS data for lithium TDM. Results revealed assay-related differences in lithium measurements and their interpretations. Users of the Vitros assay reported systematically higher lithium concentrations with consecutive biased interpretations (with regard to the consensus statement). In contrast, users of the Thermo Scientific Infinity method reported significant lower levels. Clinicians should be aware of the dependent-method decision limits for lithium TDM. Moreover, in Belgium, no interpretation of the lithium levels is standardized. In general, a lithium concentration b 0.8 mmol is considered as subtherapeutic, between 0.8 and 1.2 mmol/L as therapeutic, between 1.2 and 2.0 as potentially toxic and N 2.0 as severely toxic. A clear standardization needs

I.K. Delattre et al. / Clinical Biochemistry 48 (2015) 617–621

therefore to be defined among Belgian laboratories in order to improve the lithium monitoring. References [1] Grunze H, Vieta E, Goodwin GM, Bowden C, Licht RW, Moller HJ, et al. The World Federation of Societies of Biological Psychiatry (WFSBP) guidelines for the biological treatment of bipolar disorders: update 2012 on the long-term treatment of bipolar disorder. World J Biol Psychiatry 2013;14(3):154–219. [2] Yatham LN, Kennedy SH, Parikh SV, Schaffer A, Beaulieu S, Alda M, et al. Canadian Network for Mood and Anxiety Treatments (CANMAT) and International Society for Bipolar Disorders (ISBD) collaborative update of CANMAT guidelines for the management of patients with bipolar disorder: update 2013. Bipolar Disord 2013; 15(1):1–44. [3] Suppes T, Dennehy EB, Hirschfeld RM, Altshuler LL, Bowden CL, Calabrese JR, et al. The Texas implementation of medication algorithms: update to the algorithms for treatment of bipolar I disorder. J Clin Psychiatry 2005;66(7):870–86. [4] Geddes JR, Burgess S, Hawton K, Jamison K, Goodwin GM. Long-term lithium therapy for bipolar disorder: systematic review and meta-analysis of randomized controlled trials. Am J Psychiatry 2004;161(2):217–22. [5] Geddes JR, Goodwin GM, Rendell J, Azorin JM, Cipriani A, Ostacher MJ, et al. Lithium plus valproate combination therapy versus monotherapy for relapse prevention in bipolar I disorder (BALANCE): a randomised open-label trial. Lancet 2010; 375(9712):385–95. [6] Cipriani A, Pretty H, Hawton K, Geddes JR. Lithium in the prevention of suicidal behavior and all-cause mortality in patients with mood disorders: a systematic review of randomized trials. Am J Psychiatry 2005;162(10):1805–19. [7] Grandjean EM, Aubry JM. Lithium: updated human knowledge using an evidencebased approach: Part I: Clinical efficacy in bipolar disorder. CNS Drugs 2009;23(3): 225–40. [8] McKnight RF, Adida M, Budge K, Stockton S, Goodwin GM, Geddes JR. Lithium toxicity profile: a systematic review and meta-analysis. Lancet 2012;379(9817):721–8.

621

[9] Dasgupta N. Introduction to therapeutic drug monitoring. In: Dasgupta N, editor. Handbook of drug monitoring methods: therapeutics and drugs of abuse. New Jersey: Humana Press; 2008. p. 1–40. [10] Ratanajamit C, Soorapan S, Doang-ngern T, Waenwaisart W, Suwanchavalit L, Suwansiri S, et al. Appropriateness of therapeutic drug monitoring for lithium. J Med Assoc Thai 2006;89(11):1954–60. [11] Bettinger TL, Crismon ML. Lithium. In: Burton ME, editor. Applied pharmacokinetics and pharmacodynamics: principles of therapeutic drug monitoring. Lippincott Williams & Wilkins; 2006. p. 798–812. [12] Grandjean EM, Aubry JM. Lithium: updated human knowledge using an evidencebased approach. Part II: Clinical pharmacology and therapeutic monitoring. CNS Drugs 2009;23(4):331–49. [13] Dasgupta N, Datta P. Analytical techniques for measuring concentrations of therapeutic drugs in biological fluids. In: Dasgupta N, editor. Handbook of drug monitoring methods: therapeutics and drugs of abuse. New Jersey: Humana Press; 2008. p. 67–86. [14] Gruson D, Lallali A, Furlan V, Taburet AM, Legrand A, Conti M. Evaluation of a new lithium colorimetric assay performed on the dade behring dimension X-pand system. Clin Chem Lab Med 2004;42(9):1066–8. [15] Coucke W, China B, Delattre I, Lenga Y, Van Blerk M, Van Campenhout C, et al. Comparison of different approaches to evaluate external quality assessment data. Clin Chim Acta 2012;413(5–6):582–6. [16] Libeer JC. Quality assurance in Belgium. Ann Ist Super Sanita 1995;31(1):43–51. [17] Westgard QC. The comparison of methods experiment. http://www.westgard.com/ lesson23.htm; 2014. [Ref Type: Electronic Citation]. [18] Westgard QC. Z-14: Estimating analytical errors using regression statistics. http:// www.westgard.com/lesson44.htm; 2014. [Ref Type: Electronic Citation]. [19] Magari RT. Statistics for laboratory method comparison studies. BioPharm 2002; 28–32. [20] Linnet K. Evaluation of regression procedures for methods comparison studies. Clin Chem 1993;39(3):424–32.