Recently developed methods of longitudinal discriminant analysis allow for classification of subjects into prespecified prognostic groups using longitudinal history of both continuous and discrete biomarkers. The classification uses Bayesian estimates of the group membership probabilities for each prognostic group. These estimates are derived from a multivariate generalised linear mixed model of the biomarker's longitudinal evolution in each of the groups and can be updated each time new data is available for a patient, providing a dynamic (over time) allocation scheme. However, the precision of the estimated group probabilities differs for each patient and also over time. This precision can be assessed by looking at credible intervals for the group membership probabilities. In this paper, we propose a new allocation rule that incorporates credible intervals for use in context of a dynamic longitudinal discriminant analysis and show that this can decrease the number of false positives in a prognostic test, improving the positive predictive value. We also establish that by leaving some patients unclassified for a certain period, the classification accuracy of those patients who are classified can be improved, giving increased confidence to clinicians in their decision making. Finally, we show that determining a stopping rule dynamically can be more accurate than specifying a set time point at which to decide on a patient's status. We illustrate our methodology using data from patients with epilepsy and show how patients who fail to achieve adequate seizure control are more accurately identified using credible intervals compared to existing methods.

Department of Biostatistics University of Liverpool Liverpool U K

Department of Eye and Vision Science University of Liverpool Liverpool U K

Department of Probability and Mathematical Statistics Faculty of Mathematics and Physics Charles University Prague Czech Republic

Zobrazit více v PubMed

Fieuws S, Verbeke G, Maes B, Van Renterghem Y. Predicting renal graft failure using multivariate longitudinal profiles. Biostatistics. 2008;9(3):419‐431. PubMed

Brant LJ, Sheng SL, Morrell CH, Verbeke GN, Lesaffre E, Carter HB. Screening for prostate cancer by using random‐effects models. J R Stat Soc Series B. 2003;166(1):51‐62.

Tomasko L, Helms RW, Snapinn SM. A discriminant analysis extension to mixed models. Stat Med. 1999;18(10):1249‐1260. PubMed

Wernecke KD, Kalb G, Schink T, Wegner B. A mixed model approach to discriminant analysis with longitudinal data. Biom J. 2004;46(2):246‐254.

Lix LM, Sajobi TT. Discriminant analysis for repeated measures data: a review. Front Psychol. 2010;1(146):1‐9. PubMed PMC

Kohlmann M, Held L, Grunert VP. Classification of therapy resistance based on longitudinal biomarker profiles. Biom J. 2009;51(4):610‐626. PubMed

Morrell CH, Brant LJ, Sheng S, Metter EJ. Screening for prostate cancer using multivariate mixed‐effects models. J Appl Stat. 2012;39(6):1151‐1175. PubMed PMC

Marshall G, De la Cruz‐Mesía R, Quintana FA, Barón AE. Discriminant analysis for longitudinal data with multiple continuous responses and possibly missing data. Biometrics. 2009;65(1):69‐80. PubMed

Komárek A, Hansen BE, Kuiper EMM, van Buuren HR, Lesaffre E. Discriminant analysis using a multivariate linear mixed model with a normal mixture in the random effects distribution. Stat Med. 2010;29(30):3267‐3283. PubMed

Hughes DM, Komárek A, Czanner G, Garcia‐Fiñana M. Dynamic longitudinal discriminant analysis using multiple longitudinal markers of different types. Stat Methods Med Res. 2016. https://doi.org/10.1177/0962280216674496. PubMed DOI PMC

Marson AG, Al‐Kharusi AM, Alwaidh M, et al. The SANAD study of effectiveness of valproate, lamotrigine, or topiramate for generalised and unclassifiable epilepsy: an unblinded randomised controlled trial. The Lancet. 2007;369(9566):1016‐1026. PubMed PMC

Marson AG, Al‐Kharusi AM, Alwaidh M, et al. The SANAD study of effectiveness of carbamazepine, gabapentin, lamotrigine, oxcarbazepine, or topiramate for treatment of partial epilepsy: an unblinded randomised controlled trial. The Lancet. 2007;369(9566):1000‐1015. PubMed PMC

Komárek A, Komárková L. Clustering for multivariate continuous and discrete longitudinal data. Ann Appl Stat. 2013;7(1):177‐200.

Guglielmi A, Ieva F, Paganoni AM, Ruggeri F, Soriano J. Semiparametric Bayesian models for clustering and classification in the presence of unbalanced in‐hospital survival. J R Stat Soc Series B (Applied Statistics). 2014;63(1):25‐46.

Zhang X, Jeske DR, Li J, Wong V. A sequential logistic regression classifier based on mixed effects with applications to longitudinal data. Comput Stat Data Anal. 2016;94:238‐249.

Horrocks J, van Den Heuvel MJ. Prediction of pregnancy: a joint model for longitudinal and binary data. Bayesian Anal. 2009;4(3):523‐538.

Shah NH, Hipwell AE, Stepp SD, Chang CCH. Measures of discrimination for latent group‐based trajectory models. J Appl Stat. 2015;42(1):1‐11. PubMed PMC

Morrell CH, Sheng SL, Brant LJ. A comparative study of approaches for predicting prostate cancer from longitudinal data. Commun Stat Simul Comput. 2011;40(9):1494‐1513.

Rizopoulos D. Joint Models for Longitudinal and Time‐to‐Event Data: With Applications in R. CRC Press, Boca Raton; 2012.

Reddy T, Molenberghs G, Njagi EN, Aerts M. A novel approach to estimation of the time to biomarker threshold: applications to HIV. Pharm Stat. 2016;15(6):541‐549. PubMed

Hansen BE, Komárek A, Buster EHCJ, Steyerberg EW, Janssen HLA, Lesaffre E. Dynamic prediction of response to HBV‐treatment using multivariate longitudinal profiles In: Hansen BE, ed. Statistical Models of Treatment Effects in Chronic Hepatitis B and C, chap. 2.4 Rotterdam: Erasmus Universiteit; 2010:79‐103.

Lukasiewicz E, Gorfine M, Neumann AU, Freedman LS. Combining longitudinal discriminant analysis and partial area under the ROC curve to predict non‐response to treatment for hepatitis c virus. Stat Methods Med Res. 2011;20(3):275‐289. PubMed

Youden WJ. Index for rating diagnostic tests. Cancer. 1950;3(1):32‐35. PubMed

Freeman EA, Moisen GG. A comparison of the performance of threshold criteria for binary classification in terms of predicted prevalence and kappa. Ecol Model. 2008;217(1):48‐58.

Robert CP. The Bayesian Choice: From Decision‐Theoretic Foundations to Computational Implementations. 2nd edn. New York: Springer Science+Business Media; 2007.

R Core Team . R: A language and environment for statistical computing. R Foundation for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. 2016. http://www.R-project.org/.

Komárek A, Komárková L. Capabilities of R package mixAK for clustering based on multivariate continuous and discrete longitudinal data. J Stat Software. 2014;59(12):1‐38.

Commenges D, Jacqmin‐Gadda H. Dynamical Biostatistical Models, Vol. 86. CRC Press, Boca Raton; 2015.

Rizopoulos D, Taylor JMG, Van Rosmalen J, Steyerberg EW, Takkenberg JJM. Personalized screening intervals for biomarkers using joint models for longitudinal and survival data. Biostatistics. 2015;17(1):149‐164. PubMed PMC

van Houwelingen H, Putter H. Dynamic Prediction in Clinical Survival Analysis. CRC Press, Boca Raton; 2011.

Hastie T, Tibshirani R, Friedman J. The Elements of Statistical Learning: Data Mining, Inference, and Prediction. 2nd edn. New York: Springer‐Verlag; 2009.

Luts J, Molenberghs G, Verbeke G, Van Huffel S, Suykens JAK. A mixed effects least squares support vector machine model for classification of longitudinal data. Comput Stat Data Anal. 2012;56(3):611‐628.

Silverman BW. Density Estimation for Statistics and Data Analysis, Vol. 26. CRC press, London; 1986.

Gelman A, Carlin JB, Stern HS, Rubin DB. Bayesian Data Analysis, Vol. 2. Boca Raton, FL, USA: Chapman & Hall/CRC; 2014.

Personalized risk-based screening for diabetic retinopathy: A multivariate approach versus the use of stratification rules

Identification of patients who will not achieve seizure remission within 5 years on AEDs