Extension of the SAEM algorithm for nonlinear mixed models with two levels of random effects

This article focuses on parameter estimation of multi-levels nonlinear mixed effects models (MNLMEMs). These models are used to analyze data presenting multiple hierarchical levels of grouping (cluster data, clinical trials with several observation p…

Authors: Xavi`ere Panhard, Adeline Samson (MAP5)

Extension of the SAEM algorithm for nonlinear mixed models with two   levels of random effects
Extension of the SAEM algorithm for nonlinear mixed mo dels with t w o lev els of random eets Xa vière P anhard 1 ∗ A deline Samson 2 1 INSERM U738, P aris, F rane; Univ ersit y P aris 7, UFR de Médeine, P aris, F rane 2 Lab oratoire MAP5, UMR CNRS 8145, Univ ersit y P aris 5, F rane * Corresp onding author: INSERM U738, Univ ersité P aris 7, UFR de Médeine, 16 rue Hu hard, 75018 P aris, F rane Abstrat This artile fo uses on parameter estimation of m ulti-lev els nonlinear mixed eets mo d- els (MNLMEMs). These mo dels are used to analyze data presen ting m ultiple hierar hial lev els of grouping (luster data, linial trials with sev eral observ ation p erio ds,...). The v ariabilit y of the individual parameters of the regression funtion is th us deomp osed as a b et w een-sub jet v ariabilit y and higher lev els of v ariabilit y (for example within-sub jet v ariabilit y). W e prop ose maxim um lik eliho o d estimates of parameters of those MNLMEMs with t w o lev els of random eets, using an extension of the SAEM-MCMC algorithm. The extended SAEM algorithm is split in to an expliit diret EM algorithm and a sto  hasti EM part. Compared to the original algorithm, additional suien t statistis ha v e to b e appro ximated b y relying on the onditional distribution of the seond lev el of random ef- fets. This estimation metho d is ev aluated on pharmaokineti ross-o v er sim ulated trials, mimi king theoph yllin onen tration data. Results obtained on those datasets with either the SAEM algorithm or the F OCE algorithm (implemen ted in the nlme funtion of R soft- w are) are ompared: biases and RMSEs of almost all the SAEM estimates are smaller than the F OCE ones. Finally , w e apply the extended SAEM algorithm to analyze the pharma- okineti in teration of tenofo vir on atazana vir, a no v el protease inhibitor, from the ANRS 107-Puzzle 2 study . A signian t derease of the area under the urv e of atazana vir is found in patien ts reeiving b oth treatmen ts. 1 KEYW ORDS: Multilev el nonlinear mixed eets mo dels; SAEM algorithm; Multiple p erio ds; Cross-o v er trial; Bio equiv alene trials. 1 In tro dution The use of non-linear mixed eets mo dels (NLMEMs) inreases in sev eral elds su h as agron- om y , forestry , linial trials, p opulation pharmaokinetis (PK) and pharmao dynamis or viral dynamis to mo del longitudinal data. In some settings, data an presen t m ultiple hierar hial lev els of grouping, leading to m ultiple nested lev els of v ariabilit y . F or instane, w e ma y study patien ts that are group ed in medial servies that are themselv es group ed in to hospitals. In this artile, w e all m ultilev el non-linear mixed eets mo dels (MNLMEMs) the mo dels that desrib e su h data. MNLMEMs represen t a natural extension of mo dels with single v ariabilit y lev el, and they ha v e reen tly b een sub jet to a great deal of atten tion in statistial literature. In the eld of forestry , Hall and Clutter ( 1) analyze longitudinal measures of yield and gro wth that are mea- sured on ea h tree within a plot. In the eld of agronom y , Rek a y a et al. ( 2) onsider milk yield data where ea h o w is observ ed longitudinally during its rst three latations. Jarézi et al. (3) p erform geneti analyses of gro wth measuremen ts in b eef attle a kno wledging the fat that sev eral o ws ome from the same sire. Another example is p opulation PK, where onen tration measuremen ts ma y b e tak en with sev eral patien ts o v er sev eral distint time in terv als, that are often named p erio ds or o asions. That grouping pattern is used for instane in PK ross-o v er trials. In NLMEMs with only one lev el of v ariabilit y , often orresp onding to b et w een-sub jet v ari- abilit y , the analysis results in the estimation of the xed eets parameters and of the b et w een- sub jet v ariabilit y of the parameters, also alled in ter-sub jet v ariabilit y . When there is more than one lev el of grouping, the higher lev els of v ariabilit y an b e estimated. In the sp ei ase where the seond lev el of grouping orresp onds to m ultiple p erio ds of measuremen t, this v ariabil- it y is alled within-sub jet v ariabilit y (or in tra-sub jet v ariabilit y , or in ter-o asion v ariabilit y), and orresp onds to the v ariation of the individual parameters aross the dieren t study p erio ds or units. In the on text of pharmaokinetis, Karlsson et al. ( 4 ) demonstrate the imp ortane of mo deling this seond lev el of v ariabilit y in t w o-lev els NLMEMs. They sho w that negleting it resulted in biased estimates for the xed eets. 2 The parameter estimation of NLMEMs is not trivial b eause the lik eliho o d of NLMEMs annot b e expressed in a losed form due to the non-linearit y of the regression funtion in the random eets. Therefore, sev eral estimation metho ds ha v e b een prop osed. The First Order Conditional Estimates (F OCE) algorithm p erforms a rst order linearization of the regression funtion with resp et to the random eets (5; 6). The implemen tation of the F OCE algorithm in the NONMEM soft w are and in the nlme funtion of Splus and R enables the estimation of b oth b et w een- and within-sub jet v ariabilities. F rom our pratie, the main dra wba k of this metho d is ho w ev er that it do es not alw a ys on v erge when one estimates sim ultaneously the b et w een- and the within-sub jet v ariabilities on sev eral parameters. F urthermore, this linearization-based metho d annot b e onsidered as fully established in theory . F or instane, V onesh ( 7) and Ge et al. (8) giv e examples of sp ei designs resulting in inonsisten t estimates, su h as when the n um b er of observ ations p er sub jet do es not inrease faster than the n um b er of sub jets or when the v ariabilit y of random eets is to o large. Sev eral estimation metho ds ha v e b een prop osed as alternativ es to linearization algorithms. A ommon metho d to handle n umerial in tegration is the adaptativ e Gaussian quadrature (A GQ) metho d. An estimation algorithm of NLMEM parameters based on this lassial A GQ metho d has b een prop osed b y Pinheiro and Bates (9) and is implemen ted in the SAS pro edure NLMIXED (10). Ho w ev er, the A GQ metho d requires a suien tly large n um b er of quadrature p oin ts implying an often slo w on v ergene with v ery high omputational time. F urthermore, t w o-lev els NLMEM an b e implemen ted in the NLMIXED pro edure, but to our kno wledge, the on v ergene is diult to obtain in pratie (3 ). Impro v emen ts up on this metho d are th us needed. The seond alternativ e to linearization is the use of the Exp etation-Maximization (EM) algorithm (11 ) in order to estimate mo dels with missing or non-observ ed data su h as random eets. Beause of the nonlinearit y of the mo del, sto  hasti v ersions of the EM algorithm ha v e b een prop osed. W ei et al. (12 ); W alk er (13 ) and W u ( 14 ) prop ose MCEM algorithms, with a Mon te-Carlo appro ximation of the E-step. Ho w ev er the MCEM algorithm ma y ha v e omputa- tional problems (i.e slo w or ev en non on v ergene). As an alternativ e to address omputational problems, a sto  hasti appro ximation v ersion of EM (SAEM) has b een prop osed in ( 15 ; 16 ), whi h requires the sim ulation of only one realization of the missing data for ea h iteration, sub- stan tially reduing the omputation time. Kuhn and La vielle (16 ) prop ose to om bine the SAEM algorithm with a Mon te-Carlo Mark o v Chain (MCMC) pro edure adapted to the NLMEMs, and 3 pro v e that the resulting estimates are on v ergen t and onsisten t. T o date, none of the EM-based algorithms are diretly appliable to the ase of m ultilev el NLMEMs and ha v e to b e adapted. The ob jetiv e of this pap er is to extend the SAEM algo- rithm to MNLMEMs with t w o lev els of v ariabilit y: b oth E and M steps need to b e adapted to in tegrate higher lev els of random eets. W e also prop ose estimates of the lik eliho o d and of the Fisher information matrix. W e ev aluate this algorithm on a PK example, more preisely a t w o-p erio ds one-sequene ross-o v er trials sim ulated mimi king theoph yllin onen tration data (9). W e also apply the SAEM algorithm to the PK in teration of t w o HIV moleules (tenofo vir and atazana vir) from a PK substudy of the ANRS 107-Puzzle 2 trial. After desribing the mo del and notations (setion 2), setion 3 desrib es the SAEM algorithm. Setion 4 rep orts the sim ulation study and its results. In Setion 5, w e study the PK in teration of tenofo vir on atazana vir in HIV patien ts. Setion 6 onludes the artile with some disussion. 2 Mo dels Let us denote y ij k the observ ation in unit k ( k = 1 , . . . , K ) for sub jet i ( i = 1 , · · · , n ) and at time t ij k ( j = 1 , · · · , n ik ) . F or instane, the dieren t units an b e the dieren t p erio ds in the ase of PK trials, or the dieren t paren ts in the ase of geneti analyses. W e assume, as a kno wn fat, t w o nonlinear funtions f and g su h that the t w o-lev els non-linear mixed eets mo del linking observ ations to sampling times an b e written as: y ij k = f ( t ij k , φ ik ) + g ( t ij k , φ ik ) ε ij k , ε ij k ∼ N (0 , σ 2 ) , where φ ik is the p -v etor of the parameters of sub jet i for unit k and ε ij k is the measuremen t error. W e h yp othesize that the errors ε ij k giv en φ ik are m utually indep enden t. W e assume that the individual parameters φ ik are random v etors and that for ea h unit k , φ ik an b e brok en 4 in to: φ ik = µ + β k + b i + c ik , (1) b i ∼ N (0 , Ω) , c ik ∼ N (0 , Ψ) , where µ + β k is the mean v alue for unit k , b i is the random eet of size p of sub jet i , and c ik is the random eet of size p of sub jet i and unit k . T o ensure the iden tiabilit y of the parameters, w e assume that β 1 = 0 , ie µ is the mean of the rst unit and β k represen ts the dierene (or eet) of the k th unit in omparison to this rst unit. The random eets ( b i ) and ( c ik ) are assumed to b e m utually indep enden t. The total v ariane of the parameters is th us brok en in to a b et w een-sub jet v ariane Ω and a within-sub jet v ariane Ψ . Finally , the individual parameters pK -v etor φ i = ( φ i 1 , . . . , φ iK ) of sub jet i is distributed with a Gaussian distribution with mean v etor ( µ, µ + β 2 , . . . , µ + β K ) and a pK × pK o v ariane matrix Γ equal to Γ =          Ω + Ψ Ω . . . Ω Ω Ω + Ψ . . . . . . . . . . . . . . . Ω Ω . . . Ω Ω + Ψ          . Let θ = ( µ, β , Ω , Ψ , σ 2 ) , the v etor of all the parameters of the mo del where β denotes the v etor of unit eet β = ( β 1 , . . . , β K ) . The aim of this pap er is to prop ose an estimation of θ b y maximizing the lik eliho o d of the observ ations y = ( y ij k ) ij k . Let us denote ˜ b i := µ + b i . The lik eliho o d of y an b e written as: p ( y ; θ ) = Z p ( y , φ, ˜ b ; θ ) d ( φ, ˜ b ) where p ( y , φ, ˜ b ; θ ) is the lik eliho o d of the omplete data ( y , φ, ˜ b ) , with φ = ( φ ik ) i =1 ,...,n,k =1 , ...,K and ˜ b = ( ˜ b 1 , . . . , ˜ b n ) . Beause of the nonlinearit y of the regression funtion f with resp et to the random eets φ ik , the lik eliho o d has no losed form. Therefore, the maximization of the lik eliho o d in θ , θ ∈ Θ , is a omplex problem. W e prop ose to use a sto  hasti v ersion of the EM algorithm, whi h is presen ted in detail in the next setion. 5 3 Estimation algorithm 3.1 The SAEM algorithm The EM algorithm in tro dued b y Dempster et al. ( 11 ) is a lassial approa h to estimate param- eters of mo dels with non-observ ed or inomplete data. In t w o-lev els NLMEMs, the non-observ ed data are the individual parameters ( φ, ˜ b ) and the omplete data of the mo del is ( y , φ, ˜ b ) . Let us denote L c ( y , φ, ˜ b ; θ ) = log p ( y , φ, ˜ b ; θ ) the log-lik eliho o d of the omplete data. The priniple of the iterativ e EM algorithm is to maximize the funtion Q ( θ | θ ′ ) = E ( L c ( y , φ, ˜ b ; θ ) | y ; θ ′ ) where the exp etation is the onditional exp etation under the p osterior distribution p ( φ, ˜ b | y ; θ ′ ) , the maximization of Q b eing often easier than the diret maximization of the observ ed data log- lik eliho o d. Ea h iteration of the EM algorithm is omputed through t w o steps: the Exp etation step (E-step) and the Maximization step (M-step). A t the ℓ th iteration of the algorithm, the E step is the ev aluation of Q ( θ | b θ ℓ ) , while the M step up dates b θ ℓ b y maximizing Q ( θ | b θ ℓ ) . Let us sho w that the funtion Q an b e redued in the ase of a MNLMEM. First, let us note that as p ( ˜ b | y , φ ; θ ) = p ( ˜ b | φ ; θ ) , b y appliation of the Ba y es theorem w e ha v e: p ( φ, ˜ b | y ; θ ) = p ( ˜ b | φ ; θ ) p ( φ | y ; θ ) = n Y i =1 p ( ˜ b i | φ i ; θ ) p ( φ i | y i ; θ ) . (2) Seond, through the linearit y of the individual parameters mo del in equation ( 1), the p oste- rior distribution p ( ˜ b i | φ i ; θ ) of the i th sub jet is expliit: p ( ˜ b i | φ i ; θ ) is a Gaussian distribution N ( m ( φ i , θ ) , V ( θ )) of mean and v ariane equal to: m ( φ i , θ ) = V ( θ ) Ψ − 1 K X k =1 ( φ ik − β k ) + Ω − 1 µ ! , (3) V ( θ ) = (Ω − 1 + K Ψ − 1 ) − 1 . Due to the fatorization giv en in equation (2), funtion Q an b e rewritten as: Q ( θ | θ ′ ) = Z  Z L c ( y , φ, ˜ b ; θ ) p ( ˜ b | φ ; θ ′ ) d ˜ b  p ( φ | y ; θ ′ ) dφ. Beause of the expliit p osterior distribution of random eets ˜ b giv en in equation (3), the omputation of this onditional exp etation an b e split in to t w o parts : the omputation of 6 the in tegral with resp et to the p osterior distribution of ˜ b whi h has an analytial form, and the omputation of the in tegral with resp et to the p osterior distribution of φ whi h has no analytial form. Therefore the EM algorithm is split in to an expliit diret EM algorithm for the omputation of the rst in tegral and the use of a sto  hasti v ersion of the EM algorithm for the omputation of the seond in tegral. Let us detail the expliit omputation of the rst in tegral, denoted b y R ( y , φ, θ , θ ′ ) R ( y , φ, θ , θ ′ ) = Z L c ( y , φ, ˜ b ; θ ) p ( ˜ b | φ ; θ ′ ) d ˜ b. This in tegral has an analytial form. Indeed, the omplete log lik eliho o d L c ( y , φ, ˜ b ; θ ) is equal to L c ( y , φ, ˜ b ; θ ) = − 1 2 K X k =1 n X i =1 n ik X j =1 log(2 π σ 2 g 2 ( t ij k , φ ik )) − 1 2 K X k =1 n X i =1 n ik X j =1 ( y ij k − f ( t ij k , φ ik )) 2 σ 2 g 2 ( t ij k , φ ik ) − nK 2 log(2 π det Ψ) − 1 2 K X k =1 n X i =1 ( φ ik − ˜ b i − β k ) t Ψ − 1 ( φ ik − ˜ b i − β k ) − n 2 log(2 π det Ω) − 1 2 n X i =1 ( ˜ b i − µ ) t Ω − 1 ( ˜ b i − µ ) . As the p osterior distribution p ( ˜ b | φ ; θ ) is kno wn (equation 3 ), R ( y , φ, θ , θ ′ ) is equal to R ( y , φ, θ , θ ′ ) = − 1 2 X i,j,k log(2 π σ 2 g 2 ( t ij k , φ ik )) 1 2 X i,j,k ( y ij k − f ( t ij k , φ ik )) 2 σ 2 g 2 ( t ij k , φ ik ) (4) − nK 2 log(2 π det Ψ) − nK 2 Ψ − 1 / 2 V ( θ ′ )Ψ − 1 / 2 − 1 2 X i,k ( φ ik − m ( φ i , θ ′ ) − β k ) t Ψ − 1 ( φ ik − m ( φ i , θ ′ ) − β k ) − n 2 log(2 π det Ω) − n 2 Ω − 1 / 2 V ( θ ′ )Ω − 1 / 2 − 1 2 n X i =1 ( m ( φ i , θ ′ ) − µ ) t Ω − 1 ( m ( φ i , θ ′ ) − µ ) . Therefore Q is redued to the omputation of the seond in tegral under the p osterior distri- bution p ( φ | y ; θ ) as follo ws: Q ( θ | θ ′ ) = Z R ( y , φ, θ , θ ′ ) p ( φ | y ; θ ′ ) dφ. (5) Giv en the non-linearit y of funtion f with resp et to φ , the p osterior distribution p ( φ | y ; θ ′ ) has no losed form and the funtion Q dened b y (5) is in tratable. Th us w e prop ose to use the 7 sto  hasti v ersion SAEM of the EM algorithm prop osed b y Dely on et al. (15 ) whi h ev aluates the in tegral Q b y a sto  hasti appro ximation pro edure. Let us detail this SAEM algorithm in the ase of t w o-lev els NLMEMs. Let us note that the quan tit y R ( y , φ, θ , θ ′ ) b elongs to the regular urv ed exp onen tial family , i.e, it an b e written as R ( y , φ, θ , θ ′ ) = − Λ( θ ) + h S ( y , φ, θ ′ ) , Φ( θ ) i , (6) where h ., . i is the salar pro dut, Λ and Φ are t w o funtions t wie on tin uously dieren tiable on Θ and S ( y , φ, θ ′ ) is kno wn as the minimal suien t statistis of the omplete mo del. Those statistis are detailed later. In this ase, the Q funtion is redued to Q ( θ | θ ′ ) = − Λ( θ ) + h  Z S ( y , φ, θ ′ ) p ( φ | y ; θ ′ ) dφ  , Φ( θ ) i , In this ase, at the ℓ th iteration, the SAEM algorithm pro eeds as follo ws: • Sim ulation step: sim ulation of the missing data ( φ ( ℓ ) i ) i under the onditional distribution p ( φ | y ; b θ ℓ ) • Sto  hasti appro ximation step: omputation of a sto  hasti appro ximation s ℓ +1 of E h S ( y , φ, b θ ℓ ) | y ; b θ ℓ i = R S ( y , φ, b θ ℓ ) p ( φ | y ; b θ ℓ ) dφ , using ( γ ℓ ) ℓ ≥ 0 , a sequene of p ositiv e n um b ers dereasing to 0: s ℓ +1 = s ℓ + γ ℓ ( S ( y , φ ( ℓ ) , b θ ℓ ) − s ℓ ) . • Maximization step: up date of the estimate b θ ℓ +1 : b θ ℓ +1 = ar g max θ ∈ Θ ( − Λ( θ )) + h s ℓ +1 , Φ( θ )) i ) . The suien t statistis of the omplete mo del (4) ev aluated during the SA step of the SAEM 8 algorithm are as follo ws: s 1 ,i,ℓ +1 = s 1 ,i,ℓ + γ ℓ K X k =1 φ ( ℓ ) ik − s 1 ,i,ℓ ! , i = 1 , . . . , N , s 2 ,k,ℓ +1 = s 2 ,k,ℓ + γ ℓ n X i =1 φ ( ℓ ) ik − s 2 ,k,ℓ ! , k = 1 , . . . , K, s 3 ,ℓ +1 = s 3 ,ℓ + γ ℓ n X i =1 m ( φ ( ℓ ) i , b θ ℓ ) t m ( φ ( ℓ ) i , b θ ℓ ) − s 3 ,ℓ ! , s 4 ,ℓ +1 = s 4 ,ℓ + γ ℓ K X k =1 n X i =1  φ ( ℓ ) ik − m ( φ ( ℓ ) i , b θ ℓ )  t  φ ( ℓ ) ik − m ( φ ( ℓ ) i , b θ ℓ )  − s 4 ,ℓ ! , s 5 ,ℓ +1 = s 5 ,ℓ + γ ℓ   X i,j,k y ij k − f ( t ij k , φ ( ℓ ) ik ) g ( t ij k , φ ( ℓ ) ik ) ! 2 − s 5 ,ℓ   , The expression of the M step is obtained b y deriv ation of equation (4). The parameter estimates are as follo ws: b µ ℓ +1 = V ( b θ ℓ ) b Ψ − 1 ℓ 1 n n X i =1 s 1 ,i,ℓ +1 − K X k =1 b β k,ℓ ! + V ( b θ ℓ ) b Ω − 1 ℓ b µ ℓ , b β k,ℓ +1 = s 2 ,k,ℓ +1 n − b µ ℓ +1 , for k = 2 , . . . , K , b Ω ℓ +1 = V ( b θ ℓ ) + s 3 ,ℓ +1 n − ( b µ ℓ +1 ) t b µ ℓ +1 , b Ψ ℓ +1 = V ( b θ ℓ ) + s 4 ,ℓ +1 nK − 1 K K X k =1 ( b β k,ℓ +1 ) t b β k,ℓ +1 , c σ 2 ℓ +1 = s 5 ,ℓ +1 P n i =1 P K k =1 n ik . Comparing with the lassi SAEM algorithm for single-lev el NLMEMs, the extension of SAEM to the t w o-lev els NLMEMs is nally split in to an expliit EM algorithm and a sto  hasti EM part. F urthermore, it requires the omputation of t w o in termediate quan tities (the ondi- tional exp etations m ( φ i , θ ) and v ariane V ( θ ) of the b et w een-sub jet random eets parameters b i ) as w ell as t w o additional suien t statistis ( S 3 and S 4 ), funtions of m ( φ i , θ ) . The M-step diers from the one of the lassi SAEM for single-lev el NLMEMs, esp eially for the estimation of the v ariane matrix Ω and Ψ whi h uses the additional quan tit y V ( θ ) . As pro v ed b y (15 ; 17 ), the on v ergene of the SAEM algorithm is ensured under the follo wing assumption: 9 Assumption ( A1 ): 1. F untions Λ and Φ are t wie on tin uously dieren tiable on Θ . 2. The log-lik eliho o d log p ( y ; θ ) is d times dieren tiable on Θ , where d is the dimension of S ( y , φ, θ ′ ) . 3. F untion ¯ s dened as ¯ s ( θ, θ ′ ) = Z S ( y , φ, θ ′ ) p ( φ | y ; θ ) dφ is on tin uously dieren tiable on Θ with resp et to its rst v ariable. 4. F or all ℓ in N , γ ℓ ∈ [0 , 1] , P ∞ ℓ =1 γ ℓ = ∞ and P ∞ ℓ =1 γ 2 ℓ < ∞ . F or a on v enien t step sizes sequene γ ℓ , the assumption ( A1 ) is trivially  he k ed in our mo del. A  hoie of step sizes sequene γ ℓ is presen ted in Setion 4 . Ho w ev er, the sim ulation step of the SAEM algorithm, whi h p erforms the sim ulation of the non-observ ed v etor φ under the p osterior distribution p ( φ | y ; θ ) annot b e diretly p erformed b eause the p osterior distribution is only kno wn up to a onstan t. In this ase, Kuhn and La vielle (16 ) prop ose to om bine the SAEM algorithm with a Mark o v Chain Mon te Carlo (MCMC) pro edure for the sim ulation step. This v ersion of the SAEM-MCMC algorithm an b e used for the estimation of MNLMEMs. The MCMC pro edure used in this ase is detailed in setion 3.2. As pro v ed b y (17), the on v ergene of the SAEM-MCMC algorithm is ensured under assumption ( A1 ) and the follo wing additional assumption: Assumption ( A2 ): F or an y θ in Θ , w e assume that the onditional distribution p ( . | y ; θ ) is the unique limiting distribution of a transistion probabilit y Π θ , that has the follo wing prop erties: 1. F or an y ompat subset V of Θ , there exists a real onstan t L su h that for an y ( θ, θ ′ ) in V 2 sup { φ,φ ′ }∈E | Π θ ( φ ′ | φ ) − Π θ ′ ( φ ′ | φ ) | ≤ L k θ − θ ′ k . 2. The transition probabilit y Π θ supplies an uniformly ergo di  hain whose in v arian t proba- bilit y is the onditional distribution p ( φ | y ; θ ) , i.e. ∃ K θ ∈ R + , ∃ ρ θ ∈ ]0 , 1 [ | ∀ ℓ ∈ N k Π ℓ θ ( ·| φ ) − p ( · , ·| y ; θ ) k T V ≤ C θ ρ ℓ θ , 10 where k · k T V is the total v ariation norm. F urthermore, C = sup θ ∈ Θ C θ < ∞ and ρ = sup θ ∈ Θ ρ θ < 1 . 3. F untion S is b ound on E . A t iteration ℓ , the S-step of the SAEM-MCMC algorithm onsists th us in sim ulating φ ( ℓ ) with the transistion probabilit y Π ˆ θ ℓ  φ ( ℓ − 1) | dφ ( ℓ )  . The assumption ( A2 .2) is the most deliate to  he k, and it dep ends on the  hoie of the MCMC algorithm. This is detailed in the next subsetion after presen ting the MCMC pro edure. In pratie, the SAEM algorithm b eing a sto  hasti algorithm, there exists no deterministi on v ergene riterion whi h ould b e used to stop the iterations of the algorithm as so on as the on v ergene is rea hed. Therefore, w e reommend to implemen t the SAEM algorithm with a suien tly large n um b er of iterations and to graphially  he k the on v ergene b y plotting the v alues of the SAEM estimates obtained along iterations v ersus the iterations. Su h a gure is desrib ed in Setion 4. 3.2 MCMC algorithm for the sim ulation step Let us detail the sim ulation step of the SAEM-MCMC algorithm, whi h p erforms the sim ula- tion of the missing data φ through a Mark o v  hain whi h has p ( φ | y ; θ ) as unique stationary distribution. F or sub jet i , b y Ba y es form ula, this onditional distribution is prop ortional to p ( φ i | y i ; θ ) ∝ K Y k =1 p ( y ik | φ ik ; θ ) p ( φ i ; θ ) . W e prop ose to use a Metrop olis-Hastings (M-H) algorithm to sim ulate this Mark o v  hain. Let us reall the priniple of this algorithm. A t iteration r of the M-H algorithm, giv en the urren t v alue φ ( r ) i of the Mark o v Chain, the M-H algorithm pro eeds as follo ws: 1. Sim ulate φ c i with a prop osal distribution q ( · , φ ( r ) i ) 2. Compute the aeptane probabilit y α ( φ c i , φ ( r ) i ) = Q K k =1 p ( y ik | φ c ik ; θ ) p ( φ c i ; θ ) Q K k =1 p ( y ik | φ ( r ) ik ; θ ) p ( φ ( r ) i ; θ ) q ( φ c i , φ ( r ) i ) q ( φ ( r ) i , φ c i ) 11 3. Sim ulate u with a uniform distribution U [0 , 1] 4. Up date the Mark o v  hain with φ ( r +1) i =      φ c i if u ≤ α ( φ c i , φ ( r ) i ) φ ( r ) i else The on v ergene of the M-H algorithm strongly dep ends on the  hoie of the prop osal distribu- tion q . The on v ergene is ensured for some prop osal distributions su h as indep enden t ( q ( · , φ ( r ) i ) indep enden t of φ ( r ) i ) or symmetrial ( q ( · , φ ( r ) i ) = q ( φ ( r ) i , · ) ) prop osals (18 ). These prop osals are detailed b elo w. Giv en the dimension of φ , w e also onsider a Metrop olis-Hastings-Within-Gibbs algorithm, om bining b oth Gibbs algorithm and M-H pro edure. The adv an tage of the Gibbs algorithm is to redue the m ulti-dimensional sim ulation problem to the suessiv e sim ulations of one-dimension v etors. Finally , at iteration ℓ of the SAEM algorithm, giv en the urren t estimate b θ ℓ , w e om bine the three follo wing prop osal transitions: 1. the prior distribution of φ i , that is the Gaussian distribution N ( b µ ℓ + b β ℓ , b Γ ℓ ) , orresp onding to an indep enden t M-H algorithm, 2. the m ultidimensional random w alks N ( φ ( ℓ − 1) i , ρ b Γ ℓ ) (symmetri prop osal), where ρ is a saling v alue  hosen to ensure a satisfatory aeptation rate, namely around 30% as prop osed in (19 ), 3. a suession of K p unidimensional Gaussian random w alks (symmetri prop osal), i.e ea h omp onen t of φ i is suessiv ely up dated, leading to a Metrop olis-Hastings-Within-Gibbs algorithm, where b Γ ℓ is equal to b Γ ℓ =          b Ω ℓ + b Ψ ℓ b Ω ℓ . . . b Ω ℓ b Ω ℓ b Ω ℓ + b Ψ ℓ . . . . . . . . . . . . . . . b Ω ℓ b Ω ℓ . . . b Ω ℓ b Ω ℓ + b Ψ ℓ          . Giv en the prop osal distributions, as previously detailed, and using the theoretial on v ergene results prop osed in (18), this h ybrid Gibbs algorithm on v erges and generates an uniformly 12 ergo di  hain with p ( φ | y ; θ ) as the stationary distribution. Consequen tly , b y applying the on- v ergene theorem of Kuhn and La vielle ( 16) and under assumptions (A1) and (A2), w e pro v e that the estimate sequene ( b θ ℓ ) ℓ ≥ 0 pro dued b y the extended SAEM algorithm on v erges to w ards a (lo al) maxim um of the lik eliho o d p ( y ; θ ) . In pratie, the on v ergene of the MCMC algorithm is diult to v erify . As in Ba y esian inferene, the only on v ergene riteria existing for MCMC pro edure are graphial riteria. W e ha v e to  he k if the estimate sequene explores a suieny large domain of the Mark o v  hain. A on v ergene gure is presen ted and ommen ted in setion 4 . 3.3 Estimation of the Fisher information matrix and the lik eliho o d T o p erform statistial tests su h as W ald test or lik eliho o d ratio test, w e prop ose estimators of the Fisher information matrix and the lik eliho o d, resp etiv ely . As the Fisher information matrix has no losed form in MNLMEMs, w e prop ose to appro ximate it b y the Fisher information matrix of the m ulti-lev el linear mixed mo del dedued from the MNLMEM after linearization of the funtion f around the onditional exp etation of the individual parameters ( E ( φ i | y ; ˆ θ ) , 1 ≤ i ≤ n ) . The omputation of this linearized Fisher information matrix is diret and do es not need an y appro ximation. The estimation of the lik eliho o d of the MNLEM is based on an Imp ortane Sampling pro- edure, as prop osed b y Samson et al. ( 20) for NLMEMs. The Imp ortane Sampling pro edure has b een in tro dued to appro ximate the in tegral of the lik eliho o d with a smaller v ariane than with other Mon te Carlo metho ds. In this ase, an estimate of the on tribution p ( y i ; θ ) of the individual i to the lik eliho o d is \ p ( y i ; θ ) = 1 T T X t =1 p ( y i , φ ( t ) i ; θ ) q i ( φ ( t ) i ) where ( φ ( t ) i ) t =1 ,...,T are sim ulated using the individual instrumen tal distribution q i . As an in- dividual instrumen tal distribution q i , w e prop ose a Gaussian appro ximation of the individual onditional p osterior distribution p ( φ i | y i ; θ ) . 13 4 Sim ulation study: a PK example 4.1 Sim ulation settings The ob jetiv e of this sim ulation study is to illustrate the main statistial prop erties of the extended SAEM algorithm (bias, ro ot mean square errors, group omparison tests) and to ompare them to the F OCE algorithm. W e do not use the A GQ algorithm, sine pro edure NLMIXED do es not sueed, in pratie, in estimating su h omplex v ariane mo dels. W e use the PK data of orally administered theoph yllin to dene the p opulation mo del for the sim ulation study . These data are lassial ones in p opulation pharmaokinetis, often used for soft w are ev aluation (21 ). W e assume that onen trations an b e desrib ed b y a one-ompartmen t mo del with rst order absorption and rst order elimination: f ( t, φ ) = D K a V K a − C l  e − C l V t − e − K a t  where D is the dose, V is the v olume of distribution, K a is the absorption rate onstan t and C l is the learane of the drug elimination. These parameters are p ositiv e and distributed aording to a log-normal distribution. Th us, φ has the follo wing omp onen ts: φ = (log V , log K a , log AU C ) , with AU C = D /C l . W e assume iden tial sampling times for all sub jets: for all i in 1 , . . . , n , k = 1 , 2 , t ij k = t j for j = 1 , . . . , J . A dditiv e Gaussian random eets are assumed for ea h parameter with a diagonal o v ariane matrix Ω and a a diagonal o v ariane matrix Ψ . Let ω 2 = ( ω 2 V , ω 2 K a , ω 2 AU C ) and ψ 2 = ( ψ 2 V , ψ 2 K a , ψ 2 AU C ) denote the v etor of the v arianes of the random eets. A om bined error mo del is assumed b y setting g ( t, φ ) = 1 + f ( t, φ ) . W e set the dose for all sub jets to the v alue of 4 mg. F or all the parameters, the v alues are those prop osed b y P anhard and Men tré ( 22 ): log V = − 0 . 73 , log K a = 0 . 39 and log AU C = 4 . 6 1 , ω 2 V = 0 . 01 , ω 2 K a = 0 . 04 , ω 2 AU C = 0 . 04 , ψ 2 V = 0 . 002 5 , ψ 2 K a = 0 . 01 , ψ 2 AU C = 0 . 01 and σ 2 = 0 . 01 . W e generate n = 24 and n = 40 total n um b ers of sub jets with J = 1 0 blo o d samples p er sub jet, tak en at 15 min utes, 30 min utes, 1, 2, 3.5, 5, 7, 9, 12 and 24 hours after dosing. The individual data of one sim ulated trial are displa y ed in Figure 1. [Figure 1 ab out here.℄ 14 4.2 Ev aluation of estimates Our aim is to ev aluate and ompare the estimates pro dued b y the extended SAEM algorithm with those pro dued b y the nlme funtion of the R soft w are. W e t the sim ulation mo del and ompute the relativ e bias and relativ e ro ot mean square error (RMSEs) for ea h omp onen t of θ from 1000 repliations of the t w o trials desrib ed b elo w ( n = 24 and n = 40 total n um b er of sub jets). The sim ulation mo del inludes a treatmen t eet on all omp onen ts of θ . W e test the n ull h yp othesis { β log AU C = 0 } using the W ald test. W e also t the mo del where the treatmen t eet on log AU C is not estimated, and test the same n ull h yp othesis using the Lik eliho o d Ratio T est (LR T). The SAEM algorithm is implemen ted with 500 iterations. During the rst 200 iterations, a onstan t step size γ ℓ = 1 is  hosen, in order to let the Mark o v  hain explore the parameters do- main. Then during the last 300 iterations, the sto  hasti appro ximation s heme is implemen ted with a step size equal to γ ℓ = 1 ℓ − 200 at iteration ℓ . This  hoie of step size sequene v eries on v ergene assumption ( A1 .1). The ev olution of ea h SAEM parameter estimates is plotted against iterations (logarithmi sale) on Figure 2 . During the rst iterations of the SAEM algorithm, the estimate sequenes explore randomly some neigh b orho o ds of the initial v alues, through the Mark o v  hain sim ulation. In partiular, these b eha viors are learly visible for the xed eet parameters ( µ and β ). After 200 iterations, the estimates on v erge then rapidly to a neigh b orho o d of the maxim um lik eliho o d, due to the sto  hasti appro ximation s heme. In this example, the iteration n um b er has b een  hosen su h that the on v ergene is learly attained b efore the last iteration. [Figure 2 ab out here.℄ The relativ e bias and RMSEs obtained on the 1000 datasets with n = 24 and n = 40 sub jets are presen ted in T able 1. [T able 1 ab out here.℄ The bias and the RMSEs of the xed eets ( µ ) are small with the SAEM algorithm and are almost half of those obtained with nlme (esp eially the RMSEs). The bias and RMSEs of the unit eet ( β ) are small and on the same order with b oth metho ds. F or the b et w een-sub jet 15 v ariabilit y parameters ( Ω ), the bias are redued with SAEM, while the RMSEs are of the same order with b oth metho ds. F or the within-sub jet v ariabilit y parameters ( Ψ ), the bias and the RMSEs are satisfatory , and on the same order with b oth metho ds. The bias and RMSE for σ 2 are small and satisfatory for b oth metho ds. The t yp e I error of the W ald test and of the LR T are ev aluated on the same 1000 datasets. F or n = 24 , the t yp e I errors are 6.0% and 6.5% for SAEM and nlme, resp etiv ely , for the W ald test, and 4.6% and 5.6% for SAEM and nlme, resp etiv ely , for the LR T. F or n = 40 , the t yp e I error are 5.6% and 5.4% for SAEM and nlme, resp etiv ely , for the W ald test, and 5.8% and 5.2% for SAEM and nlme, resp etiv ely , for the LR T. 5 Appliation to the p opulation pharmaokinetis of atazana vir with tenofo vir 5.1 Study p opulation: ANRS 107 - Puzzle 2 study The Puzzle 2 - ANRS 107 trial w as a randomized op en-lab el, m ultiple-dose study supp orted b y the F ren h Agene Nationale de Re her he sur le Sida (ANRS) with HIV-infeted patien ts in treatmen t failure with their an tiretro viral therap y . P atien ts w ere randomized to reeiv e for the rst t w o w eeks either their un hanged treatmen t with PIs and n uleoside rev erse transrip- tase inhibitors (NR TIs) (group 1) or un hanged treatmen t with NR TIs in om bination with atazana vir (300 mg QD) plus ritona vir (100 mg QD) as a substitute for the failing PI therap y (group 2). F rom w eek 3 (da y 15) to w eek 26, patien ts from either group swit hed to atazana vir (300 mg QD) plus ritona vir (100 mg QD) plus tenofo vir DF at 300 mg QD and NR TIs seleted aording to the baseline rev erse transriptase genot yp e of the HIV isolated in ea h patien t. In this pap er, w e analyze onen tration data obtained from 10 patien ts from group 2 who w ere inluded and measured at 1, 2, 3, 5, 8, and 24 h after administering drug during ea h treatmen t p erio d. Those exat dosing in terv als w ere reorded. The ob jetiv e of the substudy w as to measure the pharmaokineti parameters of atazana vir (administered with ritona vir) either b efore (da y 14 [w eek 2℄) or after (da y 42 [w eek 6℄) initiation of tenofo vir DF in HIV- infeted patien ts in order to detet pharmaokineti in terations of tenofo vir on atazana vir. Data of this substudy w ere analyzed using a nonlinear mixed eet mo del b y P anhard et al ( 23 ) 16 and a signian t eet of the o-administration of tenofo vir on the pharmaokineti parameters of atazana vir w as found using the nlme funtion of the Splus soft w are. The aim of the presen t analysis is to ev aluate the eet of tenofo vir on the PK parameters of atazana vir using the SAEM algorithm and the W ald test desrib ed in setion 4. W e use the one-ompartmen t mo del with zero-order absorption prop osed b y P anhard et al. (23 ) to desrib e atazana vir onen trations: f ( t, φ ) = F D T a C l (1 − e − C l V t ) 1 t

Original Paper

Loading high-quality paper...

Comments & Academic Discussion

Loading comments...

Leave a Comment