Therapeutic drug monitoring as a tool in treating HIV infection

Introduction Despite its success, antiretroviral therapy (ART) is still associated with a significant risk of toxicity and subsequent virological failure. Pending the introduction of drugs with greater potency and durability, one of the most pressing priorities in HIV treatment is to identify ways in which to maximize the efficacy of existing therapeutic options while minimizing its toxicity. Resistance testing has been shown to be one such strategy. Therapeutic drug monitoring (TDM) for HIV drugs has been widely proposed as another tool by which treatment may be optimized. There are some parallels with resistance testing, not least the potential difficulties or pitfalls in interpreting test results and the understandable pressure for provision of this service in the face of increasingly limited treatment options, notwithstanding the current lack of data from controlled clinical studies. In this review, we shall discuss the value of TDM in the context of a 'monitoring' service (in other words, drug concentrations are provided in conjunction with interpretation and advice), rather than a facility for 'measuring' drug concentrations. This distinction is important since the interpretation of drug concentrations is fraught with complexity. TDM for all classes of drugs? It is first important to establish that the measured compound correlates directly with pharmacological activity (i.e., it is the active molecule). Of the three classes of drugs [nucleoside analogues (NA), non-nucleoside reverse transcriptase inhibitors (NNRTI) and protease inhibitors (PI)] used to treat HIV, this is true only for NNRTI and PI. NA are pro-drugs, requiring intracellular activation to their active triphosphate metabolites. Only a small proportion of parent compound penetrates intracellularly, and the amount of this that undergoes phosphorylation depends on the enzymes involved and the presence of other NA or drugs that may affect the activation pathway. Moreover, the intracellular formation of zidovudine phosphates may vary with CD4 cell count [1,2]. Studies have observed that plasma zidovudine and lamivudine concentrations correlate poorly with intracellular drug-triphosphate concentrations and thus cannot be used to predict levels of the active metabolite [1,3]. Direct measurement of intracellular NA triphosphates reveals wide inter-patient and intra-patient variability and is not practicable for clinical use. There is therefore little evidence that TDM for NA would be beneficial. In the following sections we examine the potential role of TDM for the PI saquinavir (SQV), ritonavir (RTV), indinavir (IDV), nelfinavir (NFV), amprenavir (APV) and lopinavir (LPV)/ritonavir, as well as the NNRTI nevirapine (NVP) and efavirenz (EFV). The case for TDM for PI and NNRTI There are potentially three major ways in which TDM may contribute to HIV treatment: first, by improving potency following initiation of ART; second, by improving durability of the regimen once maximal virological suppression has been achieved; and, finally, by minimizing drug toxicity. It must be stressed that the evidence that TDM may influence any of these has not yet been rigorously established. Nevertheless, there are compelling arguments for TDM. Low plasma drug concentrations correlate with virological failure This represents the strongest case for TDM. Dose-ranging studies have shown a relationship between the dose of drug administered and virological response [4-10], and considerable data have now accrued to suggest that treatment failure (as judged by viral load response) is associated with low plasma concentrations of PI and NNRTI. The following examine the evidence for a relationship between plasma drug concentrations and viral load response. Protease inhibitors Saquinavir. A strong correlation between plasma SQV levels and virological response has been reported for hard gel preparations given as monotherapy (n = 40) [4] or in combination [11] (where an association with CD4 cell increase was also observed). A similar association between virological response and area under the drug concentration-time curve (AUC) or plasma trough (end of the dosing interval) concentration (Ctrough) was also reported for soft-gel SQV given as monotherapy (n = 84) [5] or in combination in 14 PI-naïve children [12,13]. Mathematical modelling suggested a target AUC0-24 h of ≥ 20 000 ng h/ml for maximal virological suppression [5]. Plasma 'concentration ratios' of SQV (ratio between measured plasma drug level and the corresponding value from a reference population of HIV-positive patients at an identical time-point) were measured in 130 clinic attenders receiving the drug [14]. These were significantly related to viral load decline at 12, 24, 36 and 48 weeks of therapy, with sustained virological suppression (> 2.0 log10 copies/ml) in those having an estimated Ctrough > 50 ng/ml (P = 0.008). Ritonavir. An association between RTV AUC and Ctrough and the emergence of characteristic drug resistance mutations has been observed [15]. The Ctrough of RTV was associated with CD4 cell and viral load responses in a dose-ranging study of 84 PI-naïve patients [16] and in studies of children receiving RTV (n = 41 [9] and n = 31 [17]). A target Ctrough > 2100 ng/ml was proposed [16]. No correlation between RTV peak plasma concentrations (Cmax) [17] or AUC [9] and treatment response, was observed. Sampling after only 7 days in one study [9] was not ideal since this was during the window when RTV induces its own metabolism. Indinavir. Phase I/II studies of IDV monotherapy have yielded conflicting data that have both supported [18] and failed to observe [19] a concentration-response relationship. No correlation between IDV concentrations and virological response or serum bilirubin was observed in a study of 95 patients [20], but sampling details were not provided. Conversely, IDV AUC and Ctrough were significantly associated with the rate of HIV RNA decline (n = 25 [6]) at 7 days as well as longer term virological response (proportion achieving < 500 copies/ml or subsequent virological rebound) in adults (n = 43 [21] and n = 35 [22]) and children (n = 19 [23]) on combination therapy. One of these studies utilized population pharmacokinetic modelling to derive the AUC [22]. Collectively, these studies suggest a target AUC0-24 h of 20 000-25 000 ng h/ml [22,23]. A case-controlled substudy of 58 virological failures from TRILEGE (an induction-maintenance study in ART-naïve subjects) observed low/undetectable IDV concentrations (indicative of non-adherence to therapy) at baseline in 7/7 treatment failures receiving maintenance with IDV, zidovudine and lamivudine [24]. In contrast, another induction-maintenance study (ACTG 343) [25] failed to demonstrate a similar relationship in 26 subjects failing IDV therapy compared with 10 individuals who remained virologically suppressed. This was not surprising given the sampling at random time-points and patient selection. Nevertheless, virological rebound was associated with very low IDV concentrations (< 50 ng/ml; P = 0.03) in patients randomized to triple therapy. In a study of 17 late-stage patients receiving salvage therapy with IDV, NVP and lamivudine [26], a significant association between the virological response at 24 weeks and IDV Ctrough (P = 0.03) was observed, although no association was noted with IDV Cmax (P = 0.10) or with NVP concentrations. In an outpatient clinic-based study, a low plasma IDV concentration ratio in 65 patients commencing an IDV-containing triple-drug regimen was an independent factor predicting virological failure by multivariate analysis (P = 0.007) [27]. The effect of low IDV concentrations was compared with two other risk factors: high pretreatment viral loads, and previous PI exposure. The attributable effect of adequate versus low IDV concentrations was minimal in patients with neither (8% versus 11% failure) or both (100% failure) of the other two risk factors. However, low IDV concentrations were strongly associated with failure when either high baseline viral load (9% versus 56% failure) or previous PI treatment versus was These data are important since suggest that TDM may have the on patients who are at risk of failure rather than patients with and low viral loads, or those with the risk of treatment failure. from other PI in two which TDM results in drug concentrations to after This that the drug concentration may after Ctrough has been the presence of an active metabolite that has similar activity to may also since of is to by the Despite concentration-response have been observed. In a study of patients receiving zidovudine and lamivudine the of response 48 weeks of were baseline viral load (P = rate of HIV RNA decline at (P = and plasma concentrations (P = This correlation was for = than for Ctrough = the may have been to variability from of of previous pharmacokinetic modelling in patients receiving reported a significant association between Ctrough and treatment in multivariate analysis (P < concentrations at random time-points and compared and a level of were associated with subsequent virological failure by multivariate analysis risk = P = in a analysis (n = of the study The correlation may in have been by the sampling strategy. In the 50 children AUC sampling for and at weeks and The measured trough and AUC concentrations of drug with the proportion of patients achieving viral load at weeks of < or < 50 Moreover, those children with the AUC of or at the most in HIV RNA between weeks and These similar evidence in adults of a concentration-response relationship for plasma trough concentrations of were measured in individuals receiving a of of The of the characteristic in the protease in to but also has viral was significantly associated with individuals having Ctrough (P = This correlates with dose-ranging studies of the while the pathway. In a dose-ranging study of patients receiving monotherapy for the Cmax and Ctrough of with viral load the first weeks of therapy Studies in adults and children have observed a relationship between the of and virological There are data on for These would be of given that Ctrough of that of the concentration for A small study has reported a significant association between the of and virological response PI SQV The study patients therapy with and Plasma drug 'concentration ratios' in a of patients a greater rate of HIV with concentration of SQV (P = and (P = SQV Studies of SQV RTV are limited by small and sampling and have yielded conflicting Plasma drug concentrations of limited plasma and random from but PI-naïve patients commencing SQV RTV were significantly associated with treatment response (> log10 in HIV RNA at for SQV (P < and RTV (P = However, no correlation was observed between SQV and RTV and treatment response in a study of 24 patients study patients receiving a combination two RTV SQV and patients were to NNRTI. and plasma PI concentrations were measured at weeks and No significant in drug concentrations were observed between patients with (< 50 copies/ml) HIV RNA and those failing to virological response at 24 reverse transcriptase inhibitors well NNRTI from PI in having longer and achieving a plasma level with to the for However, their low for drug resistance the NNRTI potentially to failure in the presence of drug concentrations. Plasma drug concentrations were measured in patients randomized to NVP in the study A significant correlation with both and responses was observed. NVP concentrations during the first weeks significantly with the rate of HIV from plasma by multivariate analysis (P = and was an independent for predicting to achieving viral load < 20 copies/ml (P = significant were observed for patients plasma NVP concentrations levels from and viral load at These results to be at with another pharmacokinetic modelling of plasma NVP concentrations from patients was used to predict Ctrough and AUC to for a relationship The study patients (n = who were as load < of viral suppression or rebound) or No association between virological response and plasma NVP concentrations was in An association between CD4 cell increase at and NVP concentrations was observed only in study of 17 patients with late-stage receiving IDV and NVP as of a salvage regimen failed to demonstrate any relationship between the Cmax (P = and Ctrough (P = of NVP and virological response at 24 weeks These results from the substudy that and once the that the value of TDM may between of The AUC of and in 50 HIV-positive children after weeks of treatment was significantly and related to virological and children with the AUC of or the most in HIV RNA the of No relationship with toxicity was Plasma concentrations were measured with viral load and a of The study population a of 130 HIV-positive patients on combination therapy for load were not controlled for of therapy, viral or previous drug were data for viral responses to sampling h was utilized given the and plasma concentration-time of A significant correlation was observed between patients with concentrations < ng/ml versus ng/ml versus > for viral load copies/ml versus copies/ml versus P = and of virological failure (> versus versus P = data from studies of utilized population pharmacokinetic modelling to predict trough plasma concentrations in HIV-positive patients The Ctrough of in treatment versus failure was versus ng/ml versus and that this was a factor in (P = A of treatment by a factor of for increase in plasma failure was three as when the Ctrough of was < a on the in for the variability Plasma PI concentrations are to a of and it is not to observe a variability in trough concentrations of IDV, and variability has also been observed for NNRTI. this of it is to predict plasma concentrations in any given on TDM potentially be used to identify patients at risk of treatment Plasma concentration for trough from patients by the TDM to The a plasma ratio of < were ng/ml; IDV, ng/ml; ng/ml; 2100 ng/ml; ng/ml; ng/ml; Only ritonavir ≥ were trough were for all efavirenz on a of 36 h were from a plasma drug concentrations may predict toxicity The evidence for this by with the There is no between of or risk of or and plasma PI although studies are such as or are to be related to drug rather than to the amount of drug in Nevertheless, there are limited data some to plasma drug concentrations. Plasma at significantly with Ctrough of SQV (P = and RTV (P = 0.03) in 25 patients on RTV SQV trough concentrations of both drugs with one it was to the potential of A in was in HIV-positive adults receiving a of RTV although this was not observed in children were also observed between Ctrough and AUC of RTV and plasma and may also be related to the Cmax of of 17 patients receiving RTV reported requiring in patients The most were and of patients reported the of these between and h after These were in two patients who would have therapy, by RTV h and Cmax while adequate commencing RTV with or (n = significantly RTV Cmax (P = Ctrough (P = and h (P < than those toxicity (n = the first in plasma drug concentrations were to be to of since these were on The of in patients on compared with three IDV that high peak concentrations of IDV may be to toxicity. IDV 'concentration ratios' were than in all but one of patients with or and in were the for a reference population (P < in patients in adequate therapeutic virological and individuals on In a study of 130 patients receiving a strong correlation between random plasma concentrations and was observed. toxicity in the three of patients with plasma concentrations < ng/ml versus ng/ml versus > ng/ml were and (P = TDM may have a role in monitoring to The of PI is although of RTV to wide or of drug in plasma is a of but concentrations are of limited An adequate or high plasma drug level only the rather than in the or not TDM to a drug has yet to be established. PI and NNRTI are by PI are inhibitors of and, in RTV is an of and NNRTI may either or (NVP) although is both an and an of These drugs may not only affect the of other drugs that the but may be by those drugs are PI or drug are the of this and are or are not in patients receiving salvage are as an important factor in drug with is an of and may plasma concentrations of and NVP such as and may also have significant with PI or NNRTI. TDM potentially be used to concentrations in such in children from In and is but it in who greater of drugs compared with therefore be to area or rather than One study in in a of and in combination with and administered to 26 to HIV-positive at all three after the first 7 to or The AUC of also by a of between 14 and A drug in compared with was observed. Of at the dose of than of all target Ctrough or AUC important between adults and children low and in to These all contribute to the high variability in concentrations observed in have been observed (as for for RTV and These studies also that children at least as not variability in plasma concentrations of TDM may to identify and dosing in this of may in to in plasma and plasma The of during in a case with a corresponding in the of by during was to adequate trough Despite viral was not and both and a dose of NVP following variability in concentrations was observed in patients with HIV and > three of The pharmacokinetic was by (< 500 and plasma (> of compared with a reference population with significant levels of which PI and (< formation of the active metabolite The may have been to variability in the for activity as a of from other drugs or a combination of these factors. In another study, 10 subjects with with 10 for and a dose of was significantly in subjects with as judged by with Cmax was significantly versus or ng/ml versus The plasma AUC of was between the two but the plasma to be longer h versus in individuals with a NA are most by but are not for as NNRTI are by the and only a small is in is not and TDM PI also and not dose in One is IDV, which be used with In the combination of IDV RTV may in an of IDV and an proportion of IDV that is in TDM on measurement of the amount of drug and in are not and the test has limited In the presence of significant TDM may or to dose of of studies as a the of studies has shown either a correlation between low plasma concentrations and virological failure or a correlation between plasma concentrations and virological response. The most studies have shown a relationship between plasma concentrations of and and the of viral load However, must be when an studies concentration-response were monotherapy with the drug of for a the of other in a is no longer to clinical It is that may have studies such a relationship has not been observed. Conversely, some studies have failed to a concentration-response relationship. may for Sampling the of random or and was not in studies. for most studies are The plasma relationship may vary between to the pharmacokinetic and proportion of patients with and with for of and most this relationship is to be on the presence of other risk for treatment failure. This may the lack of a strong relationship for patients receiving salvage therapy for and that TDM may be at some of patients compared with The study the of viral resistance in HIV-positive failing therapy In a pharmacokinetic substudy of plasma PI were measured to the dose of drug PI concentrations were as were than the of HIV for plasma the target for PI were as ng/ml; ng/ml; ng/ml; and IDV, with other the were low for and high for plasma concentrations were in IDV RTV and SQV although the was in patients on SQV with RTV or A significant relationship was observed between plasma HIV RNA and the concentration of and multivariate that PI concentrations were an independent of HIV RNA response (P = These data are important since that achieving therapeutic plasma drug concentrations an and effect on of response to a data have been from a study of TDM receiving PI were randomized to of (n = or TDM (n = 84) trough PI concentrations at with dose at The in HIV RNA at was in both However, there are some to these was only after weeks of therapy, and the analysis only examine the effect of weeks of TDM. requiring were also which may be a of the of PI. No toxicity were the is any be on the of TDM. The study of TDM in HIV treatment is the patients have been randomized to TDM or no TDM. A analysis has now been on a of patients commencing IDV (n = and (n = treatment were observed in the TDM for IDV (P = and (P = in = failure the TDM a significantly greater proportion of patients with a viral load < 500 at for IDV (P = and (P = These data some of the evidence to that TDM may treatment for patients receiving PI. data are with TDM is the of has been on the in or for drug HIV, for plasma This although widely has some testing of may considerable variability in The of for plasma from the of effect following with serum and has not been for interpreting in data have suggested that this is a Ctrough value the for HIV is the major is a lack of data to the in of the we the Ctrough This is to with in previous to a target on HIV may be for patients on salvage therapy. the relationship between drug concentrations and a the of has been This may but the most is that of or for of the own that some of resistance testing by testing, or a from modelling on resistance is This is both and but treatment for patient to be some with interpretation still resistance and Ctrough was in patients with both and A of for was observed (< to 24, the trough of was a of virological response (< In contrast, a significant and strong correlation with the of was observed, with a response rate of and for < and > data have also been for patients receiving IDV who with RTV to virological failure a reported a association between the of IDV and subsequent virological response log10 or copies/ml viral patients with a < virological failure on compared with of patients with a > (P < these data are studies are to the relationship between and virological A dose-ranging study of patients receiving reported a significant (P < correlation between Cmax and AUC of and viral load testing of baseline viral was used to an The of was to be a of subsequent response to than Ctrough resistance and plasma were in children receiving response (as measured by the from baseline HIV was significantly associated by analysis with the ratio for plasma of and baseline multivariate only the ratio was significantly associated with virological response (P = There are potential pitfalls in the of There is lack of in the in reference of HIV used and in the utilized for these to be and for in it would still be be compared between classes of drug or drugs of the (where to resistance and pharmacokinetic Nevertheless, have some the effect of RTV in a The target for the of drug concentration has to toxicity has been most observed for RTV and IDV but this a plasma level has yet to be Only one study has to identify the for by the for therapeutic response (> and toxicity (> studies in this area are in the face of PI and drugs? The proportion of patients with PI trough concentrations the for HIV is with of However, the variability is not and a significant of patients still to target concentrations. The of is to increase the proportion of patients who are to be with there is a of viral such as very high in clinical and it is that TDM for these may be utilized to toxicity. data from the and the TDM suggest that low drug concentrations were in patients receiving IDV and NVP TDM Despite the that target concentrations were between the and sampling were and the TDM may have a population of patients at risk of low concentrations the outpatient clinic the data from both are and suggest that drug concentrations are than of patients in the and with low plasma variability and plasma PI variability in plasma trough concentrations. may for such as the effect of on of drug or of of the previous PI such as RTV and in concentrations with the that the concentration-time is than the with peak concentrations and longer plasma of the this for of these are on their own to PI and NNRTI as for TDM. similar have been for drugs such as and for which TDM is now PI are also = = other PI > to the of with and It is important that TDM is not during such as and when to There is a lack of from TDM sampling in to the population for that although most would that the of AUC is not practicable for most of now some of population pharmacokinetic these the Cmax and AUC may be from sampling at random time-points with the of the of previous dose and some clinical However, no between this and there is a lack of virological would be for two PI or a PI and in such as the very However, the to population modelling would the of and concentrations of drugs for TDM. There is an on target concentrations be for but still between An has also been for TDM The lack of data from studies on the clinical TDM it to on the of TDM in the In we have the used in the HIV for the treatment of adults an and the of and the of evidence for that for therapeutic drug monitoring to these as data on the of the combination of drug measurement with resistance that the ratio of drug concentrations vary between of In there are to TDM by sampling from directly measured or from random sampling population pharmacokinetic