Therapeutic Drug Monitoring (TDM): Ready for Prime Time?
Do We Need Another Lab Test in the Clinic?
by Mark Harrington

No one could accuse the HIV pharmacologists of being attention hogs -- but they were excited. It was the First International Workshop on Pharmacology in HIV Therapy, held on 30-31 March 2000 in Noordwijk, the Netherlands, a storm swept town on the coast of the North Sea. Stiff winds, damp weather, and nothing to do kept conference attendees inside the hotel during a two-day symposium which focused on a number of technical issues in HIV pharmacology. The talk in the hallways, however, was about therapeutic drug monitoring (TDM) for antiretroviral therapy, an issue which has assumed increasing prominence over the past year or so.

Several recent studies have shown that variable blood levels of HIV drugs can have a major impact on the success or failure of antiretroviral therapy. These variations can be caused by how much drug reaches the blood, how rapidly the drug is cleared from the blood, or by metabolic interactions with other drugs. Some studies are now taking place to prospectively assess whether or not measuring antiviral drug blood levels -- then individually tailoring doses to ensure continuously therapeutic levels-- can reduce the incidence of treatment failure.

Since suboptimal drug concentrations are likely to lead to viral rebound, TDM might offer a way to achieve optimal concentrations and thus prevent the emergence of resistance and cross-resistance. However, although TDM has already been included in French treatment guidelines (particularly in the setting of suboptimal response to PI-containing regimens), and is available to some HIV-infected individuals in several European countries, including France, the Netherlands, and the UK, the clinical benefits of TDM remain to be proven.

Plasma concentrations of antiretroviral drugs, particularly the protease inhibitors (PIs), vary significantly among individuals. Both the PIs and the non-nucleoside reverse transcriptase inhibitors (NNRTIs) are metabolized through the liver by the cytochrome P450 system, creating complex two- and three-way interactions when these drugs are used together or with other drug metabolized by this system. In addition, all three classes of drugs may be affected by cellular influx and efflux systems mediated by p-glycoprotein (PGP), which has been implicated in multi-drug resistance (MDR) to cancer chemotherapy (see the accompanying article by Yvette Delph).

Let's examine what is already known about the link between HIV drug pharmacokinetics and treatment outcomes.

I. Suboptimal Antiretroviral Dosing can Lead to Treatment Failure

We already know from several early studies of protease inhibitor monotherapy and combination studies, particularly in salvage therapy, that suboptimal drug concentrations lead to therapeutic failure, resistance and cross-resistance.

a. Invirase brand saquinavir
Remember Invirase brand saquinavir? Perhaps you'd rather not. (If you're up to the bad flashback, please refer to A Cynical Swiss Saquinavir Scam: Roche Admits Licensed Dose Suboptimal, TAGLine, 4(7), August 1997. In brief, Roche rushed saquinavir -- in vitro the most potent drug in its class -- to market without defining the maximum tolerated dose (MTD). This in spite of the fact that the then-current hard gel capsule formulation was only 4% bioavailable. A later study, ACTG 333, showed that the insufficient concentrations of saquinavir in Invirase resulted in not only treatment failure and resistance to saquinavir, but subsequently to indinavir as well.

Later, as the drug lost market share to its more potent classmates indinavir and ritonavir, Roche belatedly conducted studies of higher Invirase doses and of a new soft gel capsule saquinavir formulation, which was dosed at a concentration able to achieve more potent and durable viral suppression. Eventually the new formulation, dubbed Fortovase, was licensed by the FDA and substantially replaced Invirase (dwindling supplies of which are being dumped by Roche in developing countries). Nevertheless, because the pill count for Fortovase, when used as a single PI, remains daunting, it is most often used today in combination with ritonavir. Ritonavir, itself a PI, may be more noteworthy for its unsurpassed ability to inhibit the P450 metabolic pathway, a feature which enables drugs like saquinavir to be given in smaller doses that achieve higher plasma concentrations and longer plasma half-lives.

Two recent studies, ACTG 359 and VIRADAPT, showed that pharmacology can help explain treatment failure. However this is different from demonstrating that pharmacology, used in the clinic, can help maximize treatment success. At least one ongoing study, ATHENA, is attempting to validate the use of TDM in a prospective manner.

b. ACTG 359 -- Salvage Regimens Better Without Adefovir
ACTG 359 was a study of 277 pre-treated patients with virologic failure (median viral load at baseline 31,746 copies/mL; median CD4 count 229/mm3), who were randomly assigned to take one of two double protease regimens (saquinavir with ritonavir or saquinavir with nelfinavir) plus either delavirdine, adefovir, or both. Unexpectedly, the results showed that those assigned to any of the adefovir-containing arms -- with or without delavirdine -- did worse than those assigned to delavirdine without adefovir:

The ACTG 359 executive summary explained that "the explanation for the inferior virologic effect in the adefovir arms may be that subjects had extensive nucleoside experience or discontinued lamivudine [3TC] upon study entry... The lack of additivity or synergy in the delavirdine plus adefovir combination arms may well be due to an adverse pharmacokinetic interaction between delavirdine and adefovir first demonstrated in the intensive pharmacokinetic substudy (ACTG 884) of the current study... [which] showed that delavirdine levels were halved when co-administered with adefovir... In addition, saquinavir levels were reduced by about half in the delavirdine plus adefovir combination arms, possibly as a direct result of the decreased delavirdine levels."

So the unexpected adefovir-delavirdine effect caused a chain reaction which, in turn, caused a delavirdine-saquinavir effect, reducing levels of both delavirdine and saquinavir. Clearly it would have been preferable if the pharmacokinetic substudy had been carried out before ACTG 359.

c. VIRADAPT: Pharmacokinetics Better Predictor of Failure than Resistance
VIRADAPT was a prospective study designed to address whether providing genotypic resistance testing results would improve virologic responses to second-line or salvage therapy regimens. Forty-eight week follow-up data and a pharmacologic substudy were presented at the FDA Antiviral Drugs Advisory Committee meeting on 2 November 1999 by the principal investigator, Philippe Clevenbergh of the Hôpital de l'Archet in Nice, France. In VIRADAPT, 108 patients with viral load over 10,000 copies/mL who had been treated with protease inhibitors for at least three months and with NRTIs for at least six months were randomized to a control group (N=43) or a group which received genotypic testing results (N=65) before switching therapies. For six months study participants were treated by randomization arm, after which patients in both arms received treatment changes according to genotypic results performed every three months. The primary endpoint was HIV RNA variation from baseline at months 3 and 6. Secondary endpoints included proportion of patients with RNA < 200 copies/mL and CD4 count. Statistical analysis was by intent to treat (dropout equals failure). The 6-12 month analysis assessed RNA changes from baseline at months 9 and 12, proportion of patients with RNA < 200 copies, and used on treatment analysis. The genotyping technology used was the Visible Genetics TrueGene assay. A drug resistance table was provided. Treatment arms were well-matched by baseline viral load (4.0 log10), CD4 count (210), prior antiretroviral treatment, and baseline mutation frequency.

In this heavily pre-treated population, a virologic response of -1.15 log10 was sustained with genotypic guided therapy for twelve months. The presence of primary protease mutations and the performance of genotypically guided treatment each independently affected the results.

Seventy percent of the patients in VIRADAPT did not achieve plasma virus levels below the limit of detection (BLD). Since poor efficacy could also result from suboptimal drug exposure, particularly to protease inhibitors, the study team subsequently performed pharmacokinetic analysis of protease inhibitor plasma levels. The pharmacologic substudy included 87/108 patients. Serial PI plasma trough levels were performed in both arms throughout the study. PI levels were determined by high performance liquid chromatography (HPLC). Samples were collected before the morning dose. Analysis was conducted on batched frozen samples. Levels were determined for all four study PIs -- indinavir, nelfinavir, ritonavir and saquinavir. Data were analyzed only for subjects with at least three measurements. 604 plasma levels were obtained from 81 patients, who were well-matched at baseline between treatment arms.

Virologic response was analyzed by whether subjects achieved optimal PI concentrations (OC, 68%), with no more than one PI level less than twice the IC95, or suboptimal concentrations (SOC, 32%), with two or more levels less than twice the IC95.

VIRADAPT Efficacy Analysis by PI Drug Levels and Randomization Arm
	RNA change at 3m	%BLD at 3m	RNA change at 6m	% BLD at 6m
SOC/Control	-0.08	0%	-0.20	0%
SOC/Genotypic	-0.65	15%	-0.70	8%
OC/Control	-0.75	22.5%	-0.95	22.5%
OC/Genotypic	-1.15	34%	-1.35	40.0%

VIRADAPT Factors Predictive of Virologic Response
	OR	95% C.I.	p-value
PI Concentration (>IC95 x 2)	2.37	2.0 - 7.0	0.017
Genotypic therapy	2.24	1.22 - 19.56	0.025
Primary PI mutation	2.47	? - 5.67	0.014

In conclusion, PI drug exposure was inversely correlated with plasma HIV RNA changes for all four PIs. Genotypic guided therapy, PI concentrations and primary protease mutations independently affected responses to therapy. Finally, both resistance assays and pharmacologic drug testing may be useful to improve treatment responses in experienced patients.

II. Prospective Studies to Validate TDM

a. Concentration-controlled vs. Standard AZT, 3TC and Indinavir
Courtney Fletcher and colleagues at the University of Minnesota conducted a randomized study of concentration-controlled (C) versus standard (S) therapy with AZT, 3TC and indinavir in 24 subjects. Six month results were presented at the 39th IC"aC in September 1999 (abstract 322). All subjects received four weeks of standard therapy. Pharmacokinetics were carried out at week two, and subjects were randomized to C or S at week 4. The time to undetectable HIV RNA by PCR and proportion of patients reaching it were compared.

The investigators concluded that Athis preliminary observation supports the hypothesis that interpatient differences in antiretroviral drug concentrations contribute to heterogeneity in viral suppression. (Fletcher 1999)

b. ATHENA
ATHENA is an ongoing prospective randomized, controlled trial designed to evaluate whether TDM can contribute to improved virologic response and thus to reduced HIV-related morbidity and mortality. At the Noordwijk pharmacology workshop, David Burger of the University Medical Center St. Radboud, Nijmegen, the Netherlands, presented the design of ATHENA. They are attempting to achieve protease inhibitor concentrations between 75-200% of the minimum effective concentration (MEC). Drug levels are taken whenever viral load and CD4 levels are monitored, and for viral failure, adverse events, and non-adherence.

What to measure? Trough levels aren't ideal; areas under the curve (AUCs) are impractical; fixed time points are impractical; so they are using random sampling with population pharmacokinetic curves. 600 patients will be enrolled, of whom 50% will be treatment naive. Pharmacokinetic results are available to primary doctors within four weeks of sampling.

Although everyone's plasma is tested, half are randomized to receive TDM test results and expert advice, while control group results are not reported to the primary care physician. The study is about two-thirds enrolled. In Noordwijk, preliminary results indicated that between 26-41% of people were receiving less than 75% of the target PI concentration, and between 5-11.5% were receiving more than 200%, depending on the PI. Eighty-six percent of people on nevirapine were receiving adequate concentrations. Of course, no one knows whether the target concentration ratios are correct. It's too early to know whether physicians are using the information from the TDM assay to guide therapy. There are no correlations yet with viral load or resistance. The study is ongoing.

c. ACTG TDM Working Group
The Adult AIDS Clinical Trials Group has set up a TDM Working Group, involving members of the HIV Research Agenda Committee (RAC) and the Pharmacology Committee. Its role is to determine the role of TDM in "aCTG studies. They are considering looking at TDM in four settings:

1. PI naive;
2. optimizing PI regimens;
3. drug intensification; and
4. PI virologic failure.

III. TDM in Clinical Practice: Pro & Con

There seems to be more impetus behind TDM research in Europe than in the USA. Some US researchers have been heard to say that better drugs with longer half-lives, lower peak, and higher trough levels will take care of the problem, without necessitating the addition of another costly blood test in clinical practice. For example, coadministering low doses of ritonavir with other protease inhibitors such as amprenavir, indinavir, or saquinavir, allows the latter drugs to be taken in lower doses less often and -- in the case of indinavir -- without previous restrictions on food intake. This is because ritonavir, by inhibiting the cytochrome P450 enzyme system of the liver, slows down the metabolism of the other PIs, raising their trough levels, reducing their peak levels, and increasing the area-under-the-curve (AUC). However, the ritonavir/other PI studies to date remain relatively short-term, and these combinations have yet to be added to antiretroviral treatment guidelines (with the exception of ritonavir/saquinavir, for which 72 week data are available).

Baltimore pharmacologist Charles Flexner gave an overview of this issue in the on-line Hopkins HIV Report (January 2000). He points out that TDM can be useful in certain therapeutic settings, such as with the anticoagulant drug warfarin, the anti-convulsant phenytoin, and the immunosuppressive, cyclosporin. However, continues Flexner, ATDM is rarely useful for antibiotics... The therapeutic index for most antibiotics is high... Few studies ... demonstrate a clear-cut relationship between antibiotic concentration and outcome... The most important exception is the aminoglycoside class, where TDM is used mainly to prevent toxicity." At the 12th World AIDS Conference in Geneva, two abstracts found a relationship between protease concentrations and virologic outcomes, while three found no such relationship [abstracts 42261 and 42275 were in the Ayes@ side, while 42266, 42272, and 42276 were on the no relationship side.] Flexner goes on to state that Ait is essential to show, prospectively, that adjusting dose to achieve some pre-determined target concentration actually improves outcome." The Fletcher study cited above offers some preliminary evidence that this may be the case with AZT/3TC/IDV, but its small size (N=24), and the increasing use of IDV at the twice daily 800 milligram dose in combination with RTV 100 or 200 mg bid may reduce its clinical relevance.

Flexner notes that Asome drugs (like AZT and 3TC) exert significant anti-HIV effects yet routinely have trough concentrations of zero." This is (at least in part) because these drugs are active in their intracellular, triphosphate forms, which are not measured by plasma tests. However, this point indicates that TDM may be more useful in certain drug classes, and for certain drugs, than in other classes and for other drugs.

In another timely discussion of TDM, the on-line Medscape HIV/AIDS (1999) featured a debate between NIH pharmacokineticist Stephen Piscitelli (author of the useful and stimulating study which demonstrated that the popular herbal antidepressant St. John's wort causes dangerous reductions in indinavir concentrations, Piscitelli 2000) and Edward Acosta on the "limited value" (Piscitelli) vs. the "promise" (Acosta) of TDM in HIV infection.

Piscitelli cites "at least seven substantial obstacles that suggest that TDM will have little significance in clinical practice:"

1. Intrapatient variability. "A blood level collected on Monday ... may be markedly different from one drawn on Friday." In one AZT study in women, the mean AUC varied as much as two-fold in some patients (Cordaro 1993); in an indinavir study in women, the mean AUC varied from 20.2 nM*h in the menstrual phase to 34.0 nM*h in the follicular phase (Adams 1998). What you eat and when can also affect many antiretroviral drug concentrations.
2. Effect of protein binding. Many PIs, such as amprenavir, bind to alpha-1 acid glycoprotein (AaG), "an acute phase reactant ... whose concentrations can be increased by stress, injury, or infection." So here you'd need to monitor amprenavir concentrations along with AaG.
3. Sample timing. Would you sample the peak, the trough, the AUC?
4. What is the target concentration? Most studies assess this only in drug-naive patients. Target concentrations will vary by viral phenotype. So you have to do phenotypic resistance testing as well as TDM.
5. Logistics. To catch suspected treatment failure, you'd have to get to the clinic, give the specimen, have it sent to an outside lab, wait for the results, confer with your doctor, and act on them. By this time, drug resistance -- especially if it is to 3TC or an NNRTI -- may well have emerged and it will be necessary to change regimens in any case.
6. Interpretation. Who will interpret the results of TDM? With the nucleoside analogues, the concentration of interest is of the intracellular metabolite, not drug in the plasma. In the blood, ddI has a half-life of two hours, but in the cell, it is over 11 hours.
7. Assays. Certain labs will perform TDM, but there are no standardized, simple, and inexpensive methods, nor any quality assurance for labs carrying out TDM.

Acosta responds by citing the previously mentioned saquinavir data (Schapiro 1996) and other PI monotherapy studies. He concludes that Athe significance of TDM for PIs in the treatment of HIV ... has yet to be completely understood." He suggests that it may be useful for Arandom adherence checks@, although most pharmacologists at the Noordwijk meeting disagreed with the concept that adherence could be measured pharmacokinetically.

This raises another issue, which is that if an individual is experiencing difficulty adhering to an antiretroviral regimen, measuring and adjusting the regimen dosage may not make much difference. Some TDM studies have been distorted by the so-called Awhite coat effect@, in which patients not normally especially adherent dutifully take all of the pills they're supposed to in the knowledge that they're about to be monitored for drug levels. This would tend to overstate drug exposure over the long term.

The question of whether to measure adherence or drug levels, or both, to optimize therapeutic responses, or to prevent, or identify early and mitigate, antiretroviral treatment failure, remains unanswered at this time. Nor is it clear whether the best use of limited resources is to provide routine TDM in the clinic, or simply to develop therapeutic regimens which are more forgiving and have better pharmacokinetic profiles.

One thing that is clear, however, is that pharmaceutical manufacturers need to be more pro-active in providing access to their compounds for drug-drug interaction studies both before and after FDA approval.

IV. Some unanswered questions about TDM

1. Can TDM help optimize therapy, increase efficacy, reduce resistance and delay treatment failure?
   How many / what proportion of patients will benefit, and at what incremental extra cost, offset against what benefit (saved regimens, reduced toxicity, etc.)?
   Are there other means of obtaining the same results?

   If so, when and how should TDM be used, for which drugs, and in which patients?

   a. When/in which patients?

   When starting antiretroviral therapy, or within a 2-4 weeks (after steady-state levels are reached)?

   For early treatment failure / adherence assessment?

   When starting second line/salvage regimens?

   Should TDM be given along with resistance tests for early virologic failure, to enable raising doses hopefully to therapeutic levels?

   For toxicity, to enable lowering doses?

   b. For which drugs?

   Protease inhibitors B single, dual?

   Some non-nucleoside reverse transcriptase inhibitors [NNRTIs] B which ones?

   When people are on anti-tuberculosis drugs?

   For which other drug-drug interactions?

   How much evidence, and of what kind, will be needed to determine whether TDM should be integrated into standards of care (SOC)?

   What retrospective studies can help?

   What prospective studies are underway?

   What kind of evidence will be useful (initial antiviral response; time to first failure linked with drug exposure levels, resistance)?

   How should TDM assays be standardized and validated for clinical use? What sort of quality assurance should be used? Which parameters should be measured? Who and how will they be interpreted clinically?

   Cmax (maximum, or peak, concentration) B for toxicity

   Cmin (minimum, or trough, concentration) B for efficacy B which measurement, the IC50, IC90/95, minimum effective concentration?

   AUC (area under the curve) B population pharmacokinetics

   Other parameters?

   More Information

   Cytochrome P450 drug interaction table from Georgetown University: http://www.dml.georgetown.edu/depts/pharmacology/davetab.html

   The Department of Pharmacology & Therapeutics, University of Liverpool, UK: http://www.hiv-druginteractions.org and http://csdinfo2.liv.ac.uk/hivgroup/

Acosta EP. The promise of therapeutic drug monitoring in HIV infection. Medscape HIV/AIDS 5(4), 1999.

Adams J, Frost C, Shelton M, et al. Indinavir pharmacokinetics and menstrual cycle physiology during different phases of the menstrual cycle. Abstract 356, 5th Conference on Retroviruses & Opportunistic Infections, Chicago, 1998.

AIDS Clinical Trials Group, ACTG 359 Executive Summary, 1 June 1999.

Cordaro JA, Morse GD, Bartos L, et al. Zidovudine pharmacokinetics in HIV-positive women during different phases of the menstrual cycle. Pharmacotherapy 1993;13:369-77.

Durant J, Clevenbergh P, Delgiudice P, et al. Drug-resistance genotyping in HIV-1 therapy: the VIRADAPT randomised controlled trial. Lancet 353, 2195-99, 26 June 1999.

Fletcher CV, Kakuda TN, Anderson PL, et al. Viral dynamics of concentration-targeted (C) vs. standard-dose (S) therapy with zidovudine (ZDV), lamivudine (3TC), and indinavir (IDV). Abstract 322, 37th Interscience Conference on Antimicrobial Agents and Chemotherapy (ICAAC), San Francisco, September 1999.

Flexner C. What's in a trough? Therapeutic drug monitoring and antiretroviral regimens. The Hopkins HIV Report, January 2000.

Flexner C, Piscitelli S. Managing drug-drug interactions in HIV disease. Medscape HIV, March 2000.

Piscitelli SC. The limited value of therapeutic drug monitoring in HIV infection. Medscape HIV/AIDS 5(4), 1999.

Piscitelli SC, Burstein AH, Chaitt D, Alfaro RM, Falloon J. Indinavir concentrations and St. John's wort. Lancet 2000;355:547-548.

Schapiro JM, Winters MA, Stewart F, et al. The effect of high-dose saquinavir on viral load and CD4+ T cell counts in HIV-infected patients. Ann Intern Med 1996;124:1039-50.

Webster RD, Barr D. Adherence to H"aRT among individuals with HIV/AIDS: a compendium of H"aRT adherence research, November 1997 - November 1999; http://www.www.gwu.edu/~chsrp (click on "HIV research").

Related Articles
• Policy Statement on NDA 20-993, Gilead Sciences, Inc's Application for Approval of Preveon
• Simple Facts Sheets on delavirdine, Fortovase, ritonavir