Optimal Medical Management
Entering the Era of Truly Individualized Antiretroviral Therapy
By Combining Resistance Testing With Drug-Level Monitoring, Providers Will Be Able to Individualize therapy to a Degree Never Previously Possible
Over the past few years, there has been much speculation and debate about the potential role of drug-level monitoring in the optimal medical management of individuals with HIV infection. In its simplest form, so-called therapeutic drug monitoring, or TDM, can help a clinician determine if a particular patient is getting a suboptimal or supra-optimal dose of a particular drug -- and allow the clinician to titrate that patient's dose accordingly, to maximize efficacy while minimizing toxicities. In its most sophisticated form -- in which TDM is combined with resistance testing -- the information obtained from drug-level measurements and phenotypic assays can enable providers to individualize antiretroviral therapy to a degree that was heretofore impossible, so that HIV-infected patients receive what are, for them, the optimal doses of the most effective drugs.
While the concept of TDM has been effectively used to treat other disease states, many clinicians are persuaded that the treatment of HIV infection is too complex to allow for the individualization of a patient's antiretroviral regimen based upon circulating drug levels. Pharmacokinetic differences between antiretroviral classes, inter- and intra-patient variability, inter-laboratory variability, assay limitations -- and, of course, less-than-perfect adherence to therapy -- are often cited as drawbacks to the full clinical application of TDM in combination with resistance testing.
Until quite recently, office-based clinicians had limited access to laboratories that could measure circulating drug levels or determine viral phenotypes, and the use of these tests was largely restricted to research-related activities. Commercial tests are now available, however, and so are data from a number of recent studies that demonstrate the potential benefit of TDM in clinical practice. These findings also offer evidence that the use of plasma drug-level measurements in conjunction with phenotypic testing provides more useful information that either test does on its own.
The principles of pharmacokinetics (PK) and pharmacodynamics (PD) have been used to manage a number of diseases (see "Glossary of Terms"). Our ability to determine optimal drug levels for compounds such as the aminoglycosides, vancomycin, anti-epileptics, and theophylline helps us to prescribe effective, non-toxic doses of these drugs. In each of these instances what we know about the PK and PD of the individual compounds enables us to use them in the safest and most effective way.
Pharmacokinetic and pharmacodynamic principles also contribute to the management of
Clinical Limitations of TDM
While the clinical use of TDM to guide therapy for HIV-infected patients has been widely discussed, limitations in both laboratory methods and clinical experience have hindered the widespread adoption of this new technology (see Table 1). To begin with, the commercial laboratories that do phenotypic susceptibility testing only report concentrations that inhibit 50% of the virus -- the so-called IC50 of the drug. Ideally, of course, clinicians should know what concentrations are required to inhibit 95% or more of a viral isolate. At present, however, the limitations of the assays themselves prevent labs from reporting reliable results at these higher inhibitory values.
Moreover, the currently available drug-level tests only report total drug levels. Given that it is a drug's non-protein-bound moiety that is its active component, these tests should ideally report the free-drug level, rather than the sum of the free-drug and protein-bound drug levels. Before free-drug levels can be routinely and reliably obtained, laboratory techniques will have to be improved -- by ultra-filtration of samples, for example. And given that only 1% of replication-competent HIV is found in the plasma, methods to assess viral levels within different compartments of the body must also be developed.
Another practical limitation of TDM arises from the difficulty of collecting a "true" trough level, or Cmin, from an HIV-infected individual. This finicky and demanding clinical task requires that a patient be available for a blood draw at a particular sampling time during a dosing cycle -- an onerous obligation on the patient's part, and a tedious and exacting chore for that patient's care providers.
The clinical complexity of obtaining accurate trough levels has led some investigators to evaluate the feasibility of substituting population-based dosing curves for individual Cmin measurements (1). These population curves are derived by obtaining serial blood levels from large numbers of patients over a standard dosing interval, thereby providing clinicians with mean blood levels at fixed time points during that dosing cycle. Using this methodology, a clinician can draw a random blood sample from a given patient during that patient's dosing cycle, and then compare this level with the mean for the study population. If the patient's level is low, the care provider will want to consider increasing the dose; if the level is elevated, he or she will want to consider lowering the dose.
Researchers evaluating this technique have experienced widely varied clinical results. While this method permits sampling at any point during the dosing interval, it is flawed in that in assumes every patient has the same viral subtype and the same IC50.
The reliability of TDM is also compromised by concerns about how accurately in-office measurements of circulating drug levels reflect out-of-office levels. We know that there is an absolute correlation between adherence to therapy and the success of antiretroviral regimens (2, 3). What we don't know, with the same degree of assurance, is how adherent patients actually are. Most of us rely on our patients to report their degree of adherence to their assigned therapy -- a method of data-gathering that is notoriously unreliable.
What's more, patients whose compliance falters badly between clinic appointments often comply scrupulously with their dosing schedule in the days immediately preceding their next appointment. And because many of the protease inhibitors have short half-lives, drug levels obtained from these patients in the clinic will suggest that they are maintaining therapeutic serum concentrations of these potent agents between visits, when in reality their adherence is so poor that they rarely achieve the levels necessary to inhibit viral replication.
In the last analysis, however, the greatest barrier to applying TDM to the long-term management of patients with HIV disease may simply be our inability, at this juncture, to extrapolate clinical outcome from the results of drug-level monitoring. For diseases such as asthma and epilepsy we have well-defined therapeutic indices for various agents. Unfortunately, where HIV is concerned, the relationship between a given dose of a drug, the plasma level that dose produces, and the efficacy -- and toxicity -- of that dose has yet to be determined for most agents, although progress is being made.
Measuring Drug Levels
Both serum and plasma drug
levels can be obtained for all of the nucleoside
Furthermore, analyzing blood samples to determine the triphosphate concentration of a particular nucleoside analog requires highly specialized laboratory techniques that are not widely available, and this practical limitation effectively rules out TDM for this drug class, since serum and plasma levels of these agents provide little useful clinical information.
Fortunately, plasma levels can be readily obtained for all of the protease inhibitors, and this information can be applied directly to the management of seropositive patients. A number of studies have demonstrated a strong correlation between such pharmacokinetic parameters as AUC and Cmin and the clinical effectiveness of the protease inhibitors. With the exceptions of amprenavir and loprinavir, these agents display relatively short plasma half-lives (8).
A number of small in vitro studies have suggested that the protease inhibitors possess persistent antiretroviral activity within cells even when drug levels have fallen so far that they can no longer be measured (9). These data have yet to be validated in clinical practice, however. All but one of these agents are highly protein-bound. As a result, serum concentrations of free, biologically-active drug that range between 2% and 10%. The exception is indinavir, which is only 65% protein-bound and has a free component of 35% (8).
Plasma levels can also be obtained for all of the non-nucleoside reverse-transcriptase inhibitors, and this information can also be used to guide treatment. In contrast to the protease inhibitors, the NNRTIs have long half-lives that result in relatively constant plasma levels. In one study, plasma drug levels of nevirapine were so constant that it was possible to predict the twelve-hour AUC using a single time point (10).
Despite these constant levels, wide inter-patient variability in NNRTI plasma levels has been recorded, and these individual variations have been found to correlate with clinical failure (11-13). The biggest limitation to using TDM for the NNRTIs is not their constant plasma levels but rather their narrow therapeutic index. In situations where a "low" level of one of these agents has been identified, the clinician may feel constrained about increasing the dosage to a more effective level out of concern about the increased risk of toxicity.
Wide inter-patient variability is not
limited to the NNRTIs; it is seen in the AUC and Cmin values of patients
who are taking protease inhibitors, even among those assigned to
Although intra-patient variability also occurs in TDM, the range of variability is
not wide. Studies involving indinavir (8), amprenavir (16), and nevirapine
(10) have found modest
In addition to inter- and intra-patient variability in circulating levels of antiretroviral drugs, clinicians should also expect to encounter considerable variations in the serum and plasma levels obtained through different laboratories and different assay methods. In a recent study, eight laboratories in Europe and North America were sent spiked samples of protease inhibitors in order to assess those laboratories' accuracy within 20% of the reference compound (17). The researchers found that these eight laboratories were inaccurate anywhere from 14% to 58% of the time, depending upon which protease inhibitor was evaluated. Work done in our own labs has shown that variability can exist between serum and plasma, among different types of matrixes used for standardizing the assay, and even among different types of reference compounds used to standardize the assay (18). Given that there are no established guidelines for TDM, clinicians should be sure that the accuracy of a laboratory's assays has been validated by an outside agency before sending samples for testing.
Studies of PK/PD show that low levels of certain agents correlate with impaired viral suppression, increased risk for the emergence of drug-resistant viral strains, and the potential for a less durable clinical response. Although these findings are disturbing on their face, they can -- and should -- be exploited to guide therapy. When patients can be successfully shifted from the low end of the variability curve to the upper end of the curve without incurring undue toxicity, the potential exists to provide those patients with more potent and more durable antiretroviral therapy.
The Evidence Favoring TDM
Despite the very real limitations described above, TDM can be used to optimize and individualize the clinical management of HIV-infected individuals. The first trial to assess PK/PD in seropositive patients, conducted five years ago, measured the relationship between circulating levels of ritonavir and the development of resistant viral isolates. Molla and colleagues tallied the number of viral mutations that emerged each week among patients who were receiving ritonavir monotherapy (19). An inverse relationship between the Cmin and development of resistance was observed: the greater the ritonavir Cmin, the lower the number of resistant viral isolates. In addition, elevated Cmin values delayed the emergence of multiple mutations in the viral isolates.
In the last five years the effects of inter-patient variability on clinical outcome of HIV-infected individuals have been evaluated. Both total daily drug exposure (e.g. AUC) and concentration-specific measurements (e.g. Cmin) have been found to correlate with degree of viral suppression. In a clear demonstration of the correlation between AUC and efficacy, Murphy et al. evaluated the relationship between the inter-patient variability of circulating levels of indinavir and clinical efficacy (20). A total of 24 HIV-infected patients who were treated with 800 mg of indinavir every 8 hours had serial blood-level measurements assessed after a seven-day run-in period. A seven-fold difference in indinavir AUC was observed within this cohort, and that difference in total daily drug exposure was found to relate directly to suppression of viremia: the larger the exposure, the more pronounced the reduction in replication. In addition, the investigators found a correlation between both Cmin and Cmax values and viral suppression.
In a similar study, Schapiro and colleagues evaluated the effects of saquinavir exposure on decreases in viral load (21). A cohort of 40 HIV-infected individuals who received monotherapy with 3600 to 7200 mg of saquinavir per day had serial blood samples taken after 28 days on therapy. For each patient, the AUC was plotted against the decrease in plasma HIV RNA levels from baseline. The authors noted that a strong correlation was found between drug exposure and decrease in viral load.
Acosta et al. assessed the effects of indinavir concentrations on antiretroviral activity among 23 HIV- infected patients seen in a university-affiliated clinic (22). These treatment-naïve patients were assigned indinavir in combination with two nucleoside analogs, and their indinavir levels were measured after a standard 800-mg oral dose. When the investigators compared Cmin values in patients with detectable viral load measurements (n=9) against Cmin values for those subjects with undetectable viral loads (n=14), the latter were found to have significantly higher mean indinavir trough levels (0.55 µM versus 0.07 µM, respectively; p = 0.007).
These findings led Fletcher and colleagues to
investigate the impact of manipulating individual patient drug levels
(23). In this small trial, 24 antiretroviral-naïve patients with viral
loads greater than 5,000 copies/mL were randomly assigned to open-label,
At six months, 91% of the patients in the concentration-controlled arm achieved undetectable HIV RNA levels (< 50 copies/mL), versus only 69% of the patients in the standard-dose arm. In addition, patients randomized to the concentration-controlled arm reached undetectable levels significantly faster than patients who were receiving standard therapy (110 days versus 176 days; p = 0.056). Given that the magnitude of viral load decline predicts durability of response, these data suggest that dosage modifications can affect therapeutic efficacy and, potentially, improve durability of response.
In addition to improved antiviral efficacy, elevated drug levels of antiretroviral agents appear to improve CD4 cell responses. When Fletcher et al. evaluated changes in CD4 counts from baseline among 19 of the 24 patients in their study, they discovered that CD4 counts at weeks 52 and 80 correlated significantly with indinavir Cmax, and to a lesser extent with indinavir Cmin (25). The authors concluded that both the Cmin and Cmax of indinavir might be important in determining the virologic and immunologic response to antiretroviral therapy.
The impact of drug levels on clinical outcome is not limited to the protease inhibitors. Joshi and colleagues conducted a retrospective analysis of the relationship between efavirenz Cmin levels and clinical efficacy among patients participating in five Phase II clinical trials (12). Among patients with greater than 80% adherence, treatment failure -- which the investigators defined as the inability to attain an HIV RNA level below 400 copies/mL -- was found to be roughly three times as frequent in patients with Cmin values of less than 3.5 µM than it was in patients with Cmin values greater than 3.5 µM.
These findings support the belief, held by a growing number of experts on HIV infection
and its treatment, that there is a causal relationship between inter-patient differences in circulating drug levels and clinical response to
therapy. The unanswered question, at this point, is whether individual
differences in susceptibility to antiretroviral drugs, and corresponding
differences in the degree of viral suppression that these individual
patients achieve -- differences, that is, in inter-patient pharmacokinetics
The Merck 069 trial provides us
with a particularly pertinent example of the value of TDM as a potential
indicator, if not a predictor, of clinical outcome. This trial compared
the efficacy of ZDV and 3TC in combination with one of two indinavir
regimens: 2 400-mg capsules three times a day, or 3 400-mg capsules twice
a day (26). Preliminary data on 287 patients after 16 weeks of follow-up
showed equivalent rates of reduction of viral load to less than 400
copies/mL (72% versus 78% for the b.i.d. and t.i.d. dosing regimens,
respectively). However, in an earlier study the Cmin levels of patients
taking indinavir only twice a day were low enough to raise concerns that
those trough levels might be insufficient to inhibit 90% of
And, in fact, an interim analysis of the pharmaco-kinetic data on the first 87 patients treated for 24 weeks found a significant divergence between the treatment arms: 91% of the patients who were taking indinavir three times a day had viral load measurements less than 400 copies/mL, versus only 64% of the patients who were taking indinavir twice a day. These findings led Merck to stop the trial and recommend that all patients in the b.i.d. arm be switched back to t.i.d. dosing. These study data reinforce the concept that pharmacokinetic parameters have a direct impact on pharmacodynamic responses.
The outcome of the Merck 069 trial was a disappointment not only to the makers of indinavir but to all patients who take this widely prescribed agent and to all clinicians and other care providers who treat them -- because if it had proved possible to take this potent protease inhibitor safely twice a day, an onerous aspect of indinavir administration -- strict q8 dosing on an empty stomach -- would have been eliminated. As all of us who treat HIV-infected individuals are well aware, anything that simplifies a patient's dosing regimen -- by reducing the number of doses or the number of pills, the dietary restrictions or the rehydration requirements -- is likely to improve adherence, and that in turn improves clinical response.
What the Merck 069 trial did demonstrate is the importance of Cmin data as a potential marker of clinical outcome. The work of Durant et al. emphasizes just how important such trough measurements are as predictors of clinical response to therapy. This group assessed the impact of protease-inhibitor trough levels on HIV RNA changes from baseline in 81 patients who participated in the Viradapt study (28). Optimal plasma concentrations were defined as trough levels two-fold or more above the IC95; levels less than two-fold above the IC95 were categorized as suboptimal concentrations. After 48 weeks, the mean change in viral load from baseline was significantly greater in individuals with Cmin levels two-fold or more above the IC95: a 1.28 log reduction, compared with only a 0.36 log reduction among patients with suboptimal concentrations.
The findings of Durant and coworkers have broad implications for the treatment of HIV infection, because resistance to specific drugs -- and potential cross-resistance to drug classes -- develops when Cmin values drop to suboptimal concentrations. Data drawn from genotypic and phenotypic studies in treatment-experienced patients reveal that many of these patients evince elevations in IC50 values not just for agents to which the patients have been exposed but for agents to which they have not been exposed (29). In therapy-naïve patients, for example, the protease inhibitor ABT-378 achieves plasma concentrations that are well in excess of the wild type IC50. When ABT-378 is given to individuals who have been previously treated with one or more protease inhibitors, however, the IC50 values are often increased (29) -- although this agent appears to retain a significant degree of activity against a number of multidrug-resistant viral isolates (30).
Although ABT-378 does achieve more than adequate plasma concentrations, there is a breakpoint at which the plasma levels achieved are no longer sufficient to suppress viremia. In a study conducted by Kempf et al., 93% of the treatment-experienced patients whose IC50 increase from baseline was less than 10-fold had a positive virologic response to this new protease inhibitor, as compared with a 50% response rate in patients whose IC50 increase was greater than 40-fold (29).
A number of recent studies have investigated the effect of using two protease inhibitors to achieve higher and more consistent serum concentrations than can be achieved with either drug alone. In most cases, this strategy has involved the addition of low-dose ritonavir, for pharmacodynamic rather than pharmacotherapeutic reasons.
Condra and colleagues evaluated the potential role of dual-protease regimens on viral isolates that demonstrated resistance to one or more of the drugs in this class (31). This in vitro study compared plasma levels of various protease inhibitors -- when those agents were given in combination with low-dose ritonavir -- with protein-corrected IC95 values for both wild-type and resistant viral isolates. When Cmin values for each protease inhibitor were evaluated against mutant viral isolates, the combination of indinavir at the standard dose of 800 mg and ritonavir at a dose of 200 mg appeared to provide sufficient drug concentrations to overcome even highly-resistant viral strains. This finding is not altogether surprising, given that indinavir has the lowest protein-binding rate of all drugs in this class and therefore is able to inhibit 95% of the viral isolates at lower serum concentrations than other protease inhibitors.
The authors concluded that combination therapy with indinavir-ritonavir and possibly amprenavir-ritonavir may provide effective salvage in many instances of protease-inhibitor failure, even in individuals who have developed resistance to several drugs in this class. While the in vitro data compiled by Condra and coworkers suggest that indinavir-ritonavir based regimens may have activity against highly-resistant viral strains, it must be emphasized that these encouraging findings have yet to be validated in clinical trials -- and that even if regimens based on indinavir-ritonavir and amprenavir-ritonavir do prove clinically effective in some patients with multidrug-resistant viral isolates, wide inter-patient variability may limit drug exposure in some patients, thereby affecting virologic response.
Applying the Evidence to Clinical Practice
No one yet understands why one patient responds well to a given regimen, another has only a partial response, and a third fails to respond at all to the same regimen. All care providers who treat seropositive patients have encountered individuals who are scrupulously adherent to their assigned antiretroviral regimen but who nevertheless fail to experience a pronounced or durable antiviral effect from what should be a maximally suppressive regimen. The critical question is why.
ln situations where poor adherence is not a contributing factor, there are three critical -- and highly variable -- factors that influence response to therapy: inter-patient pharmacokinetic differences; differences among viral subtypes; and inter-patient variability in immune responses (Figure 1).
Variability in viral subtypes between patients has been an under-appreciated factor in the management of HIV-infected individuals. We know that treatment-naïve individuals often display highly sensitive virus, as indicated by very low IC50 values. We also know that heavily pretreated patients often exhibit less susceptible viral strains, as evidenced by increased IC50 values. In addition, recent data have shown that antiretroviral regimens that are not fully suppressive result in sequential elevations in viral phenotypes (32). When this information is coupled with the wide inter-patient variability that is seen whenever plasma levels of protease inhibitors or NNRTIs are measured, it is apparent that individual patients have dramatically different levels of exposure to individual drugs, even when adherence is optimal.
Just how dramatic that difference in exposure can be
is revealed in the phenotypic data obtained from two patients seen in our
clinic. Both of these patients were started on a
These PK/PD differences may well explain why population-based dosing curves fail to reliably predict clinical success. Such methods assume that each patient in a given population has the same virus, and hence the same degree of drug exposure. As a growing body of scientific evidence indicates, this is almost never the case. What should be evaluated is not the relationship of a given sample to a population-based dosing curve, but the relationship of that sample to the individual patient's viral subtype (Figure 2).
Properly undertaken in a sufficient number of HIV-positive patients, this comparison will produce a classic bell curve -- one in which "low" drug level/IC50 ratios have modest antiviral effects, "therapeutic" drug level/ IC50 relationships produce maximally suppressive antiviral responses, and "excessive" drug level/IC50 relationships result in toxicity, leading to poor adherence and modest antiviral effects.
When the relationship between therapeutic drug levels and IC50 values has been more fully delineated, it will in all likelihood reveal two important subgroups under the bell curve, one fixed and one shifting. The former group will be made up of those individuals who experience durable, long-term responses to antiretroviral therapy; the latter, of patients who achieve prompt and near-complete suppression of viremia, only to have their HIV RNA levels begin to climb inexorably upward (Figure 3). The critical question that has yet to be answered is: What level of drug-to-virus exposure is most likely to predict success or failure?
To begin to answer that question, clinicians must learn to interpret the results of both phenotypic resistance assays and drug-level monitoring (see "Clinical Considerations in Therapeutic Drug Monitoring" and "Michael T., a Longterm Survivor Who Is Failing His Fifth Salvage Regimen"). This information will enable care providers to choose the antiretroviral regimens that are likely to achieve maximal suppression of viremia in individual patients, and to titrate those regimens to their most effective -- and least toxic -- doses.
When phenotyping is combined with TDM,
care providers are able to get a much clearer idea of what is happening at
the cellular level in each patient who undergoes such combination testing.
Take the case of an extensively pretreated patient who was recently
evaluated at our clinic. This individual's phenotypic resistance assay
revealed that he had, over the previous four years, developed
significantly decreased susceptibility to all of the F.D.A.-approved
protease inhibitors. (The patient's decreased susceptibility ranged from a
22-fold reduction in susceptibility to amprenavir to a 171-fold reduction
in susceptibility to nelfinavir.) However, when we compared his IC50 and
Cmin we recognized that we could easily obtain levels well in excess of
this inhibitory concentration using a regimen that combined amprenavir
Evaluating a phenotype without a drug level -- or a drug level without a phenotype -- fails to provide the clinician with all of the information he or she needs to properly interpret the results of either of these tests. This is true of all patients -- because individual pharmacokinetic responses are so variable, even in treatment-naïve patients -- but it is especially true of heavily-pretreated patients. In patients who have been on a succession of antiretroviral regimens -- monotherapy, dual therapy, and various three- and four-drug combinations -- multiple codon mutations are the rule, not the exception. In such patients, elevated IC50 values are the norm, and it is only when the patient's IC50 is paired with the results of drug-level monitoring that the amount of actual drug exposure to particular antiretroviral agents can be determined.
With the advent of commercial tests that measure both plasma drug levels and susceptibility to therapy, clinicians finally have the tools they need to evaluate an individual patient's responsiveness to individual antiretroviral agents, information that can be used to devise truly individualized drug regimens. We now have drugs that can suppress viral replication to levels so low that HIV RNA cannot be detected by the most sensitive commercial assays -- and we have the clinical tools we need to devise patient-specific regimens that will achieve that objective.
The value of the data derived from these new tests will be only as good as the clinician's ability to interpret those data, of course -- and accurate interpretation will depend on familiarity with both the advantages and limitations of current methodologies. As more data are derived -- and as more clinicians gain comfort with the new technology -- significant advances in our understanding of the clinical implications of information derived from TDM and resistance testing will undoubtedly occur.
At the moment, these new tests offer us a unique opportunity to evaluate individual responses to antiretroviral therapies. And this clinical advantage offers us an opportunity to prevent treatment failures and increase the durability of responses more fully and more effectively than ever before. We badly need well-designed clinical trials and broad-based clinical experience to validate these theories and principles. Until such data are derived, clinicians should interpret the information they obtain from drug-level tests and phenotypic assays with care. However, our early experience with these new clinical tools teaches us that these tests provide potentially valuable assistance in devising antiretroviral regimens that achieve the greatest possible suppression of viremia with the lowest incidence of therapy-limiting toxicities.
Glossary of Terms
AUC (Area Under the Curve)
Andrew D. Luber, Pharm.D., is Executive Director of Pacific Oaks Research, Beverly Hills, CA.
W. David Hardy, M.D., is Associate Clinical Professor of Medicine, UCLA School of Medicine, Los Angeles, CA.
Back to the February 2001 HIV Newsline contents page.
This article was provided by San Francisco General Hospital. It is a part of the publication HIV Newsline.