Despite intense efforts by the pharmaceutical and biotech industries, translating the preclinical identification of new targets into new medicines that will benefit patients and address the opioid crisis has been largely unsuccessful, said John Dunlop, vice president of neuroscience research at Amgen. Preclinical models may fail to predict clinical efficacy for many reasons. To help companies move from the preclinical to the clinical space,
Walter Koroshetz said new tools, especially biomarkers to inform Phase II trials, are needed. Objective measures of pain, as well as measures of compliance, such as whether a trial participant is adhering to the study protocol, need to be incorporated into clinical trials, said Nora Volkow, noting that compliance issues, which can be easy to avoid in preclinical studies, may contribute to clinical trial failures.
In drug discovery for any condition, typically many targets are identified and supported by genetic associations and studies in animal models, said Andrew Ahn. He said the next step requires an important translational leap to enter the framework of clinical opportunity within the pharmaceutical industry. Biomarkers can offer an opportunity to reach across that divide, he said, noting that partnerships are essential in this area given that pharmaceutical companies may not have all the expertise needed to identify mechanisms, connect them to a disorder, and develop tools and assays to move from a mechanistic hypothesis to “druggable” targets and eventually, novel compounds.
Many workshop participants noted that objective markers are needed to demonstrate clinical efficacy of pain medications in development. Currently, the most commonly used pain assessment measures are subjective visual analog and numerical pain rating scales, said Tor Wager.
Although these tools are important, he said they are influenced by many factors beyond nociception, such as prior beliefs and experiences, emotional reactions, the reporting context, and cultural factors. These factors thus can have a substantial impact on clinical trials and treatment, said Wager. For example, in a trial of a dopaminergic-promoting gene therapy approach for Parkinson’s disease, the trial failed because patients receiving the placebo improved to the same degree as those receiving the active drug (Olanow et al., 2015), which may be due in part to the subjective nature of the outcome assessment. Indeed, said Wager, placebo response in clinical trials has been increasing over the years while drug effects have not increased in magnitude, shrinking the drug–placebo difference and making it more difficult to get drugs to market (Tuttle et al., 2015).
A Case Study of Biomarker-Based Drug Development
As an example of how biomarkers can fuel drug development, Ahn described the development of the monoclonal antibody galcanezumab, which Eli Lilly and Company is currently developing for the treatment of migraine and cluster headache. In 1990, investigators at Prince Henry, Prince of Wales Hospital in New South Wales, New Zealand, demonstrated an increase in blood levels of the vasoactive peptide called calcitonin gene-related peptide (CGRP) in patients with migraine, suggesting that migraine might be caused by an activation of sensory neurons in the head (Goadsby et al., 1990). Later they showed that triptan antimigraine drugs normalized blood levels of CGRP (Edvinsson, 2001), prompting the successful search by a collaboration of academic and pharmaceutical investigators for CGRP antagonists, said Ahn (Ho et al., 2008, 2014; Olesen et al., 2004).
The success of CGRP antagonists in reducing migraine attacks led to the next important step, said Ahn, the search for an antibody that neutralizes CGRP. Setting the optimal dose to test in clinical trials of these monoclonal antibody therapeutics has been facilitated in part by the availability of a biomarker of peripheral target engagement, which involves injecting into the skin a small amount of capsaicin and then measuring to what extent the investigational drug blocks an increase in dermal blood flow (Sinclair et al., 2010). Characterizing pharmacokinetic and pharmacodynamic properties of anti-CGRP antibodies required other assays to measure blood levels of both the antibody and CGRP in relation to migraine symptoms. Ahn showed data from a Phase IIb study of galcanezumab, which demonstrated target engagement and a drug-exposure-biomarker relationship using plasma CGRP level as the biomarker (Kielbasa et al., 2016). Investigators at Lilly recently demonstrated the high sensitivity of two CGRP immunoassays developed by Meso Scale Discovery and Quanterix (Chai et al., 2016).
In short, CGRP biomarkers facilitated the development of this drug by confirming the therapeutic hypothesis, demonstrating target engagement, and tailoring and predicting response, said Ahn.
To qualify as a biomarker, an objective measure of the pain process should reflect the experience of pain, said Wager. He described imaging approaches that are providing new insight about the physiology of the pain
response and that may be used to objectively assess pain. For example, animal studies have demonstrated changes in neuroplasticity in several different brain regions that are linked to the chronic pain response after nerve injury. Human studies are needed to apply these learnings to human brain circuitry, he said. In some cases, this circuitry has already been defined, for example, the CGRP pathway described by Ahn. These biomarkers enable drug developers and clinical trialists to study penetrance, pharmacodynamics, and efficacy, said Wager. The ultimate goal is to identify brain patterns that discriminate drug from placebo, he added. A proof-of-concept study by Duff and colleagues (2015) shows promise in this regard, said Wager. While these approaches can be used for both drug discovery and repurposing, he noted that a recent consensus statement of the International Association for the Study of Pain concluded that the potential of imaging approaches has not yet been fully realized (Davis et al., 2017).
Wager and others use functional magnetic resonance imaging (fMRI) to characterize pain circuitry and drug actions. He and his colleagues identified an fMRI-based neurological pain signature that captures the neurobiological correlates of evoked pain, starting with heat-induced pain, but then generalizing to other pain conditions (Wager et al., 2013). To validate this approach, they have collaborated with other research groups around the world, testing the construct in different populations and pain types (e.g., electrical, heat, mechanical). These studies suggest that the signature is sensitive and specific to different pain types, as well as being resistant to the placebo effect.
Wager and colleagues have also shown that pain may be mediated by distinct circuits. One of these circuits in the ventromedial prefrontal cortex (vmPFC) plays an important role in learned avoidance—for example, learning to avoid choices that involve pain—and is susceptible to cognitive interventions (Woo et al., 2015). He added that in humans, evidence also suggests that the shift from acute to chronic pain involves corticostriatal circuitry and a shift from classic nociceptive targets to more emotion-focused correlates of pain in the vmPFC (Baliki et al., 2012). This shift could present a problem for treatments focused on the periphery and the spinal cord, he said.
Neuroimaging thus may have implications for treatment selection and development of preventive approaches, said Wager. These objective imaging measures will augment but not replace pain reports, he said, noting that they can provide insight into mechanisms, define targets for intervention, define biotypes or groups of participants who have different neurological and neurophysiological bases for disorders, and measure
physiological component processes. Wager added that biomarkers do not always have to measure objective pain, but can also measure other outcomes such as cognitive impairment, fear, and avoidance. By linking neuroimaging to other measures of genetics, behavior, heart function, inflammation, and other markers, a clearer understanding of pain may emerge that can help validate new therapies in development, he said.
Another imaging-based pain signature was described by Sean Mackey, chief of the Division of Pain Medicine at Stanford University, working with the Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP) research network.1 A recent publication from MAPP compiled clinical data as well as structural and fMRI data from more than a thousand patients at seven sites, including patients with urological chronic pelvic pain (localized and widespread) (Kutch et al., 2017). Using machine learning approaches, a signature was identified involving increased connectivity in the frontoparietal cortex, which predicted who would get better and who would worsen, said Mackey. This type of biomarker could be used to identify and exclude patients who are likely to improve even without treatment from clinical studies, he said. Similar approaches have also been used to identify brain-based biomarkers for depression, which he said drives home the point that brain biomarkers and biological markers are needed, along with symptom reporting, to identify condition clusters and tease apart factors that affect response to treatment.
Many pain disorders are associated with aberrant expression of small, non-coding RNAs called micro RNAs (miRNAs), according to Seena Ajit. Approximately 10 years ago, scientists discovered that these regulators of gene expression circulate in the serum, suggesting that they could represent blood-based biomarkers of disease (Chen et al., 2008; Mitchell et al., 2008). Ajit and colleagues have been studying miRNA expression in complex regional pain syndrome (CRPS), a heterogeneous chronic progressive neurological disease characterized by severe pain. Using quantitative polymerase chain reaction analysis, they showed that 18 miRNAs are differentially expressed in CRPS patients, clustering into three groups based on expression levels of these 18 miRNAs (Orlova et al., 2011). Three inflammatory markers were also elevated in patients, and the levels of these
markers correlated with pain levels, suggesting that miRNA profiling could be used for patient stratification, according to Ajit.
Her lab has also explored miRNA expression in several different rodent models of neuropathic and inflammatory pain, where they identified characteristic expression profiles and two miRNAs that were common with CRPS patients (Qureshi et al., 2016). Identification of these miRNA biomarkers enabled Ajit and colleagues to do mechanistic studies that elucidated disease processes and demonstrated that miRNA profiles were altered by therapeutic intervention, said Ajit. Similarly, in patients with CRPS, good and poor responders to ketamine treatment can be differentiated by change in their miRNA profiles, suggesting the feasibility of using miRNA signatures as prognostic biomarkers (Douglas et al., 2015).
Ajit and colleagues were also able to show that one of the most down-regulated miRNAs in patients compared to controls, miR-939, targets pro-inflammatory genes. Reduced levels of miR-939 result in an increase in expression of these genes, amplifying the pro-inflammatory pain signal transduction cascade (McDonald et al., 2016).
Ajit’s lab has also shown that miRNAs are packaged into exosomes (secreted extracellular vesicles present in bodily fluids), which can cross the blood–brain barrier and facilitate transport and intercellular communication (McDonald et al., 2014). The field is young, she said, so many questions about exosome biology have yet to be answered. She noted that the stability of exosomes in serum makes retrospective studies possible and suggested using serum and plasma samples from failed clinical trials, where the therapeutic outcome is known, to search for miRNA signatures that could be used for patient stratification. For biomarker discovery, she advocated obtaining miRNA profiles beginning in Phase I studies, which may enable later stage trials to be conducted in defined patient subgroups. She added that a centralized database of miRNA profiles and clinical phenotypes could accelerate drug development, but would require adherence to standardized acquisition, analytical, and data normalization protocols.
A targeted approach to therapy development when drugs are developed and tested in strongly phenotyped, well-defined populations could represent a potentially useful foundational strategy on which to build a public–private partnership, suggested Koroshetz. Sharon Hertz, director of
the Division of Anesthesia, Analgesia, and Addiction Products at the Food and Drug Administration’s (FDA’s) Center for Drug Evaluation and Research, agreed that defining a subpopulation that is more likely to respond to a drug, and then demonstrating efficacy in that subpopulation, may be a good drug development strategy. However, if a compound in development does not target a specific aspect of a specific disease, and the drug is anticipated ultimately to be used in a more general way, there are dangers to that approach, said Hertz.
For example, anti-nerve growth factors (anti-NGFs) have been developed as analgesics (Chang et al., 2016). Preclinical and clinical testing of the anti-NGF antibody tanezumab suggested efficacy in improving pain and function in patients with osteoarthritis (Miller et al., 2017). However, clinical studies revealed a serious and unexpected adverse effect not found in the preclinical studies: patients treated with the drug had a higher incidence of joint destruction requiring replacement, including in shoulder joints that rarely need replacement, said Hertz. It was determined that there was a destructive arthropathy associated with the active treatment arms, stalling development of the drug while the factors contributing to this risk were determined. Even with this information, calculating a risk–benefit ratio will be challenging, said Hertz, because patients vary considerably in their tolerance for risk and desire for treatment benefits. For example, she noted that patients with debilitating trigeminal neuralgia may be willing to use a drug product with greater risk than patients with milder forms of pain. She also commented that had the anti-NGF antibody studies started with trigeminal neuralgia, for which the population is much smaller, the adverse signal might not have even been detected until the postmarketing stage.
Katherine Dawson, vice president for Late Stage Clinical Development in Pain, Neuromuscular, and Rare Diseases at Biogen, said her company chose to pursue treatments for trigeminal neuralgia precisely because there is an urgent need for therapies due to the severity of the disorder (i.e., increased risk of death due to suicide). She acknowledged that for serious conditions like this, treatments may be introduced even before the safety profile has been fully elaborated. Dawson elaborated that no safety profile is fully elucidated by the clinical trials, as uncommon and rare events will be missed; however, the risk–benefit profile that supports therapeutic development should not be the same as for a treatment focused on an illness or disorder that is not as serious or severe. With this approach, the full population of persons with the disorder will have an opportunity to take the therapy, rather than just those who meet the inclusion and exclusion
criteria, she said. Dawson noted that Biogen faced a similar dilemma when three patients being treated with natalizumab in clinical trials developed progressive multifocal leukoencephalopathy and two died. The drug was temporarily taken off the market, but was reintroduced with restrictions. Developing treatments for trigeminal neuralgia is particularly challenging, she said, because there are no well-defined endpoints, patients experience both spontaneous and evoked pain, and the severity fluctuates. She added that in order to pursue interventions in this population, Biogen has worked closely with patient groups and the FDA to identify what endpoints are meaningful to patients.
Moving forward with drugs that target novel pathways thus requires careful and thoughtful consideration, said Hertz. Ken Verburg, senior vice president at Pfizer and asset team leader for tanezumab, agreed that developing a drug in a small population can be risky. In the pharmaceutical industry, a failed early study means a failed drug program, he said, even though the drug may be beneficial in certain situations. To avoid this problem, he advocated testing a drug in three different models when the failed trial scenario is low. For example, is a drug intended for pain useful for molar extraction, osteoarthritis, and postherpetic neuralgia? If a network of physicians skilled in doing clinical investigational work was available, this pathway might be traversed fairly quickly, he said. In addition, Verburg cited the need for more exploratory work developing new clinical trial study paradigms.
William Maixner described another program that has used deep phenotyping of thousands of individuals, along with machine learning and clustering methodologies, to identify common and unique pathways in chronic pain conditions. Using a comprehensive array of biopsychosocial measures, the Orofacial Pain: Prospective Evaluation and Risk Assessment (OPPERA) study identified three clusters of individuals distinguished by pain sensitivity and psychological distress (Bair et al., 2016). These clusters could be reproduced by taking into account only four variables—semantic awareness, anxiety, depression, and pain sensitivity—which Maixner suggested could be incorporated into a short questionnaire given in the clinic. His team is now conducting pragmatic trials after classifying individuals with this set of variables. They are also working with Luda Diatchenko to identify molecular signatures associated with each group, which could suggest subgroup-specific targets, leading to the creation of new chemical entities or the repurposing of existing compounds to target these clusters.
High on the priority list of the Federal Pain Research Strategy is preventing the acute-to-chronic pain transition, said Koroshetz. He noted that the National Institutes of Health Common Fund team has taken on this topic as a potential project. Theodore (Ted) Price, director of Systems Neuroscience at The University of Texas at Dallas, described the transition from acute to chronic pain as a plasticity-driven event that persistently alters the responsiveness of the pain system. Important mechanisms involved in this transition include neuroimmune and peripheral immune interactions and neuronal plasticity. Phenotypic changes in subsets of neuronal, immune, and other types of cells may drive this transition, he said. Important questions he raised that emerge from this model include whether targeting these mechanisms will reverse the transition, returning someone to a pain-free or normal acute pain state, and whether treatments currently being used impact the transition or conversion to chronic pain.
Price suggested that preclinical models can provide insight into the acute-to-chronic pain transition and possibly illuminate ways to stop, reverse, or prevent it from occurring. One of these models, the hyperalgesic priming model, posits that an acute inflammatory insult, such as injection of prostaglandin E2 (PGE2), triggers transient hyperalgesia as well as long-lasting hypersensitivity (Reichling and Levine, 2009). This reaction can be blocked with a single dose of a μ-opioid agonist; however, repeated exposure to a μ-opioid agonist induces opioid tolerance. (Joseph et al., 2010). Price noted that simply exposing an animal multiple times to the opioid can also cause conversion to the chronic pain state (Araldi et al., 2015). His lab has also shown that hyperalgesic priming can be blocked by ablating neurons in the dorsal horn that express neurokinin 1 (NK-1),2 suggesting that these neurons are required for hyperalgesia priming (Kim et al., 2015). However, Price said that if animals are primed prior to blocking NK-1 neurons, there is no effect on priming, arguing against the use of NK-1 antagonists as a treatment for intractable pain.
Another preclinical model, the “latent sensitization” model, also suggests that the transition to chronic pain involves priming that causes long-
2 NK-1 receptors have been studied in the brain and spinal cord to understand affective behavior, nociception, and emesis.
lasting sensitization of pain pathways, and results in a predisposition to chronic pain (Taylor and Corder, 2014). These models offer the potential of identifying neural circuits that contribute to chronic pain, said Price. For example, work in his lab suggests that dysfunction of the dopaminergic system may play a critical role in pain chronicity (Megat et al., 2017). He added that stress primes animals for the acute-to-chronic pain transition, and is being actively investigated in conjunction with these other models. Moreover, Price stated that priming models are useful to predict efficacy of various analgesics for the treatment of chronic pain, as well as to predict whether an acute pain treatment can prevent priming and the transition to chronic pain.
He added that a better understanding of the mechanisms underlying the acute-to-chronic pain transition may suggest therapeutic approaches to reverse the transition and treat chronic pain. Price cited three hypothesized chronic pain resolution mechanisms that are currently under investigation in preclinical models: the resolvins, immune modulators, and adenosine monophosphate-activate protein kinase (AMPK) hypotheses. The resolvin hypothesis suggests that the resolvins, a unique family of lipid mediators, may reduce inflammatory pain both centrally and peripherally (Xu et al., 2010). The immune modulator hypothesis suggests that CD8+ T cells and increased interleukin-10 may promote recovery from persistent neuropathy following cancer treatment (Krukowski et al., 2016). Price mentioned two potential interventions related to this mechanism that might prevent the development of or even reverse chronic pain: an IL4-10 fusion protein delivered intrathecally (Eijkelkamp et al., 2016) and exercise-induced release of IL-10 (Grace et al., 2016). Price’s lab is pursuing a third hypothesis, the AMPK activation hypothesis, which suggests that AMPK activators may also prevent development of or reverse chronic pain (Asiedu et al., 2016).
Preclinical models also have revealed sex differences in the acute-to-chronic pain transition, said Price. For example, CGRP antagonists have profound effects in female rodents, but do nothing in males. Given that migraine is predominantly a female disease, CGRP antagonists may be extraordinarily effective, he said.
Clifford Woolf suggested it may be possible to use induced pluripotent stem cell (iPSC) lines from well-phenotyped and -genotyped patients to explore the acute-to-chronic pain conversion. While some people use the term “transition,” he said the word “conversion” more accurately reflects that change from one set of neurobiological processes that cause acute pain to another set of processes that cause chronic pain. Comparing iPSC lines
derived from patients who developed chronic pain from ones who did not could help identify the underlying pathways or targets involved, he said. Moreover, once candidate drugs have been identified, iPSCs from many different patients could be used for multiple in vitro trials, with no concerns about placebo effects.
Tony Yaksh said there are also inflammatory models that produce more persistent and long-lasting pain responses (Christensen et al., 2016), and may reflect the conversion from acute to chronic pain.
Clinical Efforts to Prevent the Acute-to-Chronic Pain Transition
There are also many efforts to develop new pain treatments, but there has been less attention devoted to developing interventions that could prevent the acute-to-chronic pain transition, said Robert Dworkin, professor of anesthesiology and perioperative medicine, neurology, and psychiatry at the University of Rochester. Such studies could test various models of prevention, including the following conditions in which the acute-to-chronic pain transition often occurs:
- Surgery, leading to chronic postsurgical pain;
- Acute low back injury leading to chronic low back pain;
- Wrist fracture, leading to CRPS;
- Cancer chemotherapy, leading to peripheral neuropathy;
- Herpes zoster and the development of postherpetic neuralgia; and
- Diabetic peripheral neuropathy.
The burden of chronic pain associated with these conditions is high, said Dworkin. For example, postsurgical pain persists in between 10 and 50 percent of cases, varying with the type of surgery (Kehlet et al., 2006), although Dworkin said it is unclear whether severe acute pain itself is a causal risk factor or concomitant to nerve or musculoskeletal damage. Mackey added that surgery is nothing more than a controlled injury and thus, the perioperative period provides a unique environment in which to study what happens before and after injury.
Dworkin suggested two broad types of studies that need to be done, preferably in tandem: (1) observational studies to develop a better understanding of risk and protective factors, transition mechanisms, and biomarkers of transition; and (2) clinical trials to test putative preventive interventions. Some of the risk factors are well established, he said. For
example, the severity of acute pain and the severity of nerve or musculoskeletal injury increase the likelihood that chronic pain will develop and persist. A prior history of chronic pain and psychosocial vulnerabilities such as catastrophizing—that is, the irrational belief that something is worse than it actually is—also raise the risk of developing chronic pain, he said. Dworkin added that some people may also be at increased risk because of underlying mechanisms that could facilitate the transition from acute to chronic pain, such as impaired conditioned pain modulation or augmented central sensitization.
Fifteen years ago, Dworkin and Dennis Turk, from the University of Washington, launched the Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT),3 which has since convened 20 meetings to build consensus on how pain trials should be conducted. Dworkin said he thinks trials of putative preventive interventions can and should be started right away using recommendations published in 2015 (Gewandter et al., 2015). For example, patients scheduled for cancer chemotherapy or surgery could receive preventive interventions a week or two before the cancer treatment or surgery. He added that multiple followups should be done. Possible primary endpoints could include incidence of and time to resolution of any chronic pain or clinically important chronic pain, pain intensity after 6 months, and various area under the curve analyses, said Dworkin. In addition, he advocated for including hypothesized transition biomarkers in all studies.
Examples of prevention trials that Dworkin said could be soon include those that would deliver anti-NGF antibodies perioperatively in patients scheduled for knee replacement or thoracotomy, and brief targeted catastrophizing interventions such as cognitive behavioral therapy in patients with acute low back injury. He also mentioned that while three clinical trials of vitamin C for prevention of CRPS have produced conflicting results, this low-risk intervention should be tested further. Ultimately, he said that preventing the acute-to-chronic pain transition will likely require multimodal intervention, including pharmacological and non-pharmacological treatments, such as physical therapy and psychological treatment.
Mackey proposed that clinical studies also explore the continuum of pain and trajectories of pain using duration as a dependent variable to determine which factors cause pain to be persistent, what leads patients to persistently use opioids, and which characteristics lead to resilience. He noted that surgery is a risk factor for persistent opioid misuse and abuse.
Thus, it provides an opportunity to better understand which factors contribute to persistent opioid use and design interventions to mitigate this problem. Volkow added that patients who are treated with opioids for chronic pain tend to have higher levels of pain, suggesting that opioids themselves may facilitate the conversion to chronic pain, possibly through common mechanisms of neuroplasticity and conditioning. Past and current history of smoking increases the risk for developing chronic pain conditions like temporomandibular disorders, said Maixner. Price said the preclinical data also support this idea—giving a μ-opioid agonist at the time of injury seems to promote neuronal plasticity, while inverse agonists4 precipitate a more transient pain state (e.g., Kandasamy and Price, 2015).
Volkow noted two other factors known to increase the risk of converting to chronic pain: being female and catastrophizing. Catastrophizing invokes multiple circuits, she said, and she wondered if differences have been identified between XY and XX cells. Catastrophizing is a heritable trait, said Diatchenko, but to study this at the genetic level first would require identification of all the elements that compose catastrophizing. Woolf added that while this could be studied in vitro, it would require many fully genotyped cell lines from a diverse set of individuals to get a sense of line-to-line variation. Robert Gereau agreed that catastrophizing is an important predictor of poor pain outcome for patients undergoing surgery. He suggested that mapping genetic susceptibility in terms of cognitive flexibility could have a substantial impact on understanding catastrophizing, and that it could have translational implications in terms of intervening to prevent long-term pain.
According to Diatchenko, some pharmaceutical companies fear that technologies identifying subgroups of patients who will respond to particular treatment will reduce the generalizability of a drug and be reflected in the labeling of the product. Hertz commented that restricted labeling would be preferable to failure. Jonathan Jarow, senior medical advisor at the FDA, noted that using predictive biomarkers in a clinical trial can lead to labeling that requires a companion diagnostic, which can create some
4 Agents that bind to the same receptor as agonists, but induce an opposite response.
problems. Moreover, he noted that predictive biomarkers can be very complex. For example, expression of programmed death-ligand 1 (PD-L1) can predict the response of some tumors to some drugs (Patel and Kurzrock, 2015) and thus, the labeling of these drugs may indicate that they should be used only for certain types of tumors that overexpress PD-L1, said Jarow. But for other tumor types, many responders do not express the biomarker, and therefore, the biomarker is not required in the labeling for those indications, he added.
Prognostic biomarkers used for trial enrichment can include more general phenotypic characteristics such as weight or physical signs and usually do not require a companion test, said Jarow. However, there is still a possibility that this could affect labeling. He added that a surrogate endpoint is “very problematic,” and will require validation that it is reasonably likely to predict clinical benefit. The FDA generally restricts the use of surrogate endpoints to serious conditions for which the event rate (e.g., death) is very low or when it takes a long time to see the event, he said. For example, in the early development of HIV drugs, viral titer at 12 weeks was accepted as a useful endpoint to get the drug to market more rapidly, said Jarow.
This page intentionally left blank.