OR WAIT null SECS
Only eight percent of CNS drugs will reach approval, in part because of the high adverse effects standards.
Central nervous system (CNS) indications, including neurologic and psychiatric drugs, remain an area of great potential due to the still-high unmet need for novel and improved medicines to treat these complex conditions. For example, due to aging populations, the incidence of neurodegenerative diseases, such as Alzheimer's disease and Parkinson's disease, is set to increase over the coming decades. Meanwhile, attrition rates for CNS drugs appear to be higher than for indications in other therapeutic areas,1 further challenging the prospects for pharmaceutical companies in bringing CNS drugs to the market. Reportedly, only 8.2% of CNS drugs that begin clinical testing will be approved, and those that are cleared take an average of 18 years to advance from the laboratory bench to the market.2
CNS drugs fail in development for various reasons. Some of the special challenges include the need to cross the blood-brain barrier; patient heterogeneity; complex diseases that often involve multiple molecular targets; the relatively low predictive validity of experimental animal efficacy models; and the relative lack of established clinical biomarkers and proof-of-concept models for these diseases.3 Another major issue is the regulatory need to establish a favorable risk-benefit profile for these drugs. While there is a relatively high tolerance for adverse effects with oncologic or cardiovascular drugs, the bar is set much higher for many CNS indications, possibly because of widespread misperceptions about the burden of CNS disease. In addition, these drugs may be associated with special safety risks unique to centrally-acting compounds that can alter the risk-benefit profile of these drugs unfavorably, even as the unmet need for these treatments grows. These include issues such as abuse liability and physical dependence, cognitive and motor impairment, suicidality, and interactions with other CNS drugs and alcohol.
Despite these challenges, it is worthwhile for drug developers to try to minimize CNS-related safety risks through better early screening and candidate selection, because the market potential for CNS drugs remains high. The CNS drug market is forecast to grow to $81.8 billion by 2015.4 In addition, CNS continues to be the second-most researched therapeutic class behind cancer, with more than 300 drugs in clinical development targeting mental illnesses, ranging from Alzheimer's disease to depression and from schizophrenia to addictive disorders.5 Alzheimer's is the most actively researched of the group, due to the disease's significant public health burden. In the United States, Alzheimer's costs the healthcare system $172 billion a year.6
Detecting relevant CNS safety signals early on can help sponsors make "go/no-go" decisions and more effectively design their late-stage development program around the understanding of those liabilities. Putting this approach into practical use is crucial amid the growing demands on drugmakers to conduct faster and more efficient early-phase studies. While assessments in early-stage development may not directly impact critical go/no-go decisions on an investigative product, resulting insights could help accelerate clinical timelines, reduce costs and ultimately represent a potential commercial benefit by helping convince investors to continue funding development into Phase II trials.
In addition, with increased regulatory expectations around the risk-benefit profiles of new CNS drugs, early and continuous monitoring of safety indicators during drug development can help avoid later delays in regulatory review and market approval.
The hunt for CNS safety indicators, however, can pose unique challenges. Due to the often subjective nature of these assessments, along with additional factors such as specialized and small subject populations in Phase I trials, the responsibility of sponsors and contract research organizations (CROs) to evolve strategies in signal detection is paramount.
Amid the specter of soaring prescription drug abuse in the United States, the nonmedical use of CNS treatments in particular has become a serious public health concern for healthcare providers, drug manufacturers, and governmental regulators. In recent years, the number of cases of prescription drug abuse in the United States has grown significantly. It is estimated that close to 20% of people over 12 years old in the United States have used prescription drugs for non-medical purposes.7 In 2010, nearly 3 million teenagers and young adults became new abusers of prescription drugs.8
In efforts to stem this tide, abuse liability assessments are required for all CNS-active products, including drugs that may show signals of depressant, stimulant, hallucinogenic, or mood enhancing effects.9 These assessments are typically conducted in later-stage development, but given the inherent risks associated with CNS drug development, the need to generate more indicators of abuse potential in the early phase is crucial.
To that end, biopharmaceutical companies and CROs are increasingly implementing abuse potential detection strategies in Phase I clinical trials in efforts to more broadly evaluate the liability profile of their CNS-active drug candidates. The key, therefore, is using these signals to aid in predicting eventual abuse liability data requirements. Strategies that can be incorporated into early Phase I studies to detect abuse potential signals can better position sponsors to anticipate later-stage requirements for abuse liability assessment.
Abuse liability assessments evaluate the likelihood that a drug may be abused and/or result in dependence once on the market. Coupled with preclinical and other clinical data, these assessments provide regulatory agencies with a scientific basis on which to make product labeling and drug scheduling recommendations in hopes of reducing the effects of abuse and addiction. In many cases, this process involves performing human abuse potential studies, which evaluate the ability of a new drug to produce positive psychoactive effects such as sedation, euphoria, and mood changes. According to guidance released by the US Food and Drug Administration (FDA) in early 2010, human abuse potential studies are usually conducted in experienced, recreational drug users who have a recent or current history of using a drug in the pharmaceutical class of the test product.10 Regardless of the pharmacologic mechanism of action, the subjects in the study should have experience with drugs that have similar psychoactive properties. The guidance noted that recent abuse potential studies have also been conducted in drug-naive healthy subjects, an area of needed research, according to the FDA. The agency says these two populations may differ in significant ways, including in their ability to identify subtle differences in drug effects that are relevant to abuse assessment.
One such vital area of research involves the collection of early-stage clinical indicators of abuse potential that are manifested in Phase I studies in healthy volunteers. Such signals can focus on abuse-related adverse events (AEs) suggesting mood elevation, euphoria, or hallucinogenic effects.11 Other indicators can point to the subjective effects and pharmacokinetics of a drug, its early pharmacodynamics, and its proof of mechanism, providing evidence of similarity to existing controlled substances.
Early-stage indicators of abuse potential can be gleaned from Phase I single-ascending-dose (SAD) and multiple-ascending-dose (MAD) studies. This can be accomplished by adding additional measures to SAD and MAD trials. It is important to recognize that these and other abuse potential measures are subjective, not confirmatory. There are no existing objective measures that accurately predict potential for abuse across drug classes. Nevertheless, the ability to categorize, tabulate, and analyze subjective effects can prove valuable in generating useful information about the target.
In Phase I studies with healthy volunteers, the primary method used for evaluating the subjective effects of drugs is through the use of subjective questionnaires. Primary endpoints typically used in human abuse potential studies, such as measures of "drug liking," "good effects," "feeling high," etc. and the Addiction Research Center Inventory are probably not appropriate to use in healthy non-drug using populations. These measures have not been validated for use in these populations, and interpretation of data can be problematic. In particular, with a few exceptions, non-drug using volunteers generally do not report these effects with drugs of abuse, leading to a very high potential for false negative results. Even if a positive "signal" is detected, it cannot necessarily be extrapolated to drug abusing populations as it is unknown how a non-drug user may interpret a question about "liking" a drug or feeling "high," since they don't have experience abusing drugs.
In the absence of more research to develop healthy volunteer-specific scales, a more useful approach is to include measures of general pharmacologic/subjective effects that may be of interest to abusers (i.e., the stimulant, depressant, hallucinogenic, and mood elevating effects). Examples of potentially useful measures include general mood scales, such as the Bond-Lader visual analog scale,12 the profile of mood states (POMS),13 or even some of the more general assessments often included in human abuse potential studies, such as visual analog scales of alertness, drowsiness, etc. Those measures can be integrated seamlessly into a SAD or MAD study and can provide invaluable information about the potential profile of the drug.
Tracking AE data in early-stage development can also give sponsors an idea of the different subjective effects of their drug. For example, if the product has a stimulant effect, clues may arise regarding the AE pattern of the drug by tracking incidents such as hyperexcitability, euphoric mood, and restlessness. AE data have traditionally been a critical component in overall safety evaluation, as well as a useful measure in setting dose level limits. In strictly providing signals of abuse potential, AE data can be prospectively monitored based on predefined low-level terms of interest. It is important that analyses include summaries and listings of coded and subject verbatim events, as well as case narratives on AEs of particular interest, such as "euphoric mood," "feeling drunk," or different types of hallucinations. Narratives can trigger a deeper understanding of the context of AE reports and whether they suggest signals of abuse potential.14
Reliable reporting and critical evaluation of AE and pharmacodynamic data can also aid in early decision making. After Phase I testing, for example, a drug that shows "signals" of abuse potential can be targeted for further abuse liability testing earlier in development than is typical, or targeted for indications where abuse liability may be less of an issue (i.e., either indications where competitors are already controlled substances or indications which may be lower risk for abuse, such as Alzheimer's disease) or where the risk-benefit profile for these types of effects is more favorable. In addition, if multiple candidates are available for a particular class/indication, this type of data can help sponsors decide which candidate would be most appropriate for advancement. The resulting indicators can also help sponsors understand the pharmacology of the product, so that if they are required to conduct an abuse potential study later in the development, they have a better idea of what comparators, measures, and populations they will need to evaluate.
Due to limited patient populations in Phase I trials, collecting reliable safety signals of abuse potential can be challenging and limitations should be acknowledged. Analysis sets in these studies contain much smaller samples, with subject characteristics, doses, and treatment durations vastly different from one another. Historically, sources of safety data, most predominately AEs reported during Phase I to Phase III trials, have not been a major focus of study by the addiction research community, due in part to the potential of error from AE reports in assessing the subjective effects of drugs.15 Limitations also arise from the absence of an established direct link between AEs in clinical trials and actual abuse. However, these limitations are less of an issue when the data are used as "signal" generation, rather than viewed as confirmatory evidence.
In addition, strategies such as integrated analysis could boost the usefulness of early signals on predicting abuse potential. Integrated analysis enables the opportunity to look at safety data (AE and pharmacodynamic/subjective) across multiple studies in a drug development program, allowing the introduction of tailored displays specific to abuse potential assessment.16 For instance, structured assessments of mood or behavior are more likely to occur in short-term Phase I studies, making the Phase I analysis set perhaps the logical integrated dataset on which to perform such analyses.
Another abuse-related liability where early-stage investigation is important is physical dependence. This refers to a state resulting from habitual use of a drug that has produced tolerance and where negative physical symptoms of withdrawal result from abrupt discontinuation or dosage reduction.17 Physical dependence can develop from low-dose therapeutic use of benzodiazepines, opioids, antiepileptics, and antidepressants, as well as misuse of recreational drugs such as alcohol, benzodiazepines, and opioids. Physical dependence may not necessarily be associated with abuse, as several classes of drugs are known to induce withdrawal syndromes but are not abused—e.g., selective serotonin reuptake inhibitors (SSRIs). Regardless, physical dependence is an important potential safety issue that may affect how a CNS drug can be used.
The European Medicines Agency (EMA) has historically focused on concerns of physical dependence, although its guidance thus far in this area refers mainly to animal/non-clinical studies. The issue has also gained increased attention from US regulators. The FDA requires that an assessment of tolerance or physical dependence be performed as part of the safety evaluation of a drug candidate and the data is considered in drug scheduling. For new drugs, the clinical assessment is usually largely based on long-term Phase III studies, as clinical pharmacology and human abuse potential studies often involve only single or short-term dosing of the test drug.
Nevertheless, careful prospective evaluations in early clinical development can help evaluate a drug candidate's potential to trigger physical dependence among users. Due to ethical concerns, physical dependence is not typically intentionally induced in human subjects, although a small number of studies have examined this issue in healthy volunteers.18, 19 Therefore, the clinical evaluation of physical dependence in healthy volunteers and patients is typically based on AEs observed upon discontinuation or taper of the drug. In early phase, sponsors can look at AE data after completion of the MAD study and discontinuation of the drug. The emergence of new or more severe AEs after complete drug discontinuation can signal the potential for physical dependence. Care must be taken, however, to ensure that the AEs observed after withdrawal are not simply reflective of residual on-treatment AEs (i.e., a true signal would show a higher incidence after discontinuation and/or a change in patterns of AEs). As with "abuse-related" AEs, interpretation of withdrawal AEs should be undertaken with caution and not considered as confirmatory evidence. Nonetheless, these types of data can help guide sponsors in determining the extent of evaluation needed in later phases.
In some cases, various quantitative scales and measurements can generate objective data to assess dependence. Examples of existing withdrawal scales include those for SSRIs,20 benzodiazepines (e.g., physician withdrawal checklist, Ashton withdrawal scale),21, 22 and opioids (e.g., Himmelsbach's withdrawal rating scale and short opiate withdrawal scale)23, 24, 25. Although there are some limitations to including this type of scale in clinical trials with new chemical entities, some of them may be useful in that they examine a fairly broad range of potential symptoms (e.g., scales for benzodiazepines and SSRIs). However, the results should be interpreted cautiously as they may indicate a statistical effect that is not necessarily clinically relevant (for example, if used in a fairly large Phase IIa study). Analyzing indicators of potential physical dependence more proactively will generate stronger data and, in turn, can ultimately be more convincing to regulatory agencies when evaluating the safety of the drug.
Besides assessing potential for abuse and dependence, strategies in early-phase development can also help detect signals for other CNS-related liabilities. Among these liabilities is cognitive or motor impairment. For example, cognitive impairment is a major factor limiting the use of benzodiazepines, as well as some sleep medications. It also increases the liability associated with a drug due to risk of accidents and falls. Similar to the approach for abuse potential, neurocognitive measures can be added to SAD, MAD, and other early-phase studies to detect risk for cognitive and performance impairment. These scales can be pulled from a battery of measures. Examples include tests of divided or selective attention, verbal learning, short-term memory, psychomotor control, and visual motor processing. Although there are some limitations to including these assessments in a SAD or MAD design (i.e., very small sample sizes and generally lack within-subject comparisons), if carefully evaluated, these measures can be used to explore whether a drug is associated with enhancement or impairment, particularly if changes from baseline (pre-dose) values are tracked and analyzed carefully to evaluate trends over time.
Cognitive impairment can also be evaluated in separate crossover studies. The EMA requires such studies for some CNS indications, including epilepsy and sleep disorders, and these may be considered for other indications such as schizophrenia, anxiety, depression, bipolar, smoking cessation, and attention deficit hyperactivity disorder (ADHD).26 The psychomotor/cognitive liabilities associated with sleep medications in particular were highlighted at a recent FDA public meeting.27 In addition, the FDA, in its guidance on abuse potential, highlights cognitive impairment as another key aspect of human pharmacology that should be investigated. The agency says psychomotor tests conducted during clinical development that determine whether the effects of a drug increase or decrease normal motor functioning can suggest whether the product may be like a known stimulant or depressant. Similarly, cognitive tests that assess whether memory, perception, attention, language ability, or consciousness are altered by a drug can indicate the presence of certain effects that drug abusers might find desirable. For example, a drug that enhances cognitive performance may be misused as a "study drug," as is the case with some medications for ADHD. While the traditional SAD/MAD design is not ideal for this type of controlled study (which would typically be a randomized, double-blind crossover), this type of study could be conducted soon after SAD and MAD data are available, or could even be incorporated as a separate part of a "super-protocol," to be conducted immediately following the MAD part of the study.
Assessing the potential interaction of CNS drugs in development with alcohol or other CNS drugs is also important to evaluate early on. Reported estimates suggest that alcohol-medication interactions may be a factor in at least 25% of all emergency room admissions. An unknown number of less serious interactions may go unrecognized or unrecorded. In addition to alcohol, concomitant administration of more than one CNS medication may also lead to additive cognitive/motor impairment or CNS depression.
Because evaluation of this type of effect requires an additional intervention, it cannot be assessed as easily within a traditional SAD or MAD design. However, early-phase strategies in this area could include adding separate cohort(s) at the end of the SAD or MAD study, which combine the investigational product with a single dose of alcohol or other CNS drug (typically CNS depressants such as benzodiazepines). In a MAD design, alcohol or other CNS drugs could also reasonably be added after data have been collected for the drug alone at steady state, although the fixed order (rather than traditional randomized crossover design) may limit interpretation of the data (i.e., such a design would be exploratory rather than pivotal). In either case, the use of standard batteries of neurocognitive, motor, and subjective measures can be used to assess the potential for additive CNS depression/impairment. With early-stage programs increasingly featuring multiple-part studies examining such factors as food effects or elderly cohorts, the introduction of an alcohol/drug interaction arm could provide useful information as well. This type of design may be particularly useful when it is known that the drug in development will likely be co-administered with CNS depressants, such as benzodiazepines, or with alcohol (i.e., medications in development for treatment of alcohol dependence).
Much attention has been paid in recent years to possible links between antidepressant treatment and an increased risk of suicidal behavior in pediatric and adult patients. Concerns around antiepileptic drugs (AEDs) have been spotlighted as well. In late 2008, the FDA issued an alert about an increased risk for suicidality during treatment with AEDs, including those used to treat psychiatric disorders, migraine headaches, and other conditions, as well as epilepsy. The agency decided to add a label warning on heightened suicide risk for users of AEDs. The FDA made the decision based on its review of 199 clinical trials of 11 AEDs. These studies showed that patients receiving AEDs had almost twice the risk of suicidal behavior or thoughts compared to patients receiving a placebo.28
In guidance released in 2010, the FDA urges that all clinical protocols for drugs developed in the agency's Division of Psychiatry Products include a prospective assessment for suicidality. This approach can ensure that subjects who are experiencing suicidality are properly recognized and adequately treated. It can also enable the collection of more timely (i.e., closer to the event) and more complete data on suicidality than have been collected in the past. This information can aid future efforts to better detect increased suicidality in individual studies and in pooled analyses. The FDA says prospective suicidality assessments should be conducted in all clinical trials involving a CNS-active drug, both inpatient and outpatient, including Phase I trials in healthy volunteers. The agency notes that treatment-emergent suicidality has been reported in short-term Phase I studies with several different antidepressants. Suicidality assessments should, therefore, be included even in SAD trials.29
The FDA recommends the use of the Columbia Suicide Severity Rating Scale (C-SSRS), which involves a series of detailed questions designed to measure possible suicidal ideation and behavior. This interview should be conducted at baseline and at each subject visit. The baseline or baseline screening versions of C-SSRS are suitable as part of a subject's first interview in a private practice or as part of a clinical study. The version for already enrolled subjects ("since last visit") assesses suicidality since the individual has begun taking part in a clinical trial and over their lifetime. This version is useful in determining any effects that a specific treatment method may have had on suicidality. In an early phase setting, where subjects are confined and carefully monitored, the C-SSRS is typically investigator-administered.30
CNS disorders will continue to present one of the toughest challenges to drug developers. Soaring clinical trial costs for conditions such as Alzheimer's disease, where patients often need to be followed for long periods of time, combined with frustratingly high attrition rates have forced many to rethink their investments in CNS research. Nevertheless, with a significant unmet need remaining and the economic burden of mental illness rising globally, opportunities for new and transformative therapies for these diseases are many. Instrumental in these efforts will be fresh approaches in early-phase development, including the use of innovative methods to detect unique CNS safety signals in Phase I clinical testing. With sponsors under increased pressure to accelerate trial timelines and regulators raising the risk-benefit standards for new CNS drugs, the ability to minimize potential liabilities through better early screening and monitoring is critical.
Kerri Schoedel,* PhD, is Scientific Director, Clinical Pharmacology, e-mail: [email protected], and Pierre Geoffroy, PhD, is Vice President, Early Phase both at INC Research, 3201 Beechleaf Court, Suite 600, Raleigh, NC.
*To whom all correspondence should be addressed.
1. I. Kola and J. Landis, "Can the Pharmaceutical Industry Reduce Attrition Rates?" Nat Rev Drug Discovery, 3 (8) 711-715 (2004).
2. K. I. Kaitin and C. P. Milne, "A Dearth of New Meds," Scientific American, (2011), http://www.scientificamerican.com/article.cfm?id=a-dearth-of-new-meds.
3. A. M. Palmer and F. A. Stephenson, "CNS Drug Discovery: Challenges and Solutions," Drug News Perspect, 18 (1) 51-57 (2005).
4. Therapeutic Drugs for Central Nervous Sys)tem Disorders: Technologies and Global Markets, BCC Research, (2010), http://www.bccresearch.com/report/drugs-central-nervous-system-disorders-phm068a.html.
5. "Medicines in Development for Mental Illnesses," Pharmaceutical Research and Manufacturers of America,
6. "Medicines in Development for Alzheimer's Disease and Other Dementias," Pharmaceutical Research and Manufacturers of America,
7. "Prescription Drug Abuse Statistics," http://www.prescriptiondrugabuse.us/statistics-facts.html.
8. "Prescription Drugs," Substance Abuse,http://www.substanceabuse.me/drug-information/prescription-drugs/.
9. Food and Drug Administration, Controlled Substances Act, (FDA, Rockville, MD, 1970), http://www.fda.gov/RegulatoryInformation/Legislation/ucm148726.htm.
10. Food and Drug Administration, Guidance for Industry, "Assessment of Abuse Potential of Drugs," (FDA, Rockville, MD, 2010), http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM198650.pdf.
11. K. A, Schoedel and E. M. Sellers, "Assessing Abuse Liability During Drug Development: Changing Standards and Expectations," Clinical Pharmacology & Therapeutics, 83 (4) 622-626 (2008).
12. A. Bond and M. Lader, "The Use of Analogue Scales in Rating Subjective Feelings," British Journal of Medical Psychology, 47 (3) 211-218 (1974).
13. M. Lorr, D. M. McNair, and L. F. Droppleman, "Profile of Mood States (POMSTM)," (1971), http://www.mhs.com/product.aspx?gr=cli&prod=poms&id=overview/.
14. R. S. Mansbach, "The Role of Adverse Events and Related Safety Data in the Pre-Market Evaluation of Drug Abuse Potential," Drug and Alcohol Dependence, 112 (3) 173-177 (2010).
15. K. T. Brady, "Assessing Abuse Liability in Clinical Trials," Drug and Alcohol Dependence, 70 (3 Suppl.) S87-S95 (2003).
16. R. S. Mansbach, "The Role of Adverse Events and Related Safety Data in the Pre-Market Evaluation of Drug Abuse Potential," Drug and Alcohol Dependence, 112 (3) 173-177 (2010).
17. PubMed Health, "Drug Dependence," (2010), http://www.ncbi.nlm.nih.gov/pubmedhealth/PMH0002490/.
18. P. Dorian, E. M. Sellers, H. Kaplan, and C. Hamilton, "Evaluation of Zopiclone Physical Dependence Liability in Normal Volunteers," Pharmacology, 27 (Suppl 2) 228-234 (1983).
19. U. E. Busto, C. A. Naranjo, K. E. Bremner, J. E. Peachey, and M. Bologa, "Safety of Ipsapirone Treatment Compared with Lorazepam: Discontinuation Effects," J Psychiatry Neurosci, 23 (1) 35-44 (1998).
20. K. Black, C. Shea, S. Dursun, and S. Kutcher, "Selective Serotonin Reuptake Inhibitor Discontinuation Syndrome: Proposed Diagnostic Criteria," J Psychiatry Neurosci, 25 (3) 255-261 (2000).
21. H. Ashton, "Benzodiazepine Withdrawal: An Unfinished Story," Br Med J (Clin Res Ed) 288 (6424) 1135-1140 (1984).
22. K. Rickels, E. Schweizer, W. G. Case, and D. J. Greenblatt, "Long-Term Therapeutic Use of Benzodiazepines. I. Effects of Abrupt Discontinuation," Arch Gen Psychiatry, 47 (10) 899-907 (1990).
23. C. K. Himmelsbach, "The Morphine Abstinence Syndrome, its Nature and Treatment," Ann Intern Med, 15 (5) 829-839 (1941).
24. M. Gossop, "The Development of a Short Opiate Withdrawal Scale (SOWS)," Addict Behav, 15 (5) 487-490 (1990).
25. L. Kolb and C. K. Himmelsbach, "Clinical Studies of Drug Addiction III. A critical Review of the Withdrawal Treatments with Method of Evaluating Abstinence Syndromes," Am J Psychiatry, 94, 759-799 (1938).
26. European Medicines Agency, "Guideline on Medicinal Products for the Treatment of Insomnia," (2011), http://www.ema.europa.eu/docs/en_GB/document_library/Scientific_guideline/2011/02/WC500102351.pdf.
27. Food and Drug Administration, "Safety and Efficacy of Hypnotic Drugs," Public Meeting, May 10-11, 2011, http://www.setonresourcecenter.com/register/2011/apr/07/2011-8285.pdf.
28. Food and Drug Administration, "FDA Requires Warnings about Risk of Suicidal Thoughts and Behavior for Antiepileptic Medications," (2008), http://www.fda.gov/NewsEvents/Newsroom/PressAnnouncements/2008/ucm116991.htm.
29. Food and Drug Administration, Guidance for Industry "Suicidal Ideation and Behavior: Prospective Assessment of Occurrence in Clinical Trials," (2012), http://www.fda.gov/downloads/Drugs/GuidanceComplianceRegulatoryInformation/Guidances/UCM225130.pdf.
30. Columbia University Medical Center, "Columbia Suicide Severity Ratings Scale," http://cssrs.columbia.edu/clinical_trials.html.