31

The United States FDA’s Three Risk Guidances of 2005

In March 2005, the U.S. Food and Drug Administration (FDA) issued three guidances, which were years in the making, summarizing the agency’s views on risk management. These documents are critical to conveying how the FDA (and other health agencies) views risk. The guidances can be found at the links below and are briefly summarized in this chapter.

1. Premarketing Risk Assessment (Web Resource 31-1)
   
2. Development and Use of Risk Minimization Action Plans (RiskMAPs) (Web Resource 31-2)
   
3. Good Pharmacovigilance Practices and Pharmacoepidemiologic Assessment (Web Resource 31-3)

The FDA begins by noting that routine risk assessment is already being done. It notes that this guidance is not aimed at use on all products but rather only on those that “pose a clinically important and unusual type or level of risk.”

The adequacy of assessment of risk depends on quantity (number of patients studied) and the data quality (“the appropriateness of the assessments performed, the appropriateness and breadth of the patient populations studied, and how results are analyzed”).

“Providing detailed guidance on what constitutes an adequate safety database for all products is impossible.” Each product is weighed on its own merits.

Size of the Safety Database

The number of patients studied depends on the novelty of the drug, the availability and safety of alternate therapies, the intended population, the condition being treated, and the duration of use.

The FDA does not give safety database size advice for products for the short term (< 6 months use) but does make suggestions for products aimed at treatment longer than 6 months. For these, the FDA and the International Conference on Harmonization (ICH) recommend 1500 subjects, with 300 to 600 exposed for 6 months and 100 exposed for 1 year using doses in the therapeutic range. Higher numbers of patients may need to be studied if:

• Specific safety issues arise from animal studies.
  
• Similar drugs or the class of drugs suggest a specific problem.
  
• Pharmacokinetic or pharmacodynamic properties of the drug are associated with certain adverse events (AEs).
  
• It is important to quantitate the occurrence of low-frequency AEs.
  
• The benefit from the product is small, and one wants to be sure there are no rare AEs that will not be picked up unless large numbers of patients are studied.
  
• The benefit is experienced by only a fraction of the treated patients (i.e., the benefit is “rare” and thus the same reason as above to look for rare AEs).
  
• The benefit is unclear (e.g., surrogate endpoints used).
  
• Statistical power requires larger a number of patients to show that an already high background rate of safety issues will not be unduly raised even more because of the drug.
  
• The proposed treatment is for healthy populations (e.g., preventive vaccines).
  
• A safe and effective alternate treatment is available.

In practice, this has been done for quite some time. The ICH/FDA figures above are somewhat arbitrary and may need to be higher, depending on the drug and population being studied. Often, statistical needs for power or significance for the efficacy endpoints drive sample size and are usually insufficient for statistical safety conclusions. What is noteworthy is the requirement for longer-term exposure (6 months and 1 year) for drugs that are used for 6 months. Even these suggestions may be conservative, and some drugs have had much longer follow-ups (e.g., Adriamycin, diethylstilbestrol), where chronic toxicity is expected or suspected to occur.

Long-Term Controlled Safety Trials

The FDA notes that, in many clinical programs, uncontrolled, single-arm, long-term safety studies are done. Although useful, the use of a control or placebo (if possible) is preferable, especially when the AEs in question are relatively common in the treated population (e.g., sudden death in patients with ischemic heart disease), or when the AE might mimic the disease being treated, such as asthma exacerbations due to inhalation treatments given for asthma (see fialuridine and hepatitis, Chapter 52). Long-term safety studies are also useful if the toxicity worsens with cumulative exposure.

Uncontrolled studies are useful in picking up certain AEs that (essentially) never occur spontaneously. The FDA gives the examples of aplastic anemia and severe hepatocellular injury (though not stated, one presumes they mean noninfectious hepatic injury).

This section highlights an issue that has been around for many years but is changing: the use of placebo-controlled trials. Historically, the FDA felt that placebo-controlled trials are the gold standard for characterizing a drug’s safety and efficacy. However, such trials are becoming more difficult to perform as the ethics of placebo use have evolved. Most placebo studies are frowned upon. The World Medical Association’s Declaration of Helsinki states:

• Where for compelling and scientifically sound methodological reasons its use is necessary to determine the efficacy or safety of a prophylactic, diagnostic or therapeutic method; or
  
• Where a prophylactic, diagnostic or therapeutic method is being investigated for a minor condition and the patients who receive placebo will not be subject to any additional risk of serious or irreversible harm (Web Resource 31-4).

In addition, sponsors often avoid head-to-head trials for fear of unexpected bad results against a competitor or even against the sponsor’s own comparator (e.g., an older drug the sponsor wishes to discontinue). This too is changing as comparative effectiveness requirements are established around the world. The United Kingdom has done this for some years (see The UK National Institute for Health and Clinical Excellence at Web Resource 31-5). Clinical effectiveness and comparisons are expected to be established in the United States within the next several years. Hence, many studies now are done against the “standard of care” (SOC—do not confuse this abbreviation with the MedDRA SOC, meaning System Organ Class), which often means that these studies are not able to show superiority (one treatment is better than the other) but only noninferiority. Medical practitioners generally prefer to see trials against the current SOC, arguing that practitioners do not treat with placebo in the real world but rather with something already out on the market.

Diversity

The premarketing safety database (mainly in phase III studies) should represent, as much as possible, the expected target population. The FDA argues that, where possible, inclusion criteria should be fairly broad to include elderly patients (and, in particular, “the very old”) with concomitant diseases and patients taking concomitant medications.

Another fundamental area of potential conflict is between sponsors (companies) and the FDA. Sponsors generally like “very clean” studies in homogeneous patients so that efficacy (and safety) can be evaluated without extraneous “noise” (confounders) and that sample size can be kept “reasonable.” The FDA, knowing that the drug will be used in the general population, only some of which resembles the test population, would like to see a much broader selection of patients tested before marketing.

Exploring Dose Effects

Normally, if one is skilled and lucky, the phase II trials establish the clinical dose that is then studied in large numbers of patients in phase III studies. The FDA argues for the study, in phase III, of more than one dose. They indicate that many efficacy and safety data can be obtained from these trials. This may or may not be feasible.

Drug Interactions

It is not possible to study all potential drug or other interactions. Those studies that are done should focus on the following:

• Drug–drug interactions should be looked at using those drugs that might be expected to produce safety issues resulting from known metabolic pathways (e.g., certain cytochrome P450 enzymes). Logical choices of test drugs should be made (e.g., for a new cholesterol-lowering treatment, examining the consequences of concomitant use of HMG CoA reductase inhibitors).
  
• Drug–demography interactions should be sought where gender, age, race, or other demographics might play a role.
  
• Drug–disease interactions should be examined if necessary.
  
• Drug–dietary supplement interactions may need to be evaluated (e.g., interactions between an antidepressant and St. John’s wort).

The FDA recommends that pharmacokinetic assessments be built into some late-phase clinical trials to see whether unexpected interactions can be picked up.

Drug–drug interactions in particular remain a difficult nut to crack. Certain standard studies looking at the key cytochrome P450 enzyme pathways or drugs known to have many interactions (e.g., Coumadin, anticonvulsants) are straightforward and generally done routinely, as are drug–food and drug–alcohol studies. It is only after marketing that “surprise” interactions are seen (e.g., St. John’s wort and human immunodeficiency virus drugs). Sometimes pharmacokinetic interactions are seen but with no pharmacodynamic or clinical implications. It may not be feasible, safe, or ethical to treat normal subjects with two drugs to see whether the drugs interact. Similarly, it may not be feasible, safe, or ethical to treat a cohort of patients already on a drug (e.g., acquired immunodeficiency syndrome patients, asthmatics) with the new drug to see whether there is an interaction.

Comparative Safety Data

The FDA next comments on safety data and comparators in clinical trials. They note that “much of the safety data in an application may be derived from placebo-controlled trials and single-arm safety studies, with little or no comparative safety data.” Placebo-controlled trials may no longer be feasible in most cases. The FDA notes that active controls or a placebo arm if possible would be useful in the following circumstances:

• The background rate of AEs is high. Using a single-arm study might show an alarmingly high rate of AEs, which would be of less concern if a placebo arm showed a similarly high rate.
  
• There is a well-established treatment already. A single-arm trial would likely be uninformative and a placebo control unethical. Thus an active comparator is the usual choice.
  
• A superiority claim for safety or efficacy is desired. A control then is obviously needed.

Certain special circumstances require a tailored approach:

• Chronically used, long-half-life drugs or drugs with dose-related toxicities should have studies to determine whether a lower dose or less frequent dose is appropriate.
  
• If a specific titration schedule is needed, this schedule should be based on specific studies to determine the best titration scheme.
  
• If certain AEs are not likely to be detected or reported without special attention or tests, studies should include these. As an example, the FDA cites a new drug with central nervous system effects and notes that it should have assessments of cognitive function, motor skills, and mood.
  
• In pediatric studies, special attention should be paid to effects on growth and neurocognitive development if the drug is used in the very young, to excipients, and to immunization recommendations.
  
• The sponsor should evaluate whether and when the collection from some or all patients of blood or other bodily tissues or fluids during phase III studies for analysis at a later time should be done. This might be useful for analyzing unusual signals or for retrospective analyses at a later date.

Large Simple Safety Studies

These are usually randomized controlled studies done to assess limited specific safety outcomes in large numbers of patients. Rarely, they may be uncontrolled if the event being assessed is uncommon. The FDA notes these are usually done as formal phase IV commitments, although they could be done occasionally earlier in development. Another place for a large study is a product developed for preventive use in at-risk subjects who are otherwise healthy and where the benefits may be small.

Medication Errors

Although, historically, medication errors were not a part of drug safety or pharmacovigilance, they are now a high FDA priority in risk management. The FDA now wants sponsors to pay close attention before marketing to possible areas of medication error such as packaging, drug name, and labeling. Should medication errors or potential for errors be noted in the premarketing clinical trials, they should be acted on to remove potential for errors. The FDA recommends drawing from the experience of Good Manufacturing Practice and device development by

• Conducting a failure mode and effects analysis
  
• Using expert panels
  
• Using computer-assisted analysis
  
• Using direct observation during clinical trials
  
• Using directed interviews of consumers and medical and pharmacy personnel to better understand comprehension
  
• Using focus groups
  
• Using simulated prescription and over-the-counter use studies

Assessing Safety During Product Development

During the development of all new small-molecule drugs, the following should be addressed as part of the New Drug Application (NDA):

• Drug-related QTc prolongation (to exclude arrhythmic potential)
  
• Drug–drug interactions
  
• Polymorphic metabolism

These experiments are conducted with “normal” doses in humans and are now known as safety pharmacology, in contrast to animal toxicology, which consists of increasing dosages until some untoward effect occurs:

• Drug-related liver toxicity
  
• Drug-related nephrotoxicity
  
• Drug-related bone marrow toxicity

Not all drugs require studies looking at all of these issues. The FDA has issued various other guidances and requirements for premarketing testing. Refer to the FDA website for specific information.

Biologic products may require additional testing:

• Immunogenicity: “both the incidence and consequences of neutralizing antibody formation and the potential for adverse events related to binding antibody formation.”
  
• Transfection of nontarget cells and infection transmissibility to close contacts should be evaluated for gene-based products.
  
• “For cell-based products, assessment of AEs related to distribution, migration and growth beyond the initial intended administration are important as are AEs related to cell survival and demise.”

Data Analysis and Interpretation

The FDA refers the reader to three ICH guidances on preparing NDAs, clinical study reports, and the common technical document. The FDA notes that the later phases of product development are primarily aimed at efficacy, with endpoints identified in advance and with statistical calculations based on these endpoints. In contrast, the safety measures are usually not addressed with any prespecified level of statistical sensitivity. Thus, premarketing safety data are often only exploratory and useful primarily for signal development.

Describing AEs to Identify Signals (Coding): The FDA recommends use of a single coding system (e.g., MedDRA), with a standardized coding convention and without updates throughout the program.

Coding Accuracy: Care should be taken with investigators and coders such that the reported verbatim terms are accurately reflected in the MedDRA codes used for these reported AEs. Severity and magnitude should not be exaggerated (e.g., coding acute liver failure when all that occurred was isolated transaminase elevation without elevated bilirubin, coagulopathy, or encephalopathy) or masked (using a nonspecific and “unimportant” term to describe a serious AE). The sponsor may “recharacterize” an AE code when appropriate but with an audit trail and with FDA consultation.

Verbatim terms from investigators should be captured, verified to be accurate, and mapped to the appropriate MedDRA term (e.g., the verbatim term “suicidal ideation” should not be captured as “emotional lability”). The sponsor should ensure that verbatim terms are consistent across studies and individual coders. The FDA also suggests that the sponsor “consider a coded event in conjunction with other coded events in some circumstances” and may define certain entities as “an amalgamation of multiple coding terms.” They give as an example “acute liver failure,” which could be used if it is based on recognized definitions of this term. The FDA does not, however, come down squarely on one side or the other in the “lumpers versus splitters” debate (i.e., combine terms or report them separately). This is an area of continued controversy, with various coding conventions proposed.

Coding Analysis: The FDA cautions about how analyses can be altered or obscured by the way coding is done. They urge the sponsor to pay attention and use coding (whether lumping or splitting) to obtain the best signals:

• Splitting: Using multiple terms may at times be more useful than combinations. They give the example that “dyspnea, cough, wheezing and pleuritis” is more sensitive and useful than “pulmonary toxicity” but warn that “constipation” might include and hide cases of “toxic megacolon.”
  
• Splitting may, however, falsely decrease the incidence of AEs. “Fluid retention” may give a better signal and truer incidence than dividing the AEs into “pedal edema, generalized edema, peripheral edema.”
  
• Prospective grouping may be useful in clinical programs where appropriate: serotonin syndrome, parkinsonism, and drug withdrawal are useful and “not well characterized by a single term.” Other groupings may be done retrospectively, although the FDA should be consulted.

Analyzing Temporal and Other Associations: The FDA next gives a high-level discussion of factors that need to be considered in analyzing signals. They note that simple comparisons of frequency may not be sufficient to analyze a signal and that temporal associations may aid in evaluating causality. This includes duration of exposure to a product, time to onset, and so on. Various statistical methods, including the Kaplan-Meier approach, may be useful for evaluating risks of AEs. The product’s pharmacokinetic and pharmacodynamic profiles as well as “an appreciation of physiologic, metabolic and host immune responses may be important in understanding the possible timing of treatment-related AEs.” Concomitant medications, the initiation or withdrawal of other therapies, and changes in the preexisting conditions over time should also be considered.

Analyzing Dose Effect: AEs should be analyzed as a function of dose. That is, are there different AE responses at different doses? It may be useful to consider weight or body surface area-adjusted doses. Subgroup analysis by dose may also be useful. The FDA also makes the important point that “the likelihood of observing false positive signals increases with the number of analyses conducted.” Consistency across studies should be investigated to help validate such findings.

Data Pooling: The FDA notes that “data pooling is performed to achieve larger sample sizes and data sets because individual clinical studies are not designed with sufficient sample size to estimate the frequency of low incidence events or to compare differences in rates or relative rates between the test drug (exposed group) and the control (unexposed group).” Further, “pooled analyses can enhance the power to detect an association between product use and an event and provide more reliable estimates of the magnitude of risk over time…and can also provide insight into a positive signal observed in a single study by allowing a broader comparison.” The FDA discusses issues for and against use of such pooled data.

The FDA next notes that all placebo-controlled trials should be considered for data pooling. The patient populations should be relatively homogeneous and have similar methods of AE and dropout ascertainment. Phase I trials should generally be excluded. Risks should not just be expressed in event frequency (e.g., AEs per 100 persons) but also in time-event analyses when appropriate.

Rigorous Ascertainment of Reasons for Withdrawals from Studies: The FDA emphasizes that all participants dropping out must be followed up to fully understand why they did so. Some reasons may be irrelevant and trivial (e.g., moved away) or very important in regard to safety (e.g., had a stroke, was intolerant to adverse reactions). Terms like “withdrew consent,” “failed to return,” “administratively withdrawn,” or “lost to follow-up” are too vague to be useful. These reasons should be followed up for more specific causes, especially if safety issues are involved. Dropouts because of abnormal laboratory tests, vital signs, or electrocardiographic findings may not be always characterized as AEs but should be followed up and accounted for. Follow-up on safety issues should be done until the AE is resolved or stabilized.

Long-Term Follow-Up: In some instances (e.g., the drug has a very long half-life, is deposited in bone or brain, or might cause irreversible AEs such as cancer) patients should be followed to the end of the study or even after the study ends. This may mean follow-up long after the drug treatment ends. This is especially true in long-term treatment and outcome studies. The FDA recommends discussions with the agency for these special cases.

Important Aspects of Data Presentation: Finally, the FDA makes several comments on data presentation in the NDA integrated summary of safety and elsewhere:

• AE rates should be presented from more restrictive (e.g., myocardial infarction) to less restrictive (e.g., myocardial ischemia).
  
• AEs for the drug that are important and seen for other drugs in the class should be discussed.
  
• Analyses of pooled data looking at gender, age, extent of exposure, concomitant medical conditions, and concomitant medications should be included.
  
• Differential discontinuation rates (e.g., placebo-treated patients may drop out of a trial earlier than drug-treated patients) must be accounted for.
  
• Case report forms submitted for patients who died or discontinued should have relevant hospital records, biopsy reports, and so forth included.
  
• Narrative summaries of important AEs should be sufficiently detailed to allow the case to be understood and analyzed.

Most of the concepts introduced in this document remain true and reflect FDA’s thinking. However, some areas are evolving, especially in light of further ICH, CIOMS, and EU activities. Nonetheless, this document is still worth reading.

Stay updated, free articles. Join our Telegram channel

Join