Chapter 15: Evaluation of Medical Causation
By: Marks, D.H. , O’Donnell JT editor
Synopsis:
15.1 What is Causation
15.2 Importance of Establishing Causation
15.3 Definitions
15.4 Degree Adverse Effects Relate to Intervention
15.5 Methodology to Investigate Causation
15.6 Structured Algorithms for Determination of Causation
15.7 Comments on the Individual Ridell Criteria for Causation
15.8 Daubert and the Evolution of Causation
15.9 Does Causation Need to be Established Before a Warning is Given?
15.10 Summary
15.1 What is Causation
Medical causation is the determination whether of an adverse effect was caused by the use of a medication, biological, vaccine, device or procedure, generally referred to as treatments. An adverse effect refers to an untoward physical sign, symptom, abnormal assessment (lab, VS, ECG, etc), or cluster of signs, symptoms, abnormal assessments within the definition of FDA reg 21CRF.1 Physical signs can include fever, hypertension, weight loss or other physical findings. A symptom can include any complaint from a patient, including nausea, headache, abdominal pain or others. A medication can refer to any FDA approved prescription or to an over the counter preparation. Biologicals refer to prepared of synthetic materials of living origin. Vaccine are preparations designed to induce a protective or therapeutic immune response. Medical devices include both large units (MRI) and miniature units (intravascular stents).
15.2 Importance of Establishing Causation
Patients who receive a treatment for an illness generally can be expected to have, as a result of their illness, physical signs and symptoms as manifestations of their illness. As a general statement, all treatments also can cause their own array of physical signs and symptoms separate from the underlying illness being treated, and these new signs and symptoms are referred to as adverse effects. This illustrates the crux of medical causation’s difficulty – distinguishing adverse effects of treatments from signs or symptoms of the underlying illness they are being given to treat.
This problem is illustrated by the medication Lotronex (Alosetron, Glaxo Wellcome), an antagonist of the 5-HT3 (serotonin) receptor. Lotronex was indicated for the treatment of severe, chronic, diarrhea-predominant irritable bowel syndrome (IBS) in women who have failed conventional therapy. IBS patients can typically present with a wide range of GI symptoms and signs. Unfortunately, Lotronex can act both as serotonergic drug and as a Serotonin Reputake Inhibitor (SRI). Serotonin is a potent neurotransmitter (NM) which activates a muscle at the neuromuscular (NM) endplate. The intensity and duration of the muscle activation is directly related to the concentration of serotonin at the NM end plate. Lotronex mimics serotonin at the active receptor and blocks serotonin resorption, thereby causing an inordinately high level of serotonin activity in the GI NM endplate. This high serotonin level can be associated with severe vasoconstriction and ischemia.2
These elevated serotonin levels are known and accepted causes of many of the serious side effects of Lotronex, including obstruction, perforation, impaction, toxic megacolon, and secondary ischemia. This led to the quandary that the signs and symptoms of an adverse effect3 from the medication – Lotronex, which could have warned of an adverse event (AE) and led to the drug’s discontinuation, were similar to the not unexpected signs and symptoms of the underlying disease being treated. Ultimately this difficulty of use led to the removal of Lotronex from use.
Another example is the development of depression leading to suicidal thought from use of metoclopramide (Marks 2007).4 This drug has both central (nausea) and peripheral (gastric motility) actions, can be an antagonist of dopamine, and sensitize gastric smooth muscle to the effects of acetylcholine stimulation. Metoclopramide has a varied CNS effect, including drowsiness, extra pyramidal syndrome (dystonias, akathesia), depression, dizziness and insomnia. It should not be surprising that a drug with antipsychotic efficacy, and which can cause akathesia , may cause an increased risk of suicide, as is pointed out in the prescribing information for Reglan.
15.3 Definitions
In general, both the pharmaceutical industry and the regulatory bodies use the same definitions for adverse effects, and for the FDA these are contained in 21CFR.1 Practicing clinicians who observe potential AE, and patients who experience AE, can use different definitions or meanings of commonly used terms such as causation, probable, possible, severe, serious. This can lead to some confusion when AE are reported to regulatory agencies. This potential for discrepancy is one of the reasons why primary Medwatch reports need to be thoroughly evaluated, rather than simply tabulated. Medwatch reports can supply critical information on unsuspected, previously reported adverse events (for example heart valve thickening and pulmonary hypertension after FenPhen), or an incidence rate which is greater than previously known (rhabdomyolysis with Baycol).
Adverse events: any unfavorable and unintended diagnosis, symptom, sign (including an abnormal laboratory finding), syndrome or disease which either occurs during the study, having been absent at baseline, or, if present at baseline, appears to worsen.
Serious adverse events (CFR 312.32): any untoward medical occurrences that: (1) result in death, (2) are life threatening, (3) require (or prolong) hospitalization, (4) cause persistent or significant disability/incapacity, (5) result in congenital anomalies or birth defects, or (6) are other conditions which in the judgment of the investigators represent significant hazards.
15.4 Degree Adverse Effects Relate to Intervention
Causation can be described as definitely, probably, possibly or unrelated to the treatment. Each of the degree of relatedness has definite meanings, and their structured and consistent application is important for patients receiving the medication and for prescribing physicians. These terms are defined in 21CFR1, and their use is consistent throughout the pharmaceutical industry, the FDA, CDC and WHO. The determination of a degree of causation must occur within structured guidelines,5 such as those of Koch, Hill 6,7 or Riddell.8 A practical definition scheme for assessing the degree of causality is as follows:
Unrelated:
The adverse event is clearly due to extraneous causes (e.g., underlying disease, environment)
Unlikely (must have 2):
The adverse event:
1) does not have temporal relationship to intervention,
2) could readily have been produced by the subject’s clinical state,
3) could have been due to environmental or other interventions,
4) does not follow a known pattern of response to intervention,
5) does not reappear or worsen with reintroduction of intervention
Possible (must have 2):
The adverse event:
1) has a reasonable temporal relationship to intervention, 9
2) could not readily have been produced by the subject’s clinical state,
3) could not readily have been due to environmental or other interventions,
4) follows a known pattern of response to intervention
Probable (must have 3):
The adverse event:
1) has a reasonable temporal relationship to intervention,
2) could not readily have been produced by the subject’s clinical state or have been due to environmental or other interventions,
3) follows a known pattern of response to intervention,
4) disappears or decreases with reduction in dose or cessation of intervention
Definite (must have all 4):
The adverse event:
1) has a reasonable temporal relationship to intervention,
2) could not readily have been produced by the subject’s clinical state or have been due to environmental or other interventions,
3) follows a known pattern of response to intervention,
4) disappears or decreases with reduction in dose or cessation of intervention and recurs with re-exposure
15.5 Methodology to Investigate Causation
Signs or symptoms associated with a treatment can be observed in a number of settings. They can be seen during the study of unlicensed drugs (clinical studies), after licensure (post-marketing), or for unlicensed herbals and food supplements. Clinical studies beyond the initial and characteristically small Phase I study are designed primarily to demonstrate efficacy. Safety data are collected as secondary parameters. Because clinical studies are varied in their design, and depend in large amount on the disease studied, the drug, and the types of information sought (safety, dosing, efficacy, disease-specific), clinical studies can be conducted in a randomized, blinded or non-randomized, non-blinded setting. It is for these reasons that referring to the randomized clinical trial (RCT) as a gold standard for demonstrating causation is overly simplistic and in some senses deceiving. Important safety data which has allowed the discovery of new indications for already-licensed drugs, and caused the removal from market of other drugs, has even come from a few cases presented in collected case series.10,11,12
Some of the structured setting in which adverse effects can be evaluated include:
RCT – however, almost always efficacy is the primary parameter, and safety is collected as a secondary parameter. An example would be adverse effects which are collected during the efficacy trial of an antihypertensive medication. Data can be collected actively (for example, actively questioned for, and data collected in case report forms, or passively collected, as when patients volunteer information on adverse effects that are not actively questioned for).
Epidemiologic studies – non-randomized, non-blinded studies which can collect incidence and prevalence data on adverse events in a specific population.
Case Controlled studies – patients who exhibit an adverse event are matched with multiples of demographically similar but non-exposed patients13,14,15
Individual case reports16 or series – physicians (used here as a general term to include other health care providers) publish their observation of novel, unreported adverse effect in an individual patient (a case) or group of patients (a series)
15.6 Structured Algorithms for Determination of Causation
Koch was perhaps the first to set down rules of causation, albeit not for adverse events. The famous Koch postulates were designed to demonstrate whether a disease was caused by an infectious organism. Some have attempted to adapt Koch’s postulates from infectious disease to drug AE causation, but this has generally been unsatisfactory.
Koch’s Postulates:
1) The suspected agent must be isolated from a patient with the disease in question.
2) The agent must be grown in the laboratory in pure culture.
3) The isolated agent, grown in pure culture, when infected into a healthy host produces exactly the same clinical disease.
4) One is able to isolate the same infective organism from the newly diseased person.
Not only are Koch’s postulates not useful for determining AE causation for medications, they are also not practical for many diseases, including HIV (intentional application of the third postulate). Certainly, analogies to drug adverse causation can be used from the other postulates, as have analogies in other systems, such as those of Hill and of Ridell.
Hill and also Riddell established systematic rules for evaluating potential drug adverse effect causation, by setting up a set of eight or nine rules. Not all of their rules need to have been met for causation to be established, and some are of a higher importance than others. One of the first rules to meet is temporality, for it is obvious that individual or complex signs and symptoms of a putative adverse event which occurs before the introduction of a treatment could not have been caused by the treatment. However, the symptom or symptom complex, even if pre-existing, can be exacerbated by the introduction of the medication. Rather than address both Hill and Riddell, I shall concentrate on Ridell, as their rules, as presented in the following table, overlap.
Hill Criteria17 / Ridell CriteriaStrength of association / Temporal eligibility
Consistency of results / Latent period
Specificity / Exclusion
Temporal relationship / De-challenge
Dose response / Re-challenge 18,19
Biologic plausibility / Singularity of the drug
Biologic coherence / Pattern
Experimentation / Drug Identification
Analogy
Hill Criteria – Expanded Discussion
1. Strength of Association
A strong association gives support to a causal hypothesis
A weak association requires other information but can be equally as important
2. Consistency of Results
Repeated findings in different populations and different settings
3. Specificity
Strengthens confidence in association.
Lack of specificity does not rule out causation
4. Temporal Relationship
Required: exposure most come before disease
5. Dose Response
Increased dose = increased risk
Holds for drugs and vaccines (doses)
6. Biological Plausibility
Known mechanism not required.
Cigarettes and lung cancer
Asbestos and lung cancer
Fen-Phen and valvular heart disease and PPH
7. Biological Coherence
Does not conflict with what is known
8. Experimentation
RCT is close to experimentation.
Removal or reduction of exposure reduces disease.
Challenge – dechallenge - rechallenge
9. Analogy
Similarities with other like exposure
Aminorex and PPH
Ergot drugs and VHD
All of these systems for determination can be simplified in an algorithm-based analysis:
Temporal relatedness
Known or reported AE of Medication
Presence of Concurrent Illnesses or medications which could present similarly
Challenge – Dechallenge - Rechallenge
15.7 Comments on the Individual Ridell Criteria for Causation
Temporal Eligibility: As a general principle, an adverse event which precedes the introduction of the treatment can not be causally related. An area of conflict is how long after the introduction of the treatment can an adverse event be attributed to the treatment. Some drugs, such as PCP and Lariam, are known to have neuropsychiatric adverse effects that can last for years after the drug is last taken.
Latent Period: Although the closer the interval between the introduction of the treatment and the adverse effect the better, a long latent period can still be consistent with a causal relatedness. Some adverse effects can have immediate action, or a latent period in seconds, such as immediate hypersensitivity reactions. Other adverse effects are more subtle, although equally lethal, such as acetaminophen liver toxicity.
Exclusion: It is important to rule out the effect of concurrent medications or underlying medical conditions as alternative hypothesis. Liver enzyme elevations in a diabetic patient with underlying fatty liver, and simultaneously taking a statin for hyperlipidemia, can be a challenge for determination of causation. Regardless, with careful attention to the timeline and details, even in such cases, a determination can be make.
De-challenge: The disappearance or improvement of an adverse event after the withdrawal of the treatment is a positive indication of relatedness. Adverse events can however persist after the treatment is withdrawn, which indicates either that there was no relatedness, or that the adverse effect is persistent injury.