Drug Safety, Climate Change, Causality, and Predictability

Abstract

Welcome to summer. As we have done in the past, there will be no Topic Update this month. We will return in August to our usual format. Instead, I have decided to write about something that has been on my mind lately—the concept of causality. I have chosen this subject because there is considerable confusion these days in the media about what is fact and what is myth or false belief. Even when there is a growing consensus about causality, there may still be doubt about when things will happen, how often they will happen, or their relevance to one’s life. Recently, some colleagues and I met to discuss possible contributions from cognitive computing and analyses of so-called big data sets to drug development and safety assessment. As I thought about what could be gleaned from comprehensive pondering, I began to wonder about the sources of bias that might enter into such efforts. These days, many of us are aware of the limitations of political polling. Both selection (sampling) and reporting bias may influence these polls. Similarly, sampling issues and bias may be important when studying drug safety. For example, how are the data from patients’ claims on the internet or from published case reports taken into account? I know from personal discussions that many clinicians who encounter an interesting or unusual adverse event choose to try to publish an article about it rather than report it to the product’s manufacturer or to the US Food and Drug Administration via a Form 3500. With the proliferation of online journals, and to a lesser extent print journals, a vast amount of untapped, sometimes unrefereed, and, in many instances, unverifiable information from clinical trials and case reports probably exists. Would these additional data sources in any way modify what we now conclude from data gathered from spontaneous reports to manufacturers, Forms 3500, clinical trials, and preclinical animal studies? I was once taught that if you presented study patients with a checklist of possible adverse effects, they were more likely to identify effects that they would not mention if just asked if they had experienced any problems. In this era of direct to consumer advertising, how do those quickly aired disclosures about rare but dangerous adverse effects affect what patients report to their doctors? I know several people who, when they see these advertisements, are reluctant to accept their clinician’s advice that these medications may be useful for them. My thoughts about causality then shifted to the subject of climate change. Most polls suggest that the majority of those polled believe that the planet is warming, and many believe that humans are the cause.1Pew Research Center. What the world thinks about climate change in 7 charts. 2016. http://www.pewresearch.org/fact-tank/2016/04/18/what-the-world-thinks-about-climate-change-in-7-charts/. Accessed May 17, 2017.Google Scholar, 2Capstick S. Withmarsh L. Poortinga W. et al.International trends in public perceptions of climate change over the past quarter century.WIREs Climate Change. 2015; 5: 1014-1020Google Scholar How do people develop their ideas about climate change? Is it from the classroom, the radio, television, magazines, or newspapers? Some people may be influenced by the teachings of their ministers. What would we learn if we could scan into a database all sermons, all newspaper columns (including op-eds), and all relevant magazine articles? Could we identify groups of individuals who would benefit from well-developed educational materials aimed at clarifying any distortions of the accumulating facts? It is possible that confirmed naysayers would dig in their heels and become more resistant, whereas there may be some who could be encouraged to open their minds. Overall the basic question is, “How do we determine causality?” Beyond that, we need to understand the limitations of predictions. There is little disagreement that freezing (at or below 0°C or 32°F) causes water to turn into ice and that heating water, once the boiling point is reached (at approximately 100°C or 212°F), turns it into steam or vapor. These scientific facts were determined by observations using accurate thermometers and by multiple replications. They were established through the invention of the mercury thermometer in the Netherlands in 1714 by Daniel Gabriel Fahrenheit; approximately a generation later, Anders Celsius proposed an alternative scaling that is now widely used outside the United States.3Benedict R.P. Fundamentals of Temperature, Pressure and Flow Measurements. 3rd ed. Wiley, New York, NY1984Crossref Google Scholar There is some variability with either scale; for example, boiling point is influenced by atmospheric pressure, which in turn is influenced by altitude. Time can also be a factor. With the same heat source, it may take a quart of water longer to come to a boil than a cup. To predict how soon water will come to a boil, volume, altitude, atmospheric pressure, and the size and strength of the heat source have to be known. One further point—a prediction is more specific than a guess. Although the science of drug safety and the science of climate change are not the same, both must be grounded in accepted criteria for establishing causality and predictability; both must also consider variability. Neither can rely on guess work. In 1965, the British epidemiologist Austin Bradford Hill developed criteria for establishing causality.4Hill A.B. The environment and disease: association or causation.Proc R Soc Med. 1965; 58: 295-300PubMed Google Scholar I paraphrase, modify, and list them according to my view of their order of usefulness: (1) temporality—the change must follow the alleged cause in a timeframe that is plausible; (2) strength—the magnitude of the effect should be large enough to overcome the idea that it is simply natural variability, although a small and consistent change could also suggest causality; (3) consistency—comparable results should be found by anybody examining data drawn from different places; (4) specificity—there should be no other equally likely explanation; (5) plausibility—the cause and effect should be linked by an understandable hypothesis or mechanism; (6) coherence—data (such as water temperature, tidal height, snow fall, or droughts) measured by different people should agree; and (7) experiment—laboratory experiments should be conducted to investigate linkages. Hill used 2 additional factors that have a bit less weight for me: (1) analogy—learning from similar causal relationships and (2) biological gradient—larger exposure should lead to greater change, although inverse relationships are also possible. When studying causality of adverse drug reactions (ADRs), the scale developed by Naranjo and colleagues5Naranjo C.A. Busto U. Sellers E.M. et al.A method for estimating the probability of adverse drug reactions.Clin Pharmacol Ther. 1981; 30: 239-245Crossref PubMed Scopus (8448) Google Scholar is frequently employed. This useful scale solicits responses to 10 questions. I have simplified them as follows: (1) previous reports with the same agent; (2) sequence of events; (3) improvement with cessation; (4) reappearance with rechallenge; (5) other possible explanations; (6) recurrence during placebo substitution; 7() measurable toxic concentrations; (8) increase or decrease with changes in the suspected agent; (9) comparable reactions to agents in the same class; and (10) support from any other objective findings. Each scale item comes with weights for different answers; score totals are then used to rate the linkage as doubtful, possible, probable, or definite. In my clinical experience and from my reading of hundreds of case reports alleging causality for drug-induced ADRs, I have found these six factors to be most important: (1) a temporal relationship within a time range that is comparable across patients; (2) the absence of another plausible explanation; (3) improvement hopefully occurring on discontinuation; (4) reappearance with reexposure but only when this can be done safely; (5) similar findings with other agents in the same class or having some degree of structural overlap; and (6) published reports of similar cases. Although this last point could come first, I have placed it last because the cases I see as a journal editor may be the first of their kind. We probably should add a new criterion involving pharmacogenomics now that such tools are becoming more readily available. Using my six criteria, any of you who read my April 2017 Editor-in-Chief’s Note will readily understand why I concluded that the observations linking pemoline and hepatic failure were not conclusive.6Shader R.I. Risk evaluation and mitigation strategies (REMS), pemoline, and what is a signal.Clin Ther. 2017; 39: 665-669Abstract Full Text Full Text PDF PubMed Scopus (1) Google Scholar I feel that causality for that drug-ADR pairing was best classified as somewhere between possible and probable; it surely was not definite. Curiously, in a single-assessor evaluation using the Naranjo items as well as the World Health Organization’s (WHO’s) criteria to score 913 clinical reports on ADRs from a number of different drugs, causality was most frequently determined to be at the possible level on both scales.7Belhekar M.N. Taur S.R. Munshi R.P. A study of agreement between the Naranjo algorithm and WHO-UMC criteria for causality assessment of adverse drug reactions.Indian J Pharmacol. 2014; 46: 117-120Crossref PubMed Scopus (75) Google Scholar In this same study, there was no agreement between the two scales for the similar rankings definite and certain. The WHO items for the rating certain are similar to my items 1 through 4, although worded slightly differently.8World Health Organization (WHO) – Uppsala Monitoring Centre. The use of the WHO-UMC system for Standardized Case Causality Assessment. http://www.who-umc.org/Graphics/24734.pdf. Accessed June 1, 2017.Google Scholar However, the WHO scale also has a fifth item that I take for granted: “Event definitive, pharmacologically or phenomenologically (i.e. an objective and specific medical disorder or a recognised pharmacological phenomenon).”8World Health Organization (WHO) – Uppsala Monitoring Centre. The use of the WHO-UMC system for Standardized Case Causality Assessment. http://www.who-umc.org/Graphics/24734.pdf. Accessed June 1, 2017.Google Scholar My concerns about the difficulties in determining causality for ADRs are echoed in a study by Arimone and colleagues.9Arimone Y. Miremont-Salamé G. Haramburu F. et al.Inter-expert agreement of seven criteria in causality assessment of adverse drug reactions.Br J Clin Pharmacol. 2007; 64: 482-488Crossref PubMed Scopus (60) Google Scholar This group recruited five heads of regional centres of pharmacovigilance in France or of “departments of pharmacovigilance of the pharmaceutical industries on the basis of their experience in the field, i.e. at least 10 years of daily routine practice”9Arimone Y. Miremont-Salamé G. Haramburu F. et al.Inter-expert agreement of seven criteria in causality assessment of adverse drug reactions.Br J Clin Pharmacol. 2007; 64: 482-488Crossref PubMed Scopus (60) Google Scholar to rate 31 summaries of ADR-drug pairings (the published report sometimes states 31 and other times 30). These experts were asked to use seven criteria and then to provide their assessment of causality. The criteria were similar to all the approaches noted above. Rather than using the outcome concepts as categories, they were asked to use a visual analog scale to express their opinions as to causality. Once done, the 100-mm lines were deconstructed into seven levels: (1) ruled out, (2) unlikely, (3) doubtful, (4) indeterminate, (5) plausible, (6) likely, and (7) certain. Straightforward results, such as ruled out and certain, were rarely chosen. More than 50% of the pairings were considered plausible or likely. The pattern of responses across the seven items varied considerably; time to onset was the most consistently selected. The study authors concluded that agreement among their expert raters was very poor and that doubtful was the sole category for which there was good agreement. I conclude from all this that establishing causality is often difficult. However, there are a few elements that, when available, could facilitate better determinations. These elements are (1) a relatively consistent time of onset; (2) the pathobiology of the ADRs is recognizable and similar across patients; (3) drug concentrations in tissue or blood exceed what would typically be expected; and (4) known pharmacogenomic vulnerabilities are present. If clinicians are trained to use these criteria and to include them in their reports and write-ups, the work of the pharmacovigilance and postmarketing surveillance experts who study this public health problem would be much easier. If the currently trending political views prevail, there will be more emphasis on safety and less on efficacy; a consistent approach to determining causality is essential. Returning to the issues of climate change and global warming, some climate change deniers may never accept currently available scientific information. It is not at all obvious what facts would encourage those who deny climate change and global warming to rethink their beliefs. For example, it is difficult to reconcile that voters in Louisiana overwhelmingly supported a presidential candidate who opposes international and federal efforts to curtail greenhouse gases with the fact that residents in their own state required relocation because of rising water levels in the Gulf of Mexico. In the first allocation of federal dollars provided by the US Congress for this purpose in 2016, Isle de Jean Charles in coastal Louisiana received a $48 million grant to relocate its population to higher ground. Even a new term has been coined to describe this relocated population—climate refugees. Similar evacuations are taking place for the villagers of Newtok in Alaska. Solely presenting more facts to the general public about sea and lake water temperature changes, heights of tides, rising sea levels, or the disappearance of certain land masses is not going to work. Instead, we need to begin with children in school. We need to develop easily understandable demonstrations, and we need to create and widely distribute a simple, less data-based narrative, perhaps illustrated by satellite and onsite photography. What follows is a sequence of simple demonstrations for children that I think might be helpful. I believe that most people would accept that ice displaces its own weight in water—this could be illustrated by showing ice cubes placed in a glass of water. This concept should then be linked to the loss of the polar icecaps and the melting of glaciers. There are photographs in real time that are very dramatic. It then becomes understandable that ocean levels should proportionately rise. Evaporation can be shown to provide more moisture into clouds and increased rainfall. This can be done by showing people how a drop of water on a countertop disappears and how it will disappear more quickly if it is near a heat source and more slowly when surrounded by ice cold soda bottles. As the ice melts at both poles, currents move the cold water toward the equator. In my mind, here is where some confusion sets in. This example simplistically suggests that we should be talking about global cooling rather than warming. In fact, I believe we have seen some harsher winters of late. However, melting is only one aspect of a complex positive feedback loop, within which it is both a consequence of increasing greenhouse gases and a cause of further warming. A variety of experiments with glass jars, thermometers, water, and ice can be used to teach children about the greenhouse effect. The National Aeronautics and Space Administration maintains a very informative and appealing website for children, parents, and teachers.10National Aeronautics and Space Administration. Climate Kids. https://climatekids.nasa.gov. Accessed June 1, 2017.Google Scholar It would be great if we could teach these ideas to our children, and if they, in turn, could help their parents to a better understanding of what may lie ahead. I suspect that many of you could come up with even better ways of conveying these and similar ideas. To be honest, I feel helpless about how we can ever put such ideas into the vastly different curricula that exist around the world and even within the United States.