Phenomena: Exploring the "natural history" of disease


Idea: Detailed observation (like a naturalist) or detective work--albeit informed by theoretical ideas--may be needed before we can characterize what the phenomenon is we are studying, what questions we need to ask, and what categories we need for subsequent data collection and analysis.

Initial notes from PT


1. The initial motivation for this class was to highlight that epidemiology does not necessarily begin with data sets to analyze. There may be exploratory, investigative, detective, anthropological, and naturalist inquiries before phenomena are even noticed, categories are defined, questions are framed. Good examples of this seemed to be provided by John Snow's work on cholera, Barker's[1] research in Uganda, on "clues from geography" of infant mortality and heart disease, and the three Lancashire towns, and Oxford's account of the conditions that provided a source for a global pandemic of the 1918 flu. Even Barker's speculation about anomalous French cardiovascular disease rates looks like someone who is able to connect dots of diverse kinds and that are spread out in time.

2. Brody's paper, in addition to drawing attention to the role of maps in this exploratory research, makes the Snow story more complicated and interesting. Snow had clear hypotheses that guided his mapping and his advocacy of stopping the water supply from the Broad St pump -- he was certainly not simply noticing patterns in the data and hypothesizing about the causes. This account opens up broader questions in philosophy of science.

2a. The way I questions is not simply to contrast "induction versus deduction ," but to identify a chain of steps in scientific inquiry in which each step involves assumptions and is open for negotiation and wider influences (Taylor 2005, chapter 2). There is also the possibility that desired outcomes for the later stages (especially the actions planned) influence decisions made at earlier steps.

All possible phenomena



Reference: Taylor, P. J. (2005). Unruly Complexity: Ecology, Interpretation, Engagement. Chicago, University of Chicago Press.

[1] Barker is a recently retired but still active epidemiologist whose reputation is linked to the "Barker hypothesis" that chronic diseases of later life are associated with fetal or early-life conditions. We'll address this hypothesis within the frame of "life course epidemiology" in week 9.

Draft ideas/thoughts for discussion (to be incorporated into the substantive statement later on)


Reading accounts of past puzzles solved – like the story of the Broad St. pump, the mysterious roots of the 1918 influenza pandemic, the slow and methodical building of a case for fetal origins of later diseases, and the breadth of factors surveyed to finally identify Mycobacterium ulcerans as the culprit in buruli disease – is at once satisfying and misleading. Satisfying because we have the pleasure of seeing the “cases closed” and misleading because there is nothing inevitable about achieving success every time.

I think that these cases challenge us to reflect on what we mean by “the process of epidemiological inquiry” – can it only be detected after the fact? The articles present us with events in which all the dots have been connected, but how much time was spent on a dot or observation that did not pan out? Peter refers above to a chain of steps in scientific inquiry in which each step involves assumptions and is open for negotiation and wider influences; do we know enough from the articles to speculate on what some of those sites of negotiation might have been? [PT: not really – see above]

My experience with the discipline of epidemiology is very scanty at this point in time. It is confined to having read and thought about another earlier epidemiological triumph – the discovery by Joseph Goldberger (with help from economist and statistician Edgar Sydenstricker) that pellagra was a nutrition deficiency disease. This case was initially built on painstaking, systematic, and methodical observation – not analysis from data sets (although, of course, data from the observations gathered by Goldberger formed the basis of his case) – which raises the question, could such a ‘‘shoe-leather” epidemiological inquiry take place today?

J.G. Elmore (1994) writes that, “In today's environment, a single investigator would probably be unable to move as gracefully as Goldberger from community observation to clinical experiment to the laboratory bench, producing excellent examples of both public health research and clinical investigation in disease.” Do you agree with this statement? What is it about ‘today’s environment’ that seems to make this kind of epidemiological work less likely?

Finally, returning to Peter’s notes above: if we wanted to draw back and look more abstractly at scientific inquiry in general, does the framework of the steps assist your understanding of what is in play?

Elmore, J.G. (1994). Joseph Goldberger: An Unsung Hero of American Clinical Epidemiology. Annals of Internal Medicine, 121:372-375. http://www.annals.org/cgi/content/full/121/5/372


Substantive statement

Phenomena: Exploring the “natural history” of disease

An examination of phenomena goes to the heart of what epidemiology can do and what epidemiology is for. For example, is it the function of epidemiology (and epidemiologists) - through its knowledge base and methodologies - to lessen the burden of disease in individuals and in populations, or is its function to study “how disease is distributed in populations and the factors that influence or determine this distribution”? (Gordis, 2004) The reality is that it is not, of course, an either/or proposition. Epidemiology both strives to identify the cause(s) of disease as well as to provide a scientific base from which regulatory, health care, and public policy decisions are made.
[PT: Consider an alternative to “as well as,” namely, the two functions are linked -- the kinds of explanations that people seek or accept about the factors that influence the distribution, incidence, and timing of disease might reveal the ways they think disease can be influenced by policy and other public health measures.]
To adequately address the idea that “observation…may be needed before we can characterize what the phenomenon is we are studying, what questions we need to ask, and what categories we need for subsequent data collection and analysis” (PT in syllabus), we need to know more about the discipline of epidemiology and how it has evolved; what has driven the choice of topics and methods of analysis; what scale or level of organization of phenomena have historically been selected for investigation; and how these and other factors impact epidemiology’s relationship to public health.
[PT: The general history of epidemiology is interesting as is the specific background to any study we read, but I do not agree that we need to know all this before admitting that observational work can have a role in epidemiology. Barker’s buruli work shows that – at least in that case – there was a role for exploratory observation before he could offer plausible speculations about the distribution, incidence, and timing of disease.]

[PT: Given the previous comment, the following survey by Jan is interesting, but possibly not essential for moving ahead through the topics of the course. Admittedly, as is often the case, it is an open pedagogical question: Should a teacher start with context so that students can position the details in that context, or does context seem abstract and not very relevant until a certain level of detail and specifics have been introduced to the students.
Another way to look at Jan’s survey is to view it as an elaboration on the debate Rose initiated about focusing epidemiological research and public health high-risk individuals versus focusing on measures that can be applied to the population as a whole – see week 2 of the course.]
A survey of selected articles that discuss the history and debates within epidemiology is a start in that direction. Such a survey reveals that there have been many interpretations and perspectives relevant to these questions. [That this is a selective survey also demonstrates the question under consideration: i.e., because the literature on debates in epidemiology is large and multifaceted, I have made choices among the articles that resonate with my interests. So, rather than honing on those articles that discuss technical and methodological differences in epidemiological studies, I have selected articles that discuss the historical evolution of the conceptual differences between social- or eco- and ‘individual risk’ approaches.]

In the mid-nineties, the “epidemiology wars’ broke out (Poole & Rothman, 1998). Representative of that debate is an article by S. Schwartz, E. Susser, and M. Susser (1999) entitled, “A Future for Epidemiology?” The three authors reviewed what they considered to be the then current state of epidemiology as well as offering their prescription for an approach that would overcome the constraints of so-called risk factor paradigm – namely, eco-epidemiology. An advantage of this approach over the older chronic disease epidemiology was that it could “address… the interdependence of individuals and their connection with the biological, physical, social, and historical contexts in which they live” (p. 26). For a variety of reasons - including the historically fraught relationship between medicine and public health (Brandt & Gardner, 2000) -the direction epidemiology took in the mid-twentieth century was reductionistic and emphasized single or individual risk factors and ignored group or population level factors. Schwartz and the Sussers characterize this “paradigm” as being too preoccupied with description of risk-factor/disease rather than explanation of causal processes (p.23).

The research study designs and analytical techniques that developed to accommodate this type of thinking included more sophisticated case-control and cohort studies and statistical modeling. These changes can be thought of as apolitical but may also reflect one of the contentions made in the “idea” for this week: that the ‘chain of steps in scientific inquiry…involve assumptions and [are] open for negotiation and wider influences.’ In other words, the choices epidemiologists (or their employers) make on what is important to study, or on the kinds of studies that get funded, are rooted in decisions made at many levels throughout the public health, academic, and governmental arenas. As Schwartz, Susser and Susser put it:


This theme was taken up in another article by Schwartz and Carpenter (1999) that addressed a fundamental discrepancy between the research question being asked and the methods used to address the question. An example of this would be studies hoping to discern the causal factors for the rise in obesity that examine interindividual differences in body mass. As they point out, interindividual differences in obesity at any point in time are not likely to provide clues to the causes of the rate increase. Schwartz and Carpenter state that since an infinite number of threats and causes could be investigated, decisions are made to limit the range of inquiry and that these decisions are “determined not only by scientific concerns but also by social, political, and economic considerations” (p.1175). Furthermore, they argue that additional value-laden constraints arise from research methods themselves because “methods are designed to examine certain types of problems that, while critical to study, may nonetheless lead to restricted types of answers” (p. 1175).

Neil Pearce did not mince words in his 1996 article, “Traditional epidemiology, Modern Epidemiology, and Public Health,” when he proclaimed that “we seem to be using more and more advanced technology to study more and more trivial issues, while the major causes of disease are ignored” (p. 678). This statement reflects the crux of the difference in viewpoint in terms of what epidemiology is for. At the risk of over-simplifying, individual-level risk factor proponents seem to prefer not to get into discussions of how the health of the public can (or should) be improved; it is enough to demonstrate an association between a risk factor and a disease. Proponents of the population perspective (be it social epidemiology or eco-epidemiology) want to strengthen (or return) epidemiology to its more traditional position as a branch of public health (p. 681).

In 1999, Nancy Krieger took issue with the framing of the debate as one between “risk factor” and “social” epidemiology. Krieger’s use of the analogy of a fly meeting its demise in a sticky spider’s web is a readily understandable and amusing account of the many factors that are involved in causation. She uses the analogy to bring up the conceptual issues that characterize the way investigators whose preference is for the micro, the macro, or the meso-levels continue to speak past each other. While not as mundane as asking whether the glass is half-full or half-empty, her point is that epidemiologists need to grapple with the “simultaneity” of these levels. Additionally, and relevant for our topic, she states that “the point is that all epidemiological studies include data on variables hypothesized to be component causes of the specified outcomes, thereby raising critical questions as to which variables are included, which are excluded, and why (p. 678).
[PT: It may be relevant to note that Nancy Krieger teaches her core epidemiology course as a historical survey.]]

A 2006 article by Eric J. Duell has the ‘future of epidemiology’ in its title. Duell sees the debate in epidemiology a decade ago between individual risk-factor level and population-level advocates as mirroring today’s landscape where the role of genomic data is now the context for discussion. He advocates for “a multilevel epidemiology that would seek to understand multiple levels of inference, from genes to individuals to populations and could combine hypothesis-driven research with aspects of data mining” (p.623). Making the interesting point that genetic factors should be thought of as “having both individual-level and population-level attributes,” he says that although we measure genetic markers in individuals,… genetic marker behavior (how they are maintained in populations) is a group-level phenomenon (p. 625). One can only speculate on how many more micro-levels we will uncover in the attempt to disentangle the etiology of disease.

These articles were selected to illustrate the contention that ‘what gets done’ in science is not value-neutral. Even in epidemiology, with its traditional charge to identify and solve problems in public health, factors outside the discipline are at play that can affect what gets noticed, what is deemed important, the choice of research methods used, the evidence that is presented (and by whom), and the outcomes achieved. It may be true that epidemiology does not begin with the analysis of data sets, but given the literature in this mini-survey, at least we now have some idea why such an idea may have taken hold.

Exploring the natural history of a disease, as in the story of Snow’s Broad Street pump or Goldberger’s “filth parties” may seem anachronistic today, where “one is left with the impression that 19-century epidemiologists used ad hoc methods that have now been placed on a sounder foundation through recent developments in methods of study design (e.g., the theory of case-control studies); data analysis (e.g., logistic regression); and exposure measurement (e.g., new molecular biology techniques)” (Pearce, 1996). An optimist would say, though, that only the tools have changed – epidemiological thinking is still alive and well. (JC)

References

Brandt, A., and Gardner, M. (2000). Antagonism and accommodation: interpreting the relationship between public health and medicine in the United States during the 20th Century. American Journal of Public Health, 90:707-715.

Duell, E.J. (2006). The future of epidemiology: methodological challenges and multilevel inference. Bundesgesundheitsbl – Gesundheitsforsch – Gesundheitsschutz, 49: 622-627.

Gordis, L. (2004). Epidemiology. 3rd ed. Philadelphia: Elsevier.

Krieger, N. (1999). Sticky webs, hungry spiders, buzzing flies, and fractal metaphors: on the misleading juxtaposition of “risk factor” versus “social” epidemiology. Journal of Epidemiology & Community Health, 53: 678-680.

Pearce, N. (1996). Traditional epidemiology, modern epidemiology, and public health. American Journal of Public Health, 86:678-683.

Poole, C. and Rothman, K.J. (1998). Our conscientious objection to the epidemiology wars. Journal of Epidemiology & Community Health, 52: 613-614.

Schwartz, S. & Carpenter, K.M. (1999). The right answer for the wrong question: consequences of type III error for public health research. American Journal of Public Health, 89:1175-1180.

Schwartz, S., Susser, E., and Susser, M. (1999). A future for epidemiology? Annual Review of Public Health, 20: 15-33.

Response to Substantive Statement
Jan does a thorough job describing epidemiology concepts in terms of epidemiology phenomena and the expectations of this branch of (medical)
science, dealing with the study of the causes, distribution, and control of disease in populations. There will always be this discourse of which process of studying diseases is correct: the study of diseases and its affect on population or the study of an individual with a disease within a population. This is evident in Charlton (1995)’s critique of Roses “population strategy’ for preventive medicine discussed in week 2 annotated references. When I first read Jan’s substantive statement, I was unsure of what direction to focus my response. As Jan had presented several articles depicting the “evolution” of epidemiology and the ongoing debates of “how” to study diseases, I referred back to this week’s assigned readings to try and determine which processes were utilized. For example, in Oxford, et.al (2004), the purpose was to examine the origin, evolution, and spread of an epidemic virus in the early 1900’s. Data was collected by soldiers and pathologists in the army and recorded in great detail. Epidemiologists examined this data for clues; ways to possibly prevent this viral epidemic from occurring again. As I read the article, I found myself referring back to Peter’s chain of steps in scientific inquiry outline, starting with “phenomenon deemed interesting” all the way to the end of “actions supported.” Several conclusions resulted from this study, but the action that was supported is the need to watch potential epicenters for another possible epidemic; attempting to slow its progression. Snow (Brody et.al, 2000) based his “plot map” of cholera deaths after he tested his hypothesis; yet history leads us to believe he determined the outbreak of cholera from initial data plotted on a map. Both of these men followed a chain of scientific inquiry.
Assumptions on any disease process needs to be open to negotiations; one can never “assume” as no two individuals are the same; hence, each may respond differently depending on biological, physical, and social factors. Epidemiology investigations are needed to address all the different challenges relating to human diseases; whether it is an individual or a public health issue. The question that continuously keeps coming to mind is the following: Does it matter if we study preventative methods to control individual diseases vs. large populations that may be exposed to a small risk? It seems to me that both can only serve to help the overall population in many ways. (SA)
[PT: Although this week’s readings were not about Rose’s argument, let me respond to SA by saying that I read Rose as saying yes, both individual and population approaches can help, but yes, it does matter which is emphasized. The Rose article spells out the advantages and disadvantages that any researcher and practitioner needs to take into account.]

References
Brody, B., Rip, M., Vinten-Johansen, P., Paneth, and N., Rachman, S. (2000). Map-making and myth-making in Broad Street: the London
cholera epidemic, 1854. Lancet, 356: 64-68.
Charlton, B. (1995). A critique of Geoffrey Rose’s ‘population strategy’ for preventive medicine. Journal of the Royal Society of Medicine,
88: 607-610.
Oxford, J., Lambkin, R., Sefton, A., Daniels, R., Elliot, A., Brown, R., and Gill, D. (2004). A hypothesis: the conjunction of soldiers, gas, pigs, ducks, geese and horses in Northern France during the Great War provided the conditions for the emergence of the “Spanish” influenza pandemic of 1918-1919. Vaccine, 23: 940-945.

Revision/addendum to substantive statement

I began my preparation for week 3 with the assumption that most epidemiological research began with detailed observation of phenomena. I could hardly think otherwise since up to this point in time I had only read about Joseph Goldberger and pellagra and the 3 assigned readings (all of which included descriptions of the "sleuthing" that the investigators had to do). I guess in my ignorance I thought that that's how all epidemiological inquiries were conducted. I felt the articles did a good job of presenting their sleuthing so I didn't choose to recapitulate what they said.

Instead, as part of my own wider explorations of what shapes epidemiological inquiries, I tried to look into the history and more recent debates about the proper character of epidemiologic inquiry. This approach probably took me in directions other than those intended for this week, but I tried to relate the survey of articles to the “chain of steps in scientific inquiry” – which also formed part of the topic. I tried to tie the debates in epidemiology to the question of why certain topics get pursued in epidemiology. The point of my questions following the Elmore article was to note that the day of epidemiologists like Snow, Goldberger, and even to a certain extent, Barker, seems to have waned. The epidemiologist as “medical detective” solving cases through sheer observation, intuition, and solid data collection, it turns out, is actually far removed from how epidemiological inquiry proceeds today.

In an attempt to learn more about the kinds of studies being published in professional journals right now, I did an informal survey today of original contributions published in the American Journal of Epidemiology, the Journal of Epidemiology & Community Health, and Epidemiology. What I found was that many, many studies DO seem to start with data sets; for example, this random article:

Sarah E. Hill, Tony Blakely, Ichiro Kawachi, and Alistair Woodward
Mortality among Lifelong Nonsmokers Exposed to Secondhand Smoke at Home: Cohort Data and Sensitivity Analyses
Am. J. Epidemiol. 2007 165:530-540; Full-Text.
http://aje.oxfordjournals.org/misc/free_articles.dtl

[from the abstract]
Evidence is growing that secondhand smoke can cause death from several diseases. The association between household exposure to secondhand smoke and disease-specific mortality was examined in two New Zealand cohorts of lifelong nonsmokers ("never smokers") aged 45–77 years. Individual census records from 1981 and 1996 were anonymously and probabilistically linked with mortality records from the 3 years that followed each census. Age- and ethnicity-standardized mortality rates were compared for never smokers with and without home exposure to secondhand smoke (based on the reported smoking behavior of other household members)....

In this article, the researchers never met the people involved in the study, never interviewed them, they only used the records from the New Zealand Census-Mortality Study. Looking through the titles of articles (and going into the Materials & Methods sections in particular) I found that very few studies actually went out and observed anything. There were a few exceptions, but I had to look long and hard for them. Consider this study published in a recent issue of the American Journal of Epidemiology:

Mario Schootman, Elena M. Andresen, Fredric D. Wolinsky, Theodore K. Malmstrom, J. Philip Miller, Yan Yan, and Douglas K. Miller

The Effect of Adverse Housing and Neighborhood Conditions on the
Development of Diabetes Mellitus among Middle-aged African Americans
Am J Epidemiol 2007;166:379–387

[from the abstract]
The authors examined the associations of observed neighborhood (block face) and housing conditions with the incidence of diabetes by using data from 644 subjects in the African-American Health Study (St. Louis area, Missouri). They also investigated five mediating pathways (health behavior, psychosocial, health status, access to medical care, and sociodemographic characteristics) if significant associations were identified. The external appearance of the block the subjects lived on and housing conditions were rated as excellent, good, fair, or poor. Subjects reported about neighborhood desirability. Self-reported diabetes was obtained at baseline and 3 years later…

This study used two independent observers who rated each of five characteristics: condition of houses, amount of noise (from traffic, industry, etc.), air quality, condition of the streets, and condition of the yards and sidewalks in front of homes where participants lived. Professional interviewers (two thirds of whom were African American) who had extensive project-specific training contacted households in person for structured interviews, and so on.

I found that this kind of study, where the researchers actually went out and made observations first and subsequently conducted various statistical analyses on the data, is definitely in the minority. It will be interesting in the weeks ahead to see whether the research we encounter has more in common with the statistical manipulation of data obtained from large, population studies like the example from New Zealand or the more descriptive, observational studies of Snow and Goldberger. Or perhaps something in between…?


Annotated additions by students

[PT: JG added a reference that was more about the Barker hypothesis than the role of keen observational sleuthing work, so I moved it to week 9 on life course epidemiology.]

See Trostle book on Epidemiology and Culture.


Response to Peter's inquiry.


In addition to Peter’s response on my entry to Jan’s substantive statement, it was recommended to go back to Rose’s articles and review the advantages and disadvantages that a researcher needs to take into account when evaluating which preventative approach is emphasized: population vs. individual. As I reviewed again his reasoning, it still brings me back to my original thought of helping all persons. Although Rose presents his views on both; outlining advantages and disadvantages, there will always be those clinicians/researchers who think that one way is better than the other and will continue with their individual viewpoint. But as I reread the last paragraph that Rose wrote, it partly reiterates what I had said in my original response: “…many diseases will continue to call for both approaches.” However, in the end, Rose believes that “the priority of concern should always be the discovery and control of the causes of the incidence.” (SA).