Article Reviewed: Conceptual Problems in the Definition and Interpretation of
Attributable Fractions - Sander Greenland and James M. Robins1
The population attributable fraction is defined as the reduction in incidence that would be achieved if the population had been entirely unexposed, compared with its current exposure pattern2. In recent days, the idea of attributable fraction gaining more importance, because of its vital role in interventions, lawsuits and legislation regarding harmful exposure3. A critical analysis of perceptions and drawbacks in calculating and understanding the attributable fractions provided by Greenland and Robins in this article. They have mentioned that the Conceptual problems in the definition of an attributable fraction, which was due to the lack of incident time when considering the exposure role in disease aetiology, creates further problems in the valid assessment of exposure impacts. They also stated that the definition of attributable risk was misinterpreted in textbooks, such as some books said that only the excess cases contribute to the attributable fraction while some others imply that all etiologic cases give to the attributable fraction and this reflected equally in algebraic definitions. Greenland and Robins demonstrated the difference between the excess fraction and etiologic fractions in this article with three examples. The etiology fraction is defined as the case fraction in which the exposure has played a contributing role in the etiology of the outcome/disease and in the excess fraction, the excess cases occurring during some period among the exposed population in comparison with the unexposed. All the excess cases consist in etiologic cases, but the vice versa is not true. It is important to quantity the aetiology fraction in the viewpoint of both biology and law. In most epidemiological studies, the population attributable fraction measures only the excess fraction which was smaller than the etiologic fraction.4 They also included the incidence density fraction in which, the attributable fraction is the ratio of the difference between the incident densities when exposure present (ID1) and absent (ID0) and the incident densities when exposure present i.e. (ID1-ID0)/ID1 which was explained by the fourth example. The incidence density fraction has two subtypes by the use of densities, instantaneous and average. According to them the relevance of the measure depends on these three fractions and not by a single measure. Even though the years of life lost is a more relevant measure of exposure impact than the etiologic fraction, this is not always true due to the relevance strongly hang on social and ethical issues. The etiologic fraction is used largely in legal aspects such as compensation for harmful exposures, however without resorting the strong biologic assumptions one cannot evaluate the etiologic fraction. In further issues, they dealt with the severity of outcome with transition times, susceptible proportions termed by Khoury et al., casual and preventive cofactors. They have mentioned that the use of attributable fractions in compensation can have serious issues due to the complex biologic interactions and the estimation techniques of attributable fraction should be interpreted with caution. They hoped that these three concepts, such as excess fraction, etiologic fraction and incident density fraction, in whole may resolve the problems in interpreting the attributable fraction in the epidemiological study as well as in legislation. Article Reviewed: Health Equity and the Fallacy of Treating Causes of Population Health as if They Sum to 100%5 - Nancy Krieger, PhD This article, "Health Equity and the Fallacy of Treating Causes of Population Health as if They Sum to 100%", by Nancy Krieger seeks to address how the sum of the health causes components cannot be 100%. The author examples three reputed articles and she identified two errors in them, which are the misuse of population attributable fractions and error in equating the health causes. In the first example, the author using the article "The Causes of Cancer: Quantitative Estimates of Avoidable Risks of Cancer in the United States Today" by Doll and Peto6. In their 1981 publication, they presented quantitative estimates of the proportion of cancer deaths and used the population attributable fraction to distribute the risk. However, Nancy Krieger argued that the valid estimation and interpretation of the population attributable fraction had several well-known coercions and mentioned the Greenland and Robins illustration of lung cancer caused by smoking and asbestos exposure. In Doll and Peto's 1981 article, the "proportions of cancer deaths attributed to various factors" equated to just below 100%. However, in 1985 update by Peto added a new line for "total" percentage was equal to 200 or more," with the note admitting that PAFs can quantity to more than 100% when a cancer has more number of causes including nature, nurture, and luck. In second example, the author criticised the article "Variation in cancer risk among tissues can be explained by the number of stem cell divisions"7 by Tomasetti C, Vogelstein B. in which they claimed that the primary prevention actions are not effective for tumours of relatively low incidence arising in organs undergoing many stem cell divisions, because they arise mostly from random mutations fixed during stem cell division, and they concluded that if 65% of cancer risk is attributable to chance then only one third is attributable to environmental factors or inherited tendencies7. Nancy Krieger mentioned the two criticisms of this article and added Rose's 1985 article regarding the distribution of cases and contributing fractions not sum to 1 by Weinberg and Zaykin et al., In the third example, Nancy Krieger has mentioned two initiatives in the United States, which are County Health Rankings & Roadmaps and America's Health Rankings. Both projects are intended to influence public and policymakers' awareness of health issues and to guide the allocation of resources to improve health rankings and health equity. However, the total component factors contributions are equal to 100% in both initiatives and even by the flawed logic of the approach used, that the county health ranking weightings should effectively be shared and a large unidentified factor should be included. Forcing causes of population health to sum to 100%, whether framed in terms of PAF or via wrongly conflating explaining variation with explaining causation, is not accurate, methodologically or substantively. The author explained the important reason for this scientific error which was due to the longstanding concern within the health sciences of pitting nature compared to nurture8 and questioning their contribution in population health. the population health and the effort to reach health equity are ill-served do not require the methodological approaches that err in assuming causes sum to 100%. The cause can be individually allocated among nature, nurture, and luck and these components are equally exclusive9. The 100% fixed equating is not possible and explained by one more reason that the population attributable fraction and disparity percentage are factually dependent. Conclusion: Attributable Fractions estimators take several different forms of calculation depending on the study design, covariates adjusted risk ratios additional to the exposure variable, and the information of the covariates/exposure prevalence patterns. The population attributable fraction estimates can help guide policymakers in planning public health interventions despite numerous articles on population attributable fraction estimation, errors in computation and interpretation persist. Greenland and Robins debated a large flaw in the reasoning of attributable risk, in the distinction that they noted between excess and etiologic cases. They provided a thorough discussion of why the population attributable fraction usually greatly underestimates the proportion of disease burden10 that is related etiologically to the exposure. According to the authors, the large number of cases in which exposure plays an etiologic role although not necessarily the causally attributable role of pushing a person from the non-diseased into the diseased group by a specified point in time, is not included in the attributable fraction. Nancy Krieger pointed that the previous historical debates and her critical analysis revealed the fallacy of health causes if they equated to 100% despite its dominance in numeral epidemiological studies and she added it is an intensely flawed interpretation that causation can be construed as nature versus nurture and the imprecise assumptions and biological methods compromises the public health claims. Finally, Nancy Krieger concluded that more focus required on the validation and appropriate methods to measure the health inequalities and their elimination. Thus, Greenland and Robins emphasised the problems in the attributable fraction interpretation and Nancy Krieger argued about the faulty methods in the quantification of percentage variation and the valuation of the population attributable fraction. In both articles, it is mentioned that the characteristics of epidemiological measures depends on the eco-social theory of disease distribution. There are certain reasons to distrust the hypothesis that the population attributable fraction provides useful information about partitioning or attributing the causes of disease in the population. The model of sufficient component causes11 dictates the population attributable fraction for different exposures considered one at a time will usually sum up to far greater than 100% for a given outcome. References:
1. Greenland S, Robins JM. CONCEPTUAL PROBLEMS IN THE DEFINITION AND INTERPRETATION OF
ATTRIBUTABLE FRACTIONS. American Journal of Epidemiology. 1988;128(6):1185-1197. 2. Rothman KJ. Modern epidemiology. 3rd ed., [thoroughly rev. and updated] ed. Philadelphia: Wolters Kluwer Health/Lippincott Williams & Wilkins; 2008. 3. Kelsey JL. Methods in observational epidemiology. New York: Oxford University Press; 1986. 4. Kleinbaum DG. Epidemiologic research : principles and quantitative methods. New York: Van Nostrand Reinhold; 1982. 5. Krieger N. Health Equity and the Fallacy of Treating Causes of Population Health as if They Sum to 100%. American Journal of Public Health. 2017;107(4):541-549. 6. Doll R. The causes of cancer : quantitative estimates of avoidable risks of cancer in the United States today. Oxford ; : Oxford University Press; 1981. 7. Tomasetti C, Vogelstein B. Cancer etiology. Variation in cancer risk among tissues can be explained by the number of stem cell divisions. Science (New York, N.Y.). 2015;347(6217):78. 8. Jevons WS. English Men of Science; their Nature and Nurture. Nature. 1874;11(270):161. 9. Wylie F. Nature or Nurture?: Neither! Australian Life Scientist. 2009;6(6):26, 28. 10. Levine B. The other causality question: estimating attributable fractions for obesity as a cause of mortality. International Journal of Obesity. 2008;32(S3):S4-7. 11. Poole C. Commentary: Positivized epidemiology and the model of sufficient and component causes. International Journal of Epidemiology. 2001;30(4):707-709.