Schmalhausen's Law

Ivan Ivanovich Schmalhausen1 was a Soviet evolutionary biologist working at the Academy of Sciences in Minsk. In the 1940's his book "Factors of Evolution" appeared and was denounced by T.D. Lysenko, whose neo-Lamarckian theories of genetics were then on the ascendency. At the close of the 1948 Congress of the Timiryazev Academy of Agricultural Science it was revealed that Stalin had endorsed Lysenko's report to the Congress in which it was affirmed that the environment can alter the hereditary makeup of organisms in a directed way by altering their development. A number of opponents of Lysenko's views then took the floor to reverse themselves and support Lysenko. Schmalhausen was one of the few who reaffirmed his opposition and spent the rest of his life in his laboratory studying fish evolution and morphology.

Click to read more ...


Schmalhausen;s Law



            Ivan Ivanovich Schmalhausen1 was a Soviet evolutionary biologist working at the Academy of Sciences in Minsk.  In the 1940's his  book "Factors of Evolution" appeared and was denounced by T.D. Lysenko, whose neo-Lamarckian theories of genetics were then  on the ascendency. At the close of the 1948 Congress of the Timiryazev Academy of Agricultural Science it was revealed that Stalin had endorsed Lysenko's report to the Congress in which it was affirmed that the environment can alter the hereditary makeup of organisms in a directed way by altering their development. A number of opponents of Lysenko's views then took the floor to reverse themselves and support Lysenko. Schmalhausen was one of the few who reaffirmed his opposition and spent the rest of his life in his laboratory studying fish evolution and morphology.

            In the West, Lysenko's views were simply dismissed. An assistant professor of chemistry, Ralph Spitzer, lost his job for daring to suggest that Lysenko's ideas should at the least be examined and tested.  But Schmalhausen could not ignore the Lysenko agenda which insisted on a more complex  interpenetration of  heredity and environment than genetics generally recognized. Along with Marxist and progressive scientists in the west  such as C.D. Waddington in the UK, he accepted rather than ignored the challenge. As a result he developed a more sophisticated approach to these interactions which explained the observations of  the better studies cited by Lysenkoists.

            Schmalhausen argued that much of natural selection is stabilizing rather than directional. That is, if a species is more or less well adapted to its conditions "the tendency to vary" that Darwin invoked causes the population characteristics  to become more spread out about its average state, and selection eliminates these variants. Genes are selected which work together with the other more common genes in such a way that most of genetic combinations  produce more or less viable and similar offspring under more or less normal conditions. However under unusual or extreme conditions where selection has not had the opportunity to operate, these genetic differences show up as increased variation.  This claim provided an alternative explanation to the observation that  populations that are apparently uniform under normal conditions show a wide range of variation under new or extreme conditions. Whereas Lysenko argued "that these populations were uniform and that the environment created new genetic variation, Schmalhausen argued that the environment revealed latent genetic differences which could then be selected.

            Waddington developed this line of reasoning further with his idea of genetic assimilation: suppose that there is some threshold condition in the environment for the development of a particular trait. Much below threshold none of the individuals show it, much above threshold they all do. But under some intermediate conditions some will be above and some below threshold. If those whose threshold is lowest are selected, we can select for low threshold and eventually produce organisms whose threshold is so low that the trait always appears under any conditions in which the organism can survive. Then the trait had become "assimilated": an environmentally induced condition had become fully genetic.

The 1950's and '60s was a time of great interest in the variability of traits in populations.  In Dobzhansky's laboratory at Columbia University, where both of us raised and counted fruit flies, there was interest in showing that genetically heterozygous individuals could tolerate a wider range of conditions than those that were genetically homozygous. Attempts were made to link population genetics with physiologcal homeostasis and the stability of development.. Some researchers in the same tradition examined the differences in the number bristles on two sides of the same fly to determine the ease with which differences in conditions on a microscopic scale could influence development. (When the scale is small enough, these microscopic events are designated "random" since we cannot influence them.)

            Schmalhausen's law is more general than the consequence of stabilizing selection and has implications in many areas.

            In biogeography: At almost any location on the earth, the ecological community is made up of species near the boundary of their distribution and also species that are in the middle of their range. When the environment changes, this has a major impact on the species near their boundary. Some may become locally extinct, other may experience great expansions of their abundance and their range, while others will remain more or less as they have been. Further, populations near their boundaries are especially sensitive to changing conditions and are more likely to show big differences from year to year. Therefore simple predictions about the effect of climate change are bound to err if they take into account only the direct physiological impact of the environmental change on species one at a time.

            The thresholds of toxicity: Tolerable levels of toxic substances are often set on the basis of experiments with animals. Usually the work is done with standardized healthy animals under well controlled conditions so as to minimize "error" due to individual differences or variation in the environment. However this methodology underestimates the impact of a toxin for a number of reasons. If an organism is exposed to a toxic substance of external or internal origin, it has various mechanisms to detoxify that substance. But thetoxin is still present., If there is a constant level of exposure, the toxin will reach some level of balance between new aborption of toxin and the rate of removal. This equilibrium depends on the level of exposure and the maximum capacity of the detoxification system to remove the poison.

            But of course we know that the environmental exposure is not constant for all members of a population or even for any one individual over time. And we also know that different members of the population differ in their detoxification capacity and that it may vary over time for the same person. Furthermore, this variability matters and cannot be averaged away.

             What good is a model that assumes constant conditions? Here we see one of the powerful ways in which models are useful in science. In physical and engineering sciences it is often possible to isolate a problem sufficiently to ignore external influences, assume that all switches are the same in what is relevant, that all salt molecules are interchangeable and so on. Then we can measure accurately and get equations that are as exact as we need. But in ecological and social sciences this is not possible. The populations are not uniform, conditions change and there is always an outside impinging on the system of interest. We cannot even believe the equations too literally. But we can still study these systems. First  we find the consequences of models under unrealistic conditions that are easily studied and give precise results. Then we ask, how do departures from those assumptions affect the expected outcomes? In this case, the standing level of toxicity, a measure of damage done to organism, is a mathematical function of v-s, the maximum detoxification capacity minus the exposure.  The maximum removal rate has to be greater than the exposure or according to the math the toxicity will accumulate without limit.  In reality, it will accumulate to the point where other processes which were negligible in the original model, take over. These might involve any of the consequences of toxicity such as cell deaths. When v is greater than s the graph of  toxicity plotted against v-s  decreases from zero as capacity exceeds exposure by greater and greater amounts. Furthermore, it is concave upward. That is, it is steeper the closer we are to v=s and flattens out when capacity is much greater than exposure.  If we measure the dose response curve in the range where capacity is much bigger than exposure then the results will show little effect of the poison and we will be reassured by claims that there is no detectable effect. Testing is often carried under optimal conditions on uniform populations of experimental animals in order to get uniform results, reduce the error, and avoid "confounding factors".

            If different stressors are confronted by the same detox pathways, they can be added at the level of  exposure and act synergistically at the level of toxicity. Therefore if we look at only one insult at a time, the other "confounding factors" increase the damage.

            In the United States, exposure varies with location and occupation. The poor, excluded and marginalized communities such as inner cities, colonias and reservations are often subject to multiple exposures due to incinerators, maquiladoras, poor water quality, malnutrition and unsafe jobs. Therefore even toxics which meet EPA standards will prove more harmful than expected. But these effects will be hard to detect since we wil observe an array of health impairments rather than a single harm appearing to different degrees.

            Similar arguments hold if the capacity to detoxify varies among individuals: the average toxicity in the population is greater than the toxicity at average detox capacity. Once again, if detox capacities are reduced then each unit of insult has a bigger effect than expected.

            We suspect that detox capacities are undermined in the course of life for all of us after the first two decades, but that adverse conditions accelerate this erosion so that vulnerability increases more rapidly and life expectancy is reduced by some [fill in] years for African-American women and [].

The variability of results: Since, when v-s is small, small differences in either one can have big effects, a population which is at a disadvantage will show big differences among people for reasons we cannot explain, and different poor communities will differ widely in the rates of adverse outcomes. This can easily be misinterpreted: it appears as if under the "same" conditions some do well and others badly, and we can then blame those who do badly. But what really is happening is that under conditions of any kind of stress, small differences have big effects.




Review of The Society and Population Health Reader. Vol. I. Income Inequality and Health.

Review of The Society and Population Health Reader. Vol. I. Income Inequality and Health. Kawachi, Ichiro, Bruce Kennedy and Richard Wilkinson, editors. The New Press, New York.

One of the more surprising finds of modern archeology is that in the transition from gatherer/hunter societies to agriculture, health deteriorated. Skeletal remains show smaller stature, indicators of tuberculosis, horizontal lines on teeth that indicate arrested growth, periodontal disease and other suggestions that "progress" isn't always good for people. The adoption of agriculture brought with it the possibilities of permanent and elaborate homes, increased food production, and long term storage of food and water. People were now productive enough to be exploitable. The transition brought with it hierarchy and then classes, to the detriment of health.

Click to read more ...


Review of The Society and Population Health Reader. Vol. I. Income Inequality and Health. Kawachi, Ichiro, Bruce Kennedy and Richard Wilkinson, editors. The New Press, New York.

            One of the more surprising finds of modern archeology is that in the transition from gatherer/hunter societies to agriculture, health deteriorated. Skeletal remains show smaller stature, indicators of tuberculosis, horizontal lines on teeth that indicate arrested growth, periodontal disease and other suggestions that "progress" isn't always good for people. The adoption of agriculture brought with it the possibilities of permanent and elaborate homes, increased food production, and long term storage of food and water. People were now productive enough to be exploitable. The transition brought with it hierarchy and then classes, to the detriment of health.   

            It has long been known that poverty is bad for people. Engels' vivid account of working class conditions and their impact on health in 19th century England is one of a long series of exposés. In the United States, where "class" is not a category of sociological or epidemiological research, proxies such as Socio-Economic Status (SES) or family income have been used to underline that deprivation imposes health costs. After a long hiatus during which biochemical markers or personality type were used as the major predictors of disease, social determination of health has reemerged as a major research interest. The growing gap between rich and poor has become obvious to everyone.  Despite the conservative argument that poverty is merely a statistical convention in affluent societies (every distribution has a lower tail!)  or a free life style choice, it is clear that the poor die sooner, recover more slowly from the same diseases as the rich, have more disability, and judge their own health as poorer.

            In the last decade or so, a highly productive group of researchers has formed in the United Kingdom and North America under the rubric of "social determinants of health". A major role in their coming together as a group was played by the Robert Wood Johnson Foundation and the Association for Health Service Research. Under the leadership of Alvin Tarlov and Barbara Krimgold, researchers were supported and connected to each other, meetings were held and findings publicized. The research of the "social determinants of health" school differs from the older emphasis on poverty in that it argues that inequality affects health throughout the gradient of income and status and not just at the lower end, and that inequality per se, independently of absolute income level, is deleterious to health.

This book is a compendium of key writings from this movement and others with overlapping interests, with a few contributions from their critics. It systematically develops a very strong case that inequality itself is a cause of poor health outcome. It is a robust conclusion: whether health outcome is measured as life expectancy at birth, life expectancy at age five, infant mortality, male and female mortalities taken separately, self-rated health, or mortality due to specific treatable causes such as heart disease, cancer, or stroke the conclusion is the same. (Interestingly, the association did not show up for untreatable causes of death). International comparisons, comparisons of states or metropolitan areas in the United States or wards in the United Kingdom give the same result. Whether the measure of inequality or relative deprivation is the fraction of total income received by the top 30%, or the share of the top 5% or inversely the bottom 20%, the Gini index, or the Robin Hood Index ( the proportion of the income of the rich that has to be taken away and given to the poor in order to have complete equality) the outcome is the same: relative deprivation is bad for your health. The critics of this claim find fault with particular statistical methods or data sets, or the possibility of confounding the effects of other factors. But despite any possible weaknesses of particular studies the evidence is robust and overwhelming.

 First, international patterns are examined . A graph plotting life expectancy against per capita income shows that up to about $5000 per capita income, life expectancy does rise with income. But beyond that point the relationship disappears and for any income level beyond that there is a spread of life expectancies. Among 11 countries of the OECD there is a very strong correlation between life expectancy and the proportion of the income earned by the least well off 70% of  the population while GNP itself shows only about a 35% correlation.  This leads Wilkinson to "…ask whether further undifferentiated economic growth is worth the environmental risks."

Studies within countries comparing states or provinces also show a strong association of inequality and poor over- all health. In all, more than 15 quite diverse studies support the general conclusion.

Although it is clear that more egalitarian societies have better health outcomes, the question of why some places are more egalitarian than others is not examined, and politics is omitted as influencing health.  An exception is Wilkinson's (ch 24) discussion of the health patterns of eastern Europe before 1989. As a result of the elimination of extreme poverty, these countries , as well as Viet Nam, China, North Korea, Mongolia and Cuba showed good life expectancies for their GNP's until the 1970's. Thereafter, eastern Europe declined relative to western Europe. Further, Albania, which did not revert to capitalism until later also suffered its decline in health later, while Cuba continues to improve.

An obvious explanation for the worsening of health outcomes in these countries is the loss of a uniformly accessible health care system. This is important especially in the coverage of childhood immunizations and in prenatal care. But medical care, for all its importance, is not the major determinant of a nation's health. The economic collapse, unemployment, the delays in wage payments, the loss of economic security and of an ideology of equality, cooperation and collectivity must also be playing their part.

             Once we have established that inequality is a "risk factor" for ill health, the next question is, how does this come about.

The editors suggest three plausible pathways by which income distribution could affect health: "a) by influencing individuals' access to life opportunities and material resources (such as access to health care, education, employment, decent housing; and b) through social processes related to social cohesion, e.g. through mutual support and the benefits of better communication and cooperation; and c) through directly psychosocial processes related for instance to latent social conflict and the quality of social relations." Evidence is presented supporting roles for each of these although the various authors stress the importance of different ones. Wilkinson prefers the last explanation. Kawachi and Kennedy develop this further in examining "social capital", "social cohesion", and "trust" as contributing to the psychosocial processes through perceptions of inequality that are stressful, frustrating, depressing or in other ways undermining.

Lynch and Kaplan suggest that "…inequitable income distribution may be associated with a set of social processes and policies that systematically underinvest  in human, physical, health and social infrastructure and that this underinvestment may have health consequences."(p212). They show in particular that places with higher inequality are also places that invest less in social income such as education. These authors also raise the question of whether income might underestimate inequality and that wealth might be a more revealing measure. (A further step would be to measure the concentration of ownership of capital. This is even more unequal than wealth, or "net worth", because the "wealth" of poorer people is concentrated in a home or car rather than means of production.)

In order for a social context of disease to really make people sick, it has to reach the physical body. Several authors look particularly at stress as initiating processes that raise the cortisol level, impair the hippocampus region of the brain and from there pass through the hypothalamus-pituitary-adrenal axis and on to make mischief with blood pressure, inflammation, the immune response, and bone formation and to risks of heart disease and memory impairment. (Bruce S. MacEwan, chapter 32).  Furthermore, this system has feedbacks that can create vicious circles as when stress impairs the stress-response mechanism and this itself amplifies stress. Sapolsky, Alberts, and Altmann(ch 35)  report on animal studies that support the general model. In their work, increased cortisol levels were associated with social subordination or social isolation among baboons.

The application of animal work to people always has to be evaluated with care. Animals of similar social rank are not organized as a class in or for itself, and while animals individually may dispute their places in a hierarchy, that hierarchy itself is not challenged and therefore does not require constant justification. Nor are social hierarchies the same in different species. Yet stress is universal in the experience of vertebrates. Physiological mechanisms for confronting stress have evolved over perhaps half a billion years. A way of understanding the similarity and difference among animal species is the following: the responses and activities of an organism are organized into packages of processes, feedback loops, correlated and coordinated actions. Each of these packages is tightly connected internally, with each part depending strongly on the others. Walking involves many muscles and nervous signals, leg movements, compensating posture adjustments, signals that verify that the actions are being carried out, visual and other information. But where you walk to or why you walk is not part of the package. These packages are relatively conservative in evolution because they function within similar mechanical and physiological constraints, have similar adaptive significance and are so tightly linked internally that changes in any of the parts would require corresponding changes in the rest of the package. Therefore their evolution is much more a coadaptation of parts to each other than adaptations to some external environments. In particular, the physiology of the stress response system of other animals is very similar to ours.

But these packages are relatively loosely linked to each other. The sensory/interpretive packages that determine that a situation is stressful are not tightly connected to the stress response itself. What is stressful varies among animals, and in the higher animals is more deeply embedded in a broader context of past and present experience. The cerebral cortex, our principle link to the social context, becomes part of the nervous system networks and transforms the activity of older parts of the nervous system so that a fluctuation in an economic indicator can be as stressful to some people as a fire or an unexpected encounter with a tiger. Therefore we are not justified in inferring from animal studies what is stressful for us. Nevertheless animal studies do demonstrate that social experience is translated into physiological states through adaptive mechanisms that can be eroded, and that these physiological states affect health. This work helps to break down the artificial psychological/physiological dichotomy.

While one line of inquiry goes from an individual's experience to physiology and pathology, research also proceeds in the opposite direction, from the individual to society.

 A common property of the social determination explanations is that they do not depend on the individual alone but also on "the environment", here usually conceptualized as the community. That is, there are community level processes which promote or undermine health regardless of the socioeconomic status of an individual living there. A poor person in a rich neighborhood may have to pay more for everything so that what income she has is less effective in buying necessities. The common culture assumes individual resources that not everyone has, such as a computer for doing homework in school. Or a poor neighborhood is near an incinerator and in that neighborhood the lungs of rich and poor alike are poisoned. Thus the Social Determination of Health school leads also into community and societal level factors. We are led to seek a characterization of unhealthy neighborhoods. One definition identifies a high-stress neighborhood as one in which there is low socio-economic status, high population density, high geographic mobility, high rates of marital breakup, and high crime. These neighborhoods are also usually areas of poorer public services, often exposed to more toxic pollutants and frequently also the pervasive pall of racism.

Taylor, Repetti and Seeman's chapter ( 31) asks, "What is an Unhealthy Environment and How Does it Get Under the Skin?" They point out that low-SES neighborhoods have higher rates of cancer, heart disease and upper-respiratory disorders including  asthma, bronchitis, and emphysema."(p356), and that women in high-violence neighborhoods are significantly more likely to experience pregnancy complications than women living in neighborhoods with little violence.            

Crowding is another aspect of poverty. People in high-density living situations have more infections, higher death rates from heart disease, respiratory disease and all-cause mortality. They report more chronic stress and also excrete more adrenalin in the urine. (But this crowding effect may be partially offset: large households also provide more mutual support, and when this is lost people are more vulnerable to economic and social stressors. This may be a factor in the anomalous data showing that first generation Mexican immigrants have better health outcomes than the next generation who have lost the support system of their own culture without having escaped from United States poverty patterns. )

The same  authors also recognize that adults spend much of their waking hours at work away from home, and that the work environment is a major aspect of "environment". The occupational health field has a long history of attention to injury and toxicity. Chronic stress is increasingly seen as a risk factor. The particular pathways from each of these macro-level explanations to health outcomes must link up with the physiological mechanisms that we already suspect or know to be relevant. Here the work of Karasek and Theorell is cited: it is not so much the work demand that undermines health as the association of a high level of demand with a low level of decision making latitude. In these situations there are higher rates of heart disease and a lower ratio of catecholamine to cortisol.  Harmful behaviors such as drug addiction are also more prevalent in high-strain jobs, and among employed women, working 40 or more hours a week (outside the home) is a risk factor for low birth weight babies.

But while work can be harmful to health, so is unemployment or the threat of unemployment. Working class status is a risk factor regardless of whether one is working or not!

The home as habitat and as the workplace (or one of several workplaces)  for women is not explored further although some of the materials used in home construction, furnishings, cleaning materials and appliances have been identified as harmful. Gender aspects of inequality are largely ignored in this volume.

The effects of class (seen here indirectly as "low SES") is strong, pervasive, acts through many pathways and shows up at the levels of individual behavior, communities and populations. Unfortunately, class is not dealt with as such. The treatment  is mostly limited to particular aspects of class and therefore lacks the richness of earlier sociological studies such as Sennett and Cobb's "The Hidden Injuries of Class".

The search for particular causal pathways connecting health outcome either to the physiological processes or to the social context can serve several purposes. First, elucidating the causes reinforces the evidence for the phenomenon. If we know that poor people live in poor neighborhoods, that poor neighborhoods have more contact with cockroaches and that cockroaches are allergenic, then perhaps we know one reason why asthma is more common among the poor.

         Secondly, separating out causes allows us to separate interventions according to how fundamental a change they require and which interests will clash before the improvement is made. Some interventions are almost cost-free and can be introduced after overcoming only mild inertia. For instance signs reminding you not to drink while driving or to wear your hard hat in the construction area can be adopted without serious struggle. Others involve costs: the removal of lead from houses, the clean-ups of dumps, improving school lunches. They will meet resistance from the direct perpetrators who do not want to redo their houses or remove their dumps or lose the school lunch contract, and also resistance from those who oppose public funding for such purposes,  but struggles over these issues can be won and are not so threatening to the power structure as to bring the capitalist class as a whole to the barricades.

Sometimes medical effort alone can make an improvement. For instance in the Boston area, ambulance service for low birth Afro-American and white babies now make them equally likely to reach the emergency room on time. The difference in their survival relates to the fact that on the average Black low weight babies have lower weights than white low weight babies and therefore their lungs are less able to respond to treatment. Or, oral rehydration is a very effective therapy for cholera. In recent outbreaks throughout the world the case mortality rate has been quite low in most places. This is a matter of medical effort and rapid access to health service. But the pandemic continues: it is related to water quality and poverty and would require widespread public works. Finally, the issues of restructuring work, guaranteeing jobs or reducing the work day would clash with the prerogatives of management and we would expect fierce resistance. Most public health recommendations focus on the minimal conflict measures.

The omission of class from the theoretical structure is more than a semantic question. Among other things, it de-emphasizes work as a factor in the environment of working people. There are a few passing references to studies that show that in addition to obvious dangers on the job from accidents and toxic substances, three aspects of the work environment are especially important for health outcomes:  "the amount of control people have over their work, the pressure of work, and the social support they get from colleagues." The social support category might include anything from a friendly pat on the back in the morning to militant shop stewards' intervention against harassment by the employer's representatives. Since the work environment is not further explored, we do not have data comparing organized and unorganized work places or the effect of the higher degree of class organization in Sweden and the United Kingdom compared to the United States. Both in looking at the work place as a habitat and in analyzing the meaning of social support, the search is directed toward identifying  "factors" rather than examining how the whole system works as a system.

         Social support is studied  mostly in terms of Putnam's notion of "social capital". Interest in community cohesiveness was stimulated in part the Rosetto study: the town of Rosetto, Pennsylvania, showed better health outcome statistics than surrounding towns although it was not richer, and in keeping with its Italian-American culture consumed high levels of fat in their diet. As long as the population was more or less of a homogeneous class make up, people were linked by many strong and weak ties. But once income differentiation intervened and the community became more like other stratified American towns, the health outcomes regressed to those of the surrounding area. The Rosetto case has been cited repeatedly in support of the idea that health is determined not at the individual but at the community level and that social cohesion is a major component. In principle this could refer to all kinds of mutual aid, encouragement, validation, solidarity and community action. But social cohesion is difficult to measure, data hard to come by. Kawachi, Kennedy[] use data from a published survey which asked people whether they trusted their neighbors. Suprisingly, this single questionaire answer was a good predictor of health status in a community.

         Sampson, Raudenbush and Earls (ch 30) develop the concept of collective efficacy as an alternative to "social capital" for characterizing neighborhoods. While "social capital" lumps such things as church attendance, feelings of trust and measures of cohesion regardless of what people cohere for, collective efficacy deals directly with a community's capacity to exercise informal social control over potentially harmful behaviors and to extract resources from outside to meet community needs. It is therefore a more activist measure than social capital. These authors are particularly interested in factors influencing violent crime. Not surprisingly, "economic stratification by race and place thus fuels the neighborhood concentration of cumulative forms of disadvantage, intensifying the social isolation of lower income, minority and single-parent residents from key resources supporting collective social control." And this in turn strongly affects the homicide rate.(Since homicide, suicide and motor vehicle accidents are major causes of death in young people, they have become increasingly important objects of study by public health). The emphasis on the neighborhood does not of course deny that the neighborhood also, like the individual, is formed in  a larger context.

         The studies of neighborhoods and communities have been done fairly recently, in a conservative period where organized activism has diminished and people are more likely to be immersed in the pursuit of individual goals. This can create a misleading notion that poor and oppressed populations are inherently less coherent or capable of collective action than the more recent middle class activism, often around education or the environment. This has not always been the case. During the height of labor and civil rights struggles there was often strong sense of solidarity and collective efficacy among the deprived sectors that now seem passive and helpless.

Since the main approaches taken in these studies as a whole do not explicitly deal with the phenomena of a class society, no attempt is made to integrate the various elements of inequality, deprivation, physical exposures and pyscho-social stressors into a coherent model of class and health. Instead, the various ways in which inequality harms people are treated as "factors", and much statistical ingenuity goes into trying to identify the major particular causes of ill health in order to separate them from "confounding factors", (other causes), to assign them relative importance and to make recommendations for intervention accordingly.

The book makes good use of the accepted methods of multivariate statistics and is sensitive to the difficulties of inferring cause from association, especially when the various "factors" are themselves highly correlated. Any criticism of the use of multivariate statistics for explanation is more general than a review of this book would allow. However several observations may be made.

In epidemiological researches of this kind, a fundamental distinction is made between "independent" and "dependent" variables. The hypothesis that is examined is that the independent variable causes differences in the dependent variable directly or indirectly, although what is detected is association between variables. Thus for example crowding is not only associated with more infection, it acts to "cause" more infection because there is closer contact between infected and susceptible people. This gives a positive correlation between crowding and disease. But the same two variables may have more than one kind of connection. If an extended family also supports people economically, boosts their immune systems, or keeps closer control over dangers threatening children, then a large household, seen in our data as crowding, reduces disease and there will be a negative correlation. When the two kinds of processes operate together, either one may predominate or they may cancel each other out statistically so that crowding would show no association with health outcomes even though it is affecting health strongly.

The usual procedure correlates the presumed causal factors with the outcome. Correlation is a linear relation. The strongest correlation (+/- 1) occurs when the "dependent" variable changes in the same proportion as the "independent" variable. But if the relation is not linear, for instance if at high concentrations of a pollutant small increments have a bigger effect than at low concentrations, the fitting of a straight line to this data will be less close and the correlation weaker. Depending on whether the data is drawn more from populations with high exposure or low exposure, the estimated impact of pollutants on health will be different, and the strength of the correlation will depend on how spread out the exposures are.

Further, the models usually make the distinction between dependent and independent variables too absolute, not allowing for feedback between them.  Wilson and Daly (ch 27) point out that age-specific mortality, a classical outcome or dependent variable, also feeds back on its own causal factors, "inspiring a 'rational' escalation of costly tactics of social competition", particularly in relation to crime, violence and the timing of child bearing. For instance, the work of Arlene Geronimus  is cited in which she shows that with a reduced life expectancy, early childbearing is a rational choice, increasing the chances of being around to raise the child and having available grandparental support.

These are positive feedbacks in that the actions taken reinforce the conditions that gave rise to them. There are also negative feedbacks, in which an initial condition reduces the circumstances that produce it. Suppose for example that a toxic substance in the blood can induce cancer, so that people with higher concentrations of the poison are more likely to have cancer. But suppose also that tissues more prone to developing cancer are the ones that absorb more of the toxin from the blood. Then people who are more likely to develop cancer may show lower concentrations in the blood! Here, if people differ in their exposure to the poison the dominant pathway is, poison in blood®(+) cancer. But if there are wide individual differences in susceptibility of their tissues to cancer, then the other branch of the feedback,  cancer-prone tissues®(-)poison in blood, and the correlation would be negative. Once again we could get inconsistent results among different studies or no correlation in spite of strong causation.

Or suppose that one variable acts by way of another. For instance suppose that low income kills people in a famine through lack of food. A model which correlates income, food intake and risk of death might show that lack of food is strongly correlated with death, and that once this correlation is taken into account then income has no effect on dying! In epidemiology, correlating the risk of heart disease with factors such as cholesterol reduces the statistical importance of class location that acts partly by way of cholesterol.

When independent variables are themselves intercorrelated, as are the various aspects of class position, one procedure would be to select the variable with the strongest correlation, and then see whether the other variables are also correlated after the first variable is taken into account. This association of stronger correlation with more likely causal importance can be very misleading for the reasons described above.

Finally, we note that statistical significance does not refer to meanginfulness. Statistical significance rather refers to properties of the data and the tests used, so that the effect of something can be statistically significant but of trivial importance, or be highly relevant but not significant statistically. For instance, suppose that of a hundred people working overtime 20 will die during five years while among those who work normal hours only 10 die. Overwork would have a relative risk of 2 compared to 1 for the others. Now introduce a new toxic paint into the shop that kills 50 people in each situation. Then with overwork 70 die and without it, 60. The relative risk associated with overwork is now 70/60 or 1.17. Statistical tests may now not be significant, but the overwork is just as harmful.

All these worries are familiar to the epidemiologists but do not regularly inform their research or their reporting of it. The appropriate methods for statistically describing and analyzing whole systems are yet to be developed. In a class society, class affects people by many pathways. Individual "factors" may have even opposite effects in different situations, confusing the usual statistical treatments. If unemployment doesn't get you, working will.

Research in social epidemiology is directed toward finding recommendations of "policy". But policy is construed in a narrow sense as referring to measures that could be adopted by government, by the "decision makers". There are rules to advising on policy, and constraints that must be accepted so as not to raise eyebrows in high places. These act not only to limit the possible recommendations but also the scope of the analysis. These limits are political, but the political aspects of inequality are not discussed. The Indian state of Kerala is referred to because here, despite a low per capita income the health status of the population is higher than more prosperous states of India. The greater economic equality in Kerala is recognized as contributing to the health outcomes, but no mention is made of the fact that Kerala has a strong Marxist movement and that for much of its post-independence history this state has had left wing governments.

         The idea that social inequality by itself is harmful could have radical programmatic implications.. Harm caused by poverty might possibly be alleviated through quantitative measures such as raising the incomes of the poor, or by economic growth while keeping the same income distribution, but the harmfulness of inequality itself suggests more radical change. The harm caused by overwork and stress on the job might lead to a call for restructuring work and improving job security. Spatial segregation as a contributor to ill health might suggest rent control, more and better public housing, people-friendly zoning. And the whole pattern-- that inequality and hierarchy exacts a toll in years of life, disability and pain, could become a devastating critique of class society and by implication a strong case for socialism. The editors are troubled by this. Therefore we are reassured early, "We emphasize that the papers in this book are not dealing with the benefits of some unreachable egalitarian utopia ". What makes an "egalitarian utopia" unrealizable is not stated nor is the subject alluded to further.

This is an important book, likely to have a strong impact on the direction of public health epidemiology. Its strengths are those of the contributors. Its weaknesses are weaknesses of the field itself, the kinds of data that are available, the subjection of policy to the existing power structure of "decision makers", and to limitations of the accepted methodologies.




Top * Writings * Categories Enable This Widget » * Ecology (1) * Isador Nabi (1) * Marxism (1) # Search # # insert new section insert new section # This container does not contain any items. configure this page post new entry add page header add page footer Friday 23Oct2009 (DRAFT) DateFriday, October 23, 2009 at 10:42PM modify remove organize post follow up # # insert new section insert new section # This container does not contain any items. configure this page post new entry add page header add page footer Friday 23Oct2009 (DRAFT) DateFriday, October 23, 2009 at 10:42PM modify remove organize post follow up QUESTIONAIRE Every once in a while the invisible hand of a polling company reaches into the virtual hat-in-the-sky and draws my name at random to ask me questions about my opinions. Sometimes, in a weak moment, I answer them. Click to read more ... AuthorRichard Levins | CommentPost a Comment | Share ArticleShare Article tagged TagIsador Nabi in CategoryIsador Nabi Friday 23Oct2009 DateFriday, October 23, 2009 at 10:35PM modify remove organize post follow up FOR THE INVESTOR AHEAD OF THE MOB (from Genewatch, newletter of the Council for Responsible Genetics) ISADOR NABI, PUBLISHER LETTER FROM THE PUBLISHER Anybody can follow a trend, extrapolate from the past and call it a prediction. All you need is a straight edge and pencil and you can pass yourself off as a financial advisor, investment counselor or Wall Street pundit. But to really be ahead of the mob you have to have a profound understanding of the potential, unrealized, invisible and latent directions. One of the truly new worlds of opportunity has not yet hit the headlines but is silently bubbling up in exclusive law offices and scientific laboratories. Only Isador Nabi is at home in both worlds and is able to see the trends before they happen. This most exciting new possibility is the patenting of emotional property.

Click to read more ...