[Stoves] Off-topic: Fine particulate matter and mortality: A critical review (Tony Cox Jr.)

[Nikhil Desai]

This is an expanded review of the review.

There is no reason to confuse attributability with causality; attributions
depend on whims of the glib tailored to the gullible.

Nor is there any justification for mistaking attributability to
avoidability. (Read  Kirk Smith.)

In short, there is no there there. WHO has bet its credibility on an
ideology of "no solid fuels". The trouble is not fuels but combustion and
ventilation technologies.

*Do causal concentration–response functions exist? A critical review
of associational and causal relations between fine particulate matter and
mortality,* Louis Anthony (Tony) Cox Jr CRITICAL REVIEWS IN TOXICOLOGY,
2017 http://dx.doi.org/10.1080/10408444.2017.1311838

>From the Abstract:

The assumption that C–R functions relating levels of exposure and levels of
response estimated from historical data usefully predict how future changes
in concentrations would change risks has seldom been carefully tested. This
paper critically reviews literature on C–R functions for fine particulate
matter (PM2.5) and mortality risks.* We find that most of them describe
historical associations rather than valid causal models for predicting
effects of interventions that change concentrations. The few papers that
explicitly attempt to model causality rely on unverified modeling
assumptions, casting doubt on their predictions about effects of
interventions. ..the assumption that estimated C–R relations predict
effects of pollution-reducing interventions may not be true.* Better causal
modeling methods are needed to better predict how reducing air pollution
would affect public health.

C = Concentration and R = Response

>From Introduction:

A concentration–response (C–R) curve shows levels of adverse health
responses in exposed populations on its vertical axis and levels of ambient
concentrations of a pollutant on its horizontal axis. Such curves, which
are usually upward-sloping, have been widely used to predict the public
health impacts of proposed reductions in air pollutants (Schwartz et al.
2002; Pope et al. 2015). These predictions are made by assuming that
reducing air pollution will reduce adverse
health effects, moving both exposure and response variables leftward and
downward along the C–R curve. Thus, a C–R curve is commonly given both of
the following two interpretations:

Associational interpretation .. (based on historical data, typically via a
regression model)

Causal interpretation.. (response predictions based on concentrations)

As illustrated in detail in Table 1 later, these two interpretations are
usually conflated, leading to causal interpretations of associations. For
example, causal claims such as that “The magnitude of the* association
*suggests
that controlling fine particle pollution *would result* in thousands of
fewer early deaths per year” (Schwartz et al. 2002, emphases added) are
very common in the epidemiological literature on air pollution health
effects, and in regulatory risk assessments based on this literature, even
though, as discussed next, there is no necessary relation between the
magnitude or direction of a historical C–R association and the effects on R
of reducing C.

In principle and in practice, there may be no single curve satisfying both
interpretations. ...

*More generally, finding a positive C–R association in historical data does
not necessarily imply anything about how changing C would change R. *It is
not generally true that if a C–R model describes past data values for C and
R, then it also predicts how changing exposure concentration from a current
level to a new level will change the average response.

..


*Many important past papers equate associational and causal C–R
relations *Current
risk assessments, benefits assessments, and recommendations for revising
standards for criteria pollutants build on a *decades-old scientific
literature *that routinely makes the following three *implicit assumptions*:

a. Associational and causal C–R functions can be modeled by the same curve
(Pope et al. 2002);
b. This C–R curve can be estimated quantitatively from relevant data via
regression models or other associational methods such as odds ratios,
relative risks, attributable risks, and burden of disease estimates,
perhaps augmented
with human judgments based on the Sir Austin Bradford Hill considerations
(Hill 1965) or other weight-of-evidence considerations (Fedak et al. 2015)
such as the strength, consistency, temporality, and biological plausibility
of associations.
c. C–R regression coefficients or other associational measures estimated
from one set of locations and times can be applied to exposure
concentrations for other locations and times to estimate excess mortalities
caused by air pollution and the potential human health benefits that would
be caused by reducing it (e.g. Chen et al. 2013).

Table 1 gives examples of statements from articles that make one or more of
these assumptions


[image: Inline image 1]

I
-----

*Summary and conclusions*

Our critical review of the literature on concentration-response (C–R)
relations for fine particulate matter and mortality has identified the
following challenges:

   - C–R functions that describe historical associations do not necessarily
   predict how changing C would change R. This is partly because associations
   may not represent manipulative causal relationships, as when positive
   associations between baby aspirin consumption and heart attack risk,or
   between nicotine-stained fingers and subsequent risk of lung cancer, do not
   allow a valid prediction that reducing one would reduce the other.
   - Almost all of the existing literature on PM2.5-mortality C–R functions
   deals with associations and not with causality (Wang et al. 2016).
   - The few papers that do attempt to model causality in C–R relations for
   PM2.5 exposure and mortality fail to distinguish among counterfactual,
   predictive, and manipulative causality. Most of these papers follow a
   counterfactual approach that relies heavily on unverified modeling
   assumptions about unobserved potential outcomes. *This amounts to
   assuming, rather than showing, that associations in regression models can
   be interpreted causally. Arguably, manipulative causation should be the
   gold standard for discussions of causality and the public health effects
   that would be caused by changing PM2.5 exposures. Although even
   state-of-the-art causal inference algorithms cannot definitively establish
   manipulative causality from observational data, they can identify several
   measures of predictive causality (Granger causality, information relations
   in DAG models, and so forth) that provide a valuable screen for evidence of
   potential manipulative causation.*


.. It seems highly desirable for future work to distinguish more clearly
among statistical association, predictive causal, and manipulative causal
C–R functions than past research has done. *The fundamental premise that
C–R functions exist that can predict the public health effects caused by
reductions in pollutant concentrations needs to be carefully reexamined and
tested, as it does not appear to hold in general*.

>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.bioenergylists.org/pipermail/stoves_lists.bioenergylists.org/attachments/20171026/3d2ebea1/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: image.png
Type: image/png
Size: 272674 bytes
Desc: not available
URL: <http://lists.bioenergylists.org/pipermail/stoves_lists.bioenergylists.org/attachments/20171026/3d2ebea1/attachment.png>