The Los Angeles Times reports that cancer risks in Southern California from air toxics declined in 2006 by 17%. Any decline in cancer risk is good news. How good is it?
Here’s what Times’ staff writer Janet Wilson says:
Cancer risk from Southern California’s air pollution has declined 17% over the last seven years but remains dangerously high across the region, particularly near ports and rail yards, along truck-laden freeways and in parts of the Inland Empire, according to a study released by regional air regulators Friday.
17% OF WHAT?
The draft Multiple Air Toxics Exposure Study (MATES III) is a product of the South Coast Air Quality Management District. SCAQMD, the regional air pollution regulatory agency, is taking public comment on the draft report through April 4, 2008.
According to the draft report, the “expected” number of additional cancer cases attributable to air toxics in Southern California is 1,200 cases per million persons exposed over 70 years (Executive Summary, PDF pp. 2-3). Because the models used to estimate cancer risk assume that it is linear with lifetime dose, this is equivalent to an incidence of (1,200 / 70 =) 17 cases per year per million persons exposed, or 1.7 cases per 100,000 (a more commonly used denominator).
The 17% reduction is a modeled difference from the previous analytic effort (MATES II) (PDF p. 6):
Compared to previous studies of air toxics in the Basin, this study found a decreasing risk for air toxics exposure, with the population weighted risk down by 17% from the analysis in MATES II.
ARE POPULATION CANCER RISKS FROM AIR TOXICS IN LOS ANGELES “LARGE” OR SMALL”?
If cancer incidence from air toxics were to have actually declined by 17%, would anyone notice? The answer is “No.” The total cancer incidence predicted by SCAQMD to result from air toxics exposure is too small to be noticed.
The federal Centers for Disease Control estimates that the age-adjusted all-sites invasive cancer incidence in the Los Angeles basin in 2004 was 502.4 per 100,000 (95% confidence intervals: 495.1 — 509.9 per 100,000). Incidence was higher for blacks (619.3 per 100,000; 95% confidence intervals: 592.7 — 646.7 per 100,000) and lower for Hispanics (402.2 per 100,000; 95% confidence intervals: 388.5 — 416.1).
If the modeled estimates in MATES III are correct, then air toxics are responsible for (1.7 / 502.4 =) 0.34% of CDC’s estimate of total cancer incidence. It is 12% of the estimation error within the 95% confidence intervals. If these cancer risks vanished, the change in cancer incidence would not be detected. A 17% reduction in a phenomenon too small to observe also is not observable.
ARE POPULATION CANCER RISKS FROM AIR TOXICS “UNACCEPTABLE”?
This is a policy question that cannot be answered by an appeal to science or economics. Nevertheless, both science and economics can provide useful insights to inform policy judgment.
The draft MATES III report is supposed to be a strictly scientific exercise, but SCAQMD has difficulty making a clear distinction science and policy. Speaking about its estimated 17% reduction, the SCAQMD report says:
While there has been improvement in air quality regarding air toxics, the risks are still unacceptable and are higher near sources of emissions such as ports and transportation corridors. Diesel particulate continues to dominate the risk from air toxics, and the portion of air toxic risk attributable to diesel exhaust is increased compared to the MATES II Study (Executive Summary PDF p. 6).
The SCAQMD does nor say in the MATES III report how much population cancer risk is acceptable. Clearly, it is taking the position that reducing annual cancer incidence from 2.0 cases per 100,000 to 1.7 per 100,000 (a 17% decrease) is not enough.
Is it? According to the economics literature, people are willing to pay between $1 million and $10 million to prevent one randomly occurring premature death. (The federal Environmental Protection Agency uses a figure of approximately $5 million, and assumes that all invasive cancer is fatal.) Even if the upper bound of this range is assumed to be “correct,” then it would be cost-effective to achieve the SCAQMD’s estimated 17% reduction in cancer incidence (0.3 case per year) if it could be accomplished for less than ($10 million x 0.3 =) $3,000,000 per 100,000 population.
The SCAQMD regulates the airshed in which the vast majority of the population of Los Angeles, Orange, Riverside, and San Bernardino Counties reside, a population of approximately 16.5 million persons. A reduction in cancer incidence of 0.3 per 100,000 is equivalent to about 50 cases per year in this population. All-site invasive cancer incidence in this population is about (502.4 per 105 x 165 x 105 population =) 83,000 cases per year (95% confidence interval: 82,000 to 84,000). The value of preventing 50 premature deaths per year (not cancer cases) ranges from $50 million to $500 million.
The draft MATES III report implies that it doesn’t matter whether it costs $50 million, $500 million, or more to reduce reduce cancer incidence by 50 cases in the airshed, or for that matter, what it would cost to eliminate all remaining 1.7 cases per 100,000 (about 280 out of 83,000 in the airshed). This policy almost certainly conflicts with the values of Angelenos, who probably prefer policies that reduce cancer incidence at least cost. That way, the maximum number of cancer cases would be prevented for any level of expenditure.
ARE THE SCAQMD’s ESTIMATED CANCER RISKS VALID?
No, they aren’t. The MATES III approach estimates cancer risk using a method virtually guaranteed to overstate actual cancer risk, and thus overstate cancer incidence. Substantial effort was devoted to obtain unbiased estimates of exposure, but these values then were multiplied by upper-bound cancer potency values. When an unbiased estimate is multiplied by a biased estimate, the product must be biased.
The draft report (p. ES-3) ignores this mathematical fact and claims that the product yields an “expected” value:
The carcinogenic risk from air toxics in the Basin based on the average concentrations at the fixed monitoring sites is about 1,200 per million. This risk refers to the expected number of
additional cancers in a population of one million individuals that is exposed over a 70 year lifetime.
This is false. The SCAQMD’s cancer risk estimate cannot be an “expected value” because the cancer potency factors used are not “expected value” potency estimates. Rather, they are — and are intended to be — highly precautionary, near worst-case values that almost certainly overstate the expected value, and probably by a lot. SCAQMD obtained the cancer potency values from the California Office of Health Hazard Assessment (OEHHA) (see Appendix IV PDF), which does not derive expected value estimates. OEHHA’s policy and practice is to derive estimates that are virtually certain not to underestimate cancer potency. That means OEHHA’s estimates are also virtually certain to overestimate cancer potency. The likelihood is very small that any one of these values is, by accident or error, unbiased and thus an “expected value” estimate. The likelihood that accident or error led OEHHA to derive unbiased estimates for all of them is, for all practical purposes, zero.
In a booklet published for nonspecialists, OEHHA acknowledges that it makes “health protective” assumptions that tend to overstate expected value cancer risks:
For chemicals that cause cancer, the general assumption in risk assessment has been that there are no exposures that have “zero risk” unless there is clear evidence otherwise (p. 8).
An individual’s actual risk of contracting cancer from exposure to a chemical is often less than the theoretical risk to the entire population calculated in the risk assessment (p. 10).
“Often less than the theoretical risk” is an extreme understatement; it would be much more accurate to say “almost never as high as the theoretical risk.” In any case, OEHHA does not reveal, either in this booklet for lay people or in the risk assessments it prepares for specific chemicals, the magnitude of bias that results from using these “health protective” assumptions. Just one of these health-protective assumptions — the assumption that “zero risk” is infeasible — results in a bias that is infinitely great for any substance in which a biological threshold exists. (Any positive number divided by zero yields infinity.)
Thus, on this margin alone the draft MATES III report is highly misleading. The expected number of annual cancer cases attributable to air toxics is much lower than SCAQMD’s implied figure of 1.7 cancer cases per year. An unbiased estimate might be closer to 0.17, 0.017, or 0.0017 case per year. Or less.
What the draft MATES III report shows is that even if the cancer risk from air toxics is exaggerated, the resulting cancer incidence is very small. If reducing cancer incidence is the objective, regulating air toxics is a very expensive way to accomplish next to nothing.
THE EFFECT OF BIAS IN CANCER POTENCY FACTORS ON
CANCER INCIDENCE AND VALUE OF REGULATORY CONTROLS
|Amount of Bias in Cancer Potency Values Used by SCAQMD||Unknown but True
Cancer Incidence Reduction in SCAQMD Airshed
|Annual Value of
Air Toxics Regulation by SCAQMD
Aimed at Reducing Cancer Risks*
|None**||0.3 case/year x 165 x 105||$50 to $500 million|
|10x||0.03 case/year x 165 x 105||$5 to $50 million|
|100x||0.003 case/year x 165 x 105||$500,000 to $5 million|
|1,000x||0.0003 case/year x 165 x 105||$50,000 to $500,000|
|* Assumes range of $1 million to $10 million for preventing one random premature death, and that all invasive cancer cases are fatal.
** Assumption implied in draft MATES III report. 0.3 case per year is the 17% reduction claimed in the draft report.
Note: Updated January 8, 2007 1800 to correct editing errors in the original.