Friday, March 20, 2015

One Day Closer to Death

Disease Incidence Rates

One of the primary functions of the Centers for Disease Control is to tabulate the rates of occurrence of various diseases in the United States.  For example, here is a graph (from HHS, 2014) that shows annual cancer mortality rates for U.S. males:



If you are a public health official trying decide which diseases you should try to reduce the incidence of, this is pretty useful information.  But for an individual, there are better ways to present the same information.

Cumulative Lifetime Risk

A cumulative lifetime risk converts population incidence rates to a probability that a particular cancer will develop in an individual over a lifetime.  Cancer risk estimates from laboratory animal studies are typically expressed as lifetime risks, which is very simple to do since each animal is observed over their entire lifetime.  There are several methods for estimating lifetime risk from human statistical data.  The preferred method is a “life table” analysis that takes into account the current age distribution and other causes of mortality at different ages (Schouten et al, 1994).  This method can also be used to estimate the probability of a disease occurring at any given age, which is the sort of thing life insurance companies are interested in.  But for present purposes, there is also the easy way:


Cases per lifetime = Cases per year * years per lifetime

Life expectancies in the United States, calculated with life tables from 2010 (Arias, 2014), are 76.2 for males and 81.0 for females.  So using the lung cancer rate from the graph above, the cumulative life risk of dying from lung cancer in males is:

57.9 cases per 100000/year * 76.2 years per lifetime = 4412 cases per 100000 lifetimes

Stated as an individual probability: The chance of a male person dying from lung cancer in the United States is about 4.4%.

Life Expectancy Increments

A like expectancy estimate can also be used to represent the individual impact of a change in population disease incidence.  Life expectancy depends on when the disease occurs as well as frequency of occurrence; as disease that occurs in children will have a bigger impact than a disease that occurs primarily in the elderly (NCI, 2014).  Disease-specific estimate of reduced life-expectancy per case may be used to estimate increments in life expectancy than can be anticipated by the small theoretical risks that may occur from food additives and contaminants (the smallest change in disease incidence that even a very well designed epidemiology study can reliably detect is about one in a thousand).  For example, using the years-lost per case estimate of 15.7 for lung cancer, the average change is life expectancy associated with the usual range of risk estimates are as follows:

Change in Risk
Average Change in Life Expectancy
1 in a million (10-6)
8 minutes
1 in 100,000 (10-5)
1.4 hours
1 in 10,000 (10-4)
13.6 hours
1 in a thousand (10-3)
5.7 days

But that’s just the average.  Since no one is really average, what will happen to any one individual?  Here are two theories:

The One-in-a-Million Theory.  One person dies 15.7 year earlier, while everyone else is just fine.  According to one hit cancer theory, this is the more likely interpretation for genotoxic carcinogens.

The One Day Closer to Death Theory.  Everyone gets their lung cancer progression moved up bit, so everyone who gets lung cancer gets it a little earlier, plus one person dies of lung cancer instead of something else a day later.  Probably the better theory for nongenotoxic carcinogens, where the chemical or substance is more likely to be a risk factor than the sole cause of a disease.
Reality could also be somewhere in between these two theories.  In addition, there is undoubtedly some individual variability.    For example, even if the average is one day closer to death, there may be a range of 0.01 to 3 days.

There is plenty of empirical evidence that demonstrates that increasing the dose not only increases the population rate at which diseases occur, it also makes diseases occur earlier.  This best example is the greatest toxicology study ever; the ED01 study conducted by the USFDA National Center for Toxicological Research, which initiated a debate over how time of occurrence should be included in the risk assessment process.  The abstract of a paper written by a Society of Toxicology Task Force (1981) concludes:
The capability and need to do time based risk assessment exists now.
Yet time-to-tumor analyses are typically not used to extrapolate from high to low doses.  There really is no good reason for that. 

References

Arias, L (2014).  United States Life Table, 2010. National Vital Statistics Reports 65

National Cancer Institute (2014).  SEER Cancer Statistics Review.  1975-2011.

Schouten LJ,  Straatman H, Kiemeney LA, and Verbeek AL (1994).  Cancer incidence: life table risk versus cumulative risk.  J Epidemiol Community Health. 48(6): 596–600.

Society of Toxicology Task Force (1981).  Re-examination of the ED01 Study – Risk Assessment Using Time.  Fund Appl Toxicol 1:88-123.

U.S. Department of Health and Human Services (2014).  U.S. Cancer Statistics Working Group. Centers for Disease Control and Prevention and National Cancer Institute.  United States Cancer Statistics: 1999–2011 Incidence and Mortality Web-based Report.

Official Post Soundtrack

Lovich, Lene (1979). One in a Million.  In: Stateless, Track 10


Pink Floyd (1973).  Time.  In: Dark Side of the Moon, Track 4.


Post Note

Thesis Post No. 14.  Part of a dose-response modeling series.  The ED01 study deserves a whole post.  This a two soundtrack post:  One in a million is suitable for the first to sections, while Time goes with Life Expectancy

No comments:

Post a Comment