STATS ARTICLES 2010

2010 | 2009 |2008 | 2007 | 2006 | 2005 | 2004 | 2003

Is Childhood Cancer Becoming More Common?
Trevor Butterworth, May 28, 2010
According to New York Times columnist Nick Kristof it is, and he claims that the Presidentís Cancer Panel backs him up. But what do the numbers really say?

In a recent New York Times column, Nicholas Kristof warned readers that “some cancers are becoming more common, particularly in children. We don’t know why that is,” he continued, “but the proliferation of chemicals in water, foods, air and household products is widely suspected as a factor.” Kristof drew his support for this statement from the President’s Cancer Panel, which he called “the Mount Everest of the medical mainstream.”

Because Kristof appears to have either gotten an advance copy of the report, his glowing praise for the two-member panel and the conclusions of their hearings was dramatically out of sync with the backlash from the American Cancer Society and other experts about the report’s failings, which were widely reported in the media (See STATS’ investigation of the report here).

The minutes of the Panel’s meeting on chemicals reveal that the driving force for the data on childhood cancer was Dr. Philip J. Landrigan, professor of pediatrics at the Mount Sinai School of Medicine in New York and chairman of the school’s department of preventive medicine.

Landrigan’s statement to the Panel hearings is summarized thus:

“While mortality from childhood cancer has gone sharply down, incidence rates are increasing. There has been a 55% increase from 1975 to 2005 in the incidence of leukemia in 0 to 14-year-olds and an 81% increase for acute lymphocytic leukemia—the most common type of leukemia. …The explanation for this increase may be due in part to better diagnostics, but this alone does not account for the continued inexorable rise. Serious consideration must be given to the possibility that environmental factors are involved.”

The final Panel report notes:

“The causes of this increase are not known, but as a meeting presenter emphasized, the changes have been too rapid to be of genetic origin. Nor can these increases be explained by the advent of better diagnostic techniques such as computed tomography (CT) and magnetic resonance imaging (MRI). Increased incidence due to better diagnosis might be expected to cause a one-time spike in rates, but not the steady increases that have occurred in these cancers over a 30-year span. The extent to which environmental exposures are responsible for this trend remains to be determined.”

The panel also portrays the increase by extending the usual definition of childhood as age 0 to 14 in discussions of cancer incidence to include those aged 15-20, by including the graph below depicting an increasing trend. As we note below, including the older group has the effect of increasing the overall incidence rates.

 

SEER Delay-Adjusted Incidence and U.S. Mortality
All Childhood Cancers, Under 20 Years of Age
Both Sexes, All Races, 1975-2006

cancer graph

But do the actual childhood cancer statistics reveal “a continued inexorable rise” that cannot be explained by earlier diagnosis or changes in lifestyle?

One problem with presenting percentages shorn of the numbers that generated them is that, when the number of cases is low,  a small increase in actual numbers can yield a large percentage increase. For example, an increase from 1 to 2 cases represents a 100 percent increase, while an increase from 11 to 12 cases represents only a nine percent increase. Indeed, USA Today once ran a misleading story on allegedly soaring methamphetamine use among workers based on just such a a small increase in actual numbers.  If there are only small numbers of cases of childhood cancer in the population overall, then you will see greater fluctuations from year to year than if there were large numbers. This a common and confusing statistical point: the larger the numbers, the less variation, and the smaller the numbers, the larger the variation one should expect from statistical “noise.” Therefore, dramatic percentage increases will seem loaded with ominous meaning, just as percentage decreases will suggest that a problem is disappearing.

A second problem is that 15 to 19 year olds are substantially more likely to be diagnosed with cancer than those aged 5 to 14. The PCP’s decision to plot increasing incidence by extending childhood to 20 may give the appearance of a more sharply escalating trend. This has added relevance given that that the types of cancer that occur among 15 to 19 year olds are different from those that occur in children.

Finally, Martha Linet MD, of the National Cancer Institute’s of the National Cancer Institute's Radiation Epidemiology Branch, has warned of treating all childhood cancers equally. As she noted at a 1998 meeting of the Environmental Protection Agency to discuss the causes of childhood cancer, "Many epidemiologists believe that total childhood cancer is not an etiologically meaningful entity.  "Each of the 12 major internationally recognized categories of childhood cancer has its own very clear descriptive epidemiology; they differ by age, by gender, by race, and many other factors. There's not a totally consistent picture..."

So let’s look at the National Cancer Institute’s childhood cancer data for overall incidence and for the two most common childhood cancers, those of the brain and central nervous system and the most common kind of leukemia, acute lymphocytic.

 

AGE-ADJUSTED SEER CANCER INCIDENCE RATES, 1975-2004
By Primary Cancer Site
All Races, Males and Females
Year of Diagnosis
Ages 0-14

Age/Site 1975 1980 1985 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004
All Sites 11.5 12.9 14.5 14.2 15.1 13.4 14.9 13.9 14.0 14.7 14.0 15.3 14.4 15.4 15.6 15.2 12.9 14.8
Brain & Other Nervous 2.3 2.8 3.0 3.5 3.5 3.2 3.4 3.3 3.3 3.1 2.7 3.2 3.4 3.4 3.8 3.6 2.9 3.2
Leukemia 3.3 4.0 4.6 4.4 4.6 4.0 4.2 3.7 4.4 4.6 4.2 4.8 4.7 4.6 4.4 4.8 4.0 4.9
Acute lymphocytic 2.2 3.1 3.5 3.4 3.5 3.1 3.1 2.9 3.6 3.5 3.5 3.9 3.6 3.5 3.5 4.0 3.1 3.6

SEER 9 areas (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, and Atlanta). Rates are per 100,000 and are age-adjusted to the 2000 US Std Population

 

Between 1975 and 2004, there is an increase of 28.6 percent in overall cancer incidence from 11.5 cases per 100,000 children aged 14 and under to 14.8 cases per 100,000. But if you drop the data from 1975, a particularly low year, the total cancer increase from 1980 to 2004 was a 15.5 percent. Similarly, by dropping the last data point in 2004, the increase from 1975 to 2003 is 12 percent. The numbers are more extreme if one does some cherry picking: for example, the cancer rates in 1980 and in 2003 are the same, and cancer rates measured from 1993 (a year with a high rate of incidence) went down 13.4 percent by 2003.

As the NCI noted, there is a statistically significant annual percentage change in all cancers from 1975 to 2004 (0.6 percent annually), but the annual percentage change for 1990-2004 is significantly different from the annual change for 1975-1989 (0.2 percent). So while there is evidence of an upward trend, the increases seem less substantial than suggested by the Panel.

If we look at cancers of the brain and central nervous system, and acute lymphocytic leukemia, the two major causes of childhood cancer, we see, again, a leap between 1975 and the early 1980s, and then a pattern of overall stability within a range of fluctuating rates. For instance, the incidence rate for acute lymphocytic leukemia was identical in 1980 and 2003.

In other words, it is not immediately clear whether the increases reported are a matter of small fluctuations in data and an unfortunate choice of endpoints, or whether there is a “real” and large increase.  Perhaps after the major increase in the early 1980s, we are only measuring the kind of statistical fluctuation that you would expect with such small overall numbers of cases.

Here’s the NCI’s breakdown of overall and annual percentage changes:

AGE-ADJUSTED SEER CANCER INCIDENCE TRENDS, 1975-2004, 1975-1989, 1990-2004a
By Primary Cancer Site
All Races, Males and Females 0-14

% change
APC 1975 – 2004
APC 1975-1989
APC 1990 - 2004
All sites   
28.6
0.6*
1.2*
0.2#
Brain etc 
39.9
1.0*        
2.1*        
0.0#
Leukemia
49.6                  
0.7*        
1.5*
0.6
Acute Lymphocytic
63.5
0.8*        
2.2*            
0.7#

The APC is the Annual Percent Change over the time interval.
SEER 9 areas (San Francisco, Connecticut, Detroit, Hawaii, Iowa, New Mexico, Seattle, Utah, and Atlanta).
* The APC is significantly different from zero (p<.05)
# The APC for 1990-2004 is significantly different from the APC for 1975-1989 (p<.05).

 

As can be seen in the table, the annual percentage change for brain and central nervous system cancers was 2.1 percent between 1975 and 1989, and then fell to 0 percent between 1990 and 2004. More recent NCI data from 2003 to 2007 show that the incidence of brain cancer is in decline.

While the President’s Cancer Panel dismisses improvements in cancer diagnosis – such as the introduction and widespread use of MRI technology – as even partial explanations for the rise cancer incidence in the early 1980s, the National Cancer Institute believes this is a likely explanation. As it notes,

“For childhood brain tumors, the overall incidence rose from 1975 through 2004, from 2.3 to 3.2 cases per 100,000 (2), with the greatest increase occurring from 1983 through l986. An article in the September 2, 1998, issue of the Journal of the National Cancer Institute suggests that the rise in incidence from 1983 through 1986 may not have represented a true increase in the number of cases, but may have reflected new forms of imaging equipment (magnetic resonance imaging or MRI) that enabled visualization of brain tumors that could not be easily visualized with older equipment (3). Other important developments during this time period included the changing classification of brain tumors, which resulted in tumors previously designated as “benign” being reclassified as “malignant,” and improvements in neurosurgical techniques for biopsying brain tumors. Regardless of the explanation for the increase in incidence that occurred from 1983 to 1986, childhood brain tumor incidence has been essentially stable since the mid-1980s.”

This view was echoed in the study Cancer Surveillance Series: Recent Trends in Childhood Cancer Incidence and Mortality in the United States, which noted,

“Our analysis found no large increases or decreases in incidence from 1975 through 1995 for major categories of pediatric cancers in the United States. The slight increase in childhood brain tumors from 1983 through 1986 is consistent with enhancements in diagnostic techniques and changes in classification. Reasons are unknown for the short-term increase in leukemia rates observed from 1983 to 1984…

“… it can be misleading to estimate an overall incidence rate change by the use of differences between the most recent and the earliest years, without considering variations in the entire interval. Other factors may influence the observed patterns, and they include the advent of new diagnostic techniques, such as magnetic resonance imaging (9,23), or the expanding use of an existing diagnostic aid, such as prenatal ultrasound testing (24), the specificity of histopathologic designation or other changes in diagnostic criteria over time (25), and the introduction of a new classification system (11,14-20)”

Certainly, the existence of environmental causes for cancer cannot be dismissed. In October, the American Cancer Society’s Cancer and the Environment committee estimated that pollutants may cause up to 5% of all cancers., But the President’s Cancer Panel’s unwillingness to accept changes in diagnostic techniques and classification, and the historical introduction of new imaging technologies, as explanatory causes for even some changes in cancer incidence in children, while at the same time simplifying the incidence data, seems disingenuous. Another possible culprit for the observed increases is diet. American children are getting more and more obese, which likely plays an important role in their health; even the type of food we eat is different now than it was 30 year ago. These demographic shifts were not commented upon by the Panel.

To be sure, the President’s Cancer Panel pointed to many possible environmental concerns for children, and even the small percentage of cancers attributed to these environmental factors is too many. But we have to be careful about how we make conclusions about specific toxins. While Kristof’s New York Times column focused on choosing toys without BPA and buying organic foods, the actual Panel report notes a wide range of known environmental causes for cancer, many of which have more convincing proof that they are carcinogens and have large impact on cancer incidence. These include radon home exposure, living in polluted cities, or living on a farm that uses pesticides, tobacco use and exposure, and sun exposure. Kristof does mention that one should “check radon levels” but he puts this on equal footing with storing water in plastics that do not contain BPA. Radon is a confirmed carcinogen with known impact (noted in the Panel report), while BPA continues to be a contentious issue, largely because the science is not wholly conclusive. The Panel lists many other sources of potential,suspected, or known carcinogens that were also dismissed by the media. These include cell phone usage, medical radiation exposure, contaminated drinking water, and air travel.

With so many suspected contributors to the cancer incidence – and so many potential additional factors such as our dietary habits and choices, we have to rely on better ways of evaluating our risks, such as controlled and observational studies, rather than conjectural and unquantified statements of risk. Fingering BPA as a culprit may allay some fears, as throwing out our Tupperware might be easier than losing weight and refusing to travel by air. But rest assured, cancer rates are not skyrocketing (and skyrocketing is not causing huge jumps in cancer rates). While we cannot know everything about what causes cancer, reducing exposure to the largest known carcinogens (for most people: tobacco, radon, and sunlight) will bring the biggest benefit.

 

 


Digg!

Technorati icon View the Technorati Link Cosmos for this entry


Share