By Roy Richard Grinker
Special to The Washington Post
Tuesday, October 30, 2007

When my wife was in labor in 1991, the doctor attached an electronic fetal monitor to her belly to record her contractions and the baby's heart rate. I remember being so transfixed by the monitor that I forgot about my wife. At one point, she said, "Wow, it hurts." But because the graph on the monitor had yet to rise above the level signaling a contraction, I said, "No, it doesn't."

Even the best data can be subject to misinterpretation.

Americans are confronted every day with graphs, trends, averages and percentages, and few of us have the expertise to grasp them.

"There is terror in numbers," wrote Darrell Huff in his 1954 book "How to Lie With Statistics." If, for example, a small increase in the incidence of a disease over a five-year period is projected out 20 or 30 years, with the same rate of increase, he notes, the disease suddenly looks like an "epidemic," a word scientists use to describe a sudden outbreak of a disease in a discrete population.

Indeed, our child, who was born on the day I incurred my wife's wrath, became part of the so-called autism epidemic.

The autism rate (the number of cases divided by the total population) has gone from fewer than four in 10,000 in 1990 to more than 66 in 10,000 in 2007. Those statistics have led to widespread concern and social traction: a proliferation of advocacy organizations, scores of new, unproven therapies and thousands of lawsuits filed by parents who believe the government's vaccine program is responsible for the higher rates. Between 1997 and 2007, when government funding for most diseases was unchanged, autism funding at the National Institute of Mental Health increased from $22 million to $108 million.

But numbers and rates rise for many reasons. There may actually be more of the disease -- or there may be greater awareness, better detection, improved record-keeping and/or a redefinition of terms.

In 1979, Stuart Gunn, a chemist at Lawrence Livermore National Laboratory in California, died of melanoma, the deadliest form of skin cancer. A year later, a state study found that the rate of melanoma for Livermore employees was five times higher than for residents of surrounding communities. Fear of an epidemic raced through the lab. It did so despite the fact that melanoma had never been linked to any occupational exposure.

The increase was an illusion. In response to Gunn's death, scores of employees went to doctors to have their skin examined, and a few had thin, curable lesions, probably in proportions no higher than would have been found among non-employees -- if they, too, had been seen by a dermatologist. More doctor visits, more biopsies, more cases detected, and soon you had what looked like an epidemic.

Dermatologists report a nearly 300 percent increase in the incidence of melanoma in the last 20 years. Some are convinced there has been a steady increase that began in the 1940s when it became fashionable to have a tan.

But others, such as H. Gilbert Welch, a professor of medicine at Dartmouth Medical School and co-director of the Veterans Affairs Outcomes Group, argue that melanoma is diagnosed earlier because more patients are aware of the sun's dangers, see doctors on a regular basis and get biopsies, and because malpractice insurance companies encourage doctors to do more biopsies. What's more, Welch says, "pathologists are more likely to read a given biopsy as melanoma."

Many scientists believe that what happened in Livermore with skin cancer is happening with other illnesses.

Between 1987 and 1992, the incidence of prostate cancer jumped 85 percent. Why? Because something dramatic happened in 1987: Doctors started to perform a simple blood test to screen for prostate specific antigen, or PSA, levels, a marker for prostate cancer.

Almost overnight, more early-stage cases were detected and treated. What looked like an epidemic of prostate cancer was, in fact, major progress. Before that time, cases were detected only if a man received a transurethral resection of the prostate, or TURP; the procedure involved inserting a small telescope into the prostate through the penis and chipping away pieces of the prostate for analysis. Not surprisingly, doctors performed the procedure only when truly necessary.

Then, between 1992 and 1996, rates of prostate cancer dropped almost as steeply as they had risen.

"This didn't mean there was necessarily more disease during the rise or less disease during the decline," says Thomas M. Pisansky, a professor of oncology at the Mayo Clinic College of Medicine. "Most researchers agree that the rise was due to the PSA, and the fall was due to getting all those men diagnosed."

It's as if a fisherman suddenly found a better way to catch fish. At first, he'd have a high yield, but over time there would be fewer fish to catch.

Lowering the threshold for diagnosis can quickly change the prevalence of a disease. Take hypertension (high blood pressure), a condition that affects more than 50 million Americans. Hypertension awareness campaigns since the 1960s have lowered mortality from coronary heart disease and stroke. But the prevalence of hypertension has risen over the past 10 years.

Fast food and a more sedentary lifestyle likely have played a part, but so has a redefinition of the disease.

In 2003, an expert committee concluded that individuals with a diastolic pressure (the bottom number in a blood pressure reading) of 80 to 89 (then on the high end of "normal") were at risk of developing hypertension-related disease and should be called "prehypertensive."

Physicians soon began to treat such patients for hypertension, and, for insurance and medical records, they were coded in the same way as someone with a much higher blood pressure. The number of diagnosed cases of hypertension thus rose.

The average patient with what was newly considered "high blood pressure" also fared better because the pool of patients with hypertension now included some previously "normal" people. This might also help explain part of the drop in mortality among people with hypertension: The average patient with the diagnosis was now healthier.

Thresholds have dropped for other common diseases, including obesity and diabetes, and the criteria for disease classifications have broadened. Autism, for example, once a narrowly defined disorder, is now used to describe a wide spectrum of severity, from the profoundly mentally retarded person to the socially awkward mathematics professor.

The methods scientists use to count also affect measures of incidence.

If, for example, you count the number of cases of a disease by examining insurance records, you miss those who do not have insurance (more than 40 million Americans). If you count cases through health-care providers, you miss those who did not seek treatment. If those people are minorities, immigrants and others for whom there are disparities in access to care, the prevalence of the disease will appear lower in those populations.

One reason for the higher rates of many diseases is that researchers are being more thorough in their methods, and many of the records they analyze are computerized and better organized.

For example, the Centers for Disease Control and Prevention recently searched medical and educational records and found that the proportion of children with autism was more than three times higher in New Jersey than in Alabama. The most likely explanation for the disparity is that Alabama lags behind New Jersey in providing medical and educational services for autism. Without services, many autistic people in Alabama could not be counted because there was no sign of them in the records the CDC analyzed.

When the numbers were released in early 2007, New Jersey newspapers were filled with articles about the "epidemic," but the statistics could just as well be interpreted as confirmation of how much New Jersey is doing for autistic children.

Despite all the tragedies we read about in the news, our world is actually safer than it has ever been. Yet we live in dread of epidemics and anxiously await the release of the latest figures from the country's health-care leaders.

Ironically, many of our fears are the result of the knowledge generated by the many real advances in medicine. So the next time you see statistics documenting the increase of a disease, take at least a moment to consider whether they may be evidence not of harm, but of good. ¿

Roy Richard Grinker is professor of anthropology and the human sciences at George Washington University.