by Peter N. Dunn
Are wafer fabs a safe working environment? The Semiconductor Industry Association (SIA) has formed an independent scientific panel of experts (including specialists in epidemiology, industrial hygiene, toxicology and occupational health), which will conduct a review of existing data on potential cancer risks in the chipmaking industry.
An SIA spokesman said the panel, which will be “thoroughly objective, independent and authoritative,” is still being formed. The review's time frame will be established by the panel, but the spokesman said a period of 12 to 16 months is envisioned. The SIA's Environmental, Safety, and Health committee, with representatives from each member company, will work with the independent panel and receive information from it. No specific budget has been set for the study; the spokesman said the panel's chairperson would have some say in how much will be spent.
Whenever personal health and safety are involved, there's every reason to move quickly on any potential problems. The SIA deserves credit for organizing a conference last year on semiconductor industry worker health issues, and for agreeing to assemble a data review panel. But it's worth noting the comments of Patricia Buffler, professor of epidemiology at the University of California at Berkeley, who spoke at the conference and recommended an aggressive toxicology program to study cancer rates among chip workers. She suggested that chip companies do a surveillance of employee cancer rates and compare those incidence rates to California and US cancer registries. While noting that there is no evidence today of problems with chip worker cancer rates, she said, “we don't know that there isn't. To do it well…is going to take quite an effort. We're talking about hundreds of thousands of employees. It would take several years.”
She also noted, “It's good business. You're defenseless without it.” This would require gathering of new data, not just a review of what already exists. To get more perspective on the situation, I did what I've been doing for several decades now: I asked my big sister, Julie. Not just because she's older and therefore must know better (she has explained this to me on several occasions), but because she is an epidemiologist at a major American university. Julie said studies of occupational exposure to potential disease-causing substances are “a methodological hell; there are a whole lot of difficulties.” The groups in question tend to be relatively small, and there can be data-gathering problems due to uncooperative attitudes among both companies and the workers being studied. Finding a suitable control group is very difficult, in part because of the “healthy worker effect:” people in the workforce tend to differ from non-workers in many ways, including overall health. And without a good control, you can't draw firm conclusions.
Moreover, industrial situations often involve “unknown confounders” factors other than the ones under study, which can have profound effects on health. These can include worker demographics, interactions among chemicals, and a host of other issues. Sometimes a problem is blatant, as when a cluster of cases of liver cancer, a rare disease in the US, was found among factory workers exposed to vinyl chloride. But in most cases, the process of conclusively demonstrating an association between workplace chemicals and disease is, in Julie's words, “one of the more challenging branches of epidemiology.”
All of which provides an excellent rationale for moving the SIA data review along as quickly as possible, even if it costs more. The Association and its member companies, and their counterparts in other regions, should also ensure quick funding for additional data collection, so that any needed further study can begin without delay. After all, this is an industry that uses the scientific method to completely reinvent its products every couple of years it should be able to move at least as quickly when lives may be at stake.
Peter N. Dunn is founder of and contributing editor to WaferNews. This article is reprinted from WaferNews V 6.48 (December 6, 1999).