By Olga Kharif The Semiconductor Industry Assn. (SIA) recently took a step many chip-industry workers had awaited for years. The group, which represents 85% of U.S. chip concerns employing a total of 225,000 people, announced it would fund the first-ever industry-wide study examining whether semiconductor plant workers face a higher risk of getting cancer. The decision came more than a year after a feasibility study that showed that such a survey could be done.
To be conducted by Vanderbilt-Ingram Cancer Center at Vanderbilt University by the spring of 2009, the epidemiological study comes after years of academic spats among researchers and a barrage of employee lawsuits -- some settled, others thrown out, some still pending -- against companies such as IBM (IBM) and their chemicals suppliers. The suits allege that working in certain areas of chip plants, or fabs -- semiconductor foundries -- poses higher risks of getting cancer.
"VERY DIFFICULT." Out of numerous proposals, the SIA picked Vanderbilt-Ingram oncologist Joseph McLaughlin, famed for his studies of long-term effects of medical implants. McLaughlin didn't respond to repeated requests seeking comment.
But don't expect this study, expected to cost at least $5 million, to confirm or disprove the cancer-risk allegations -- or even to answer any questions. While the survey will be massive, involving mortality and health records of more than 85,000 workers employed at eight chip companies, including Intel (INTC) and Texas Instruments (TXN), in the past 30 years, "there's a good chance that the study will be inconclusive," cautions Harvey Checkoway, occupational health sciences professor at the University of Washington. Checkoway was part of a panel that recommended that SIA commission a study several years ago.
Here's why: "It's very difficult to go into the past and find out what people were exposed to," says Checkoway. Adding to the challenge: The study will look at various types of fabs, employing different production processes and chemicals. Also, many medical records used in the study will likely be incomplete. Thus, establishing a link between workplace chemicals use and cancer might prove impossible.
EASIER TO DISCLAIM RESPONSIBILITY? Whatever the results, they could have a big impact on ongoing and future suits. "If a study of this magnitude is done properly and it turns up significant findings, it could have a dramatic effect on litigation," says Ronald Allen, professor of law at Northwestern University in Chicago, who has not been involved in the chip-industry legal cases relating to cancer risks.
It could give those suing semiconductor companies some proof of a connection between the fab work and illness. More likely, though, it will make it easier for chip concerns to disclaim responsibility in the cases brought against them.
That begs the question: Is an industry-funded study scientifically valid? "You don't go to [the former Enron auditor] Arthur Andersen to audit [fraud at] Enron," says Daniel Teitelbaum, adjunct professor of environmental toxicology at the University of Colorado who testified on behalf of employees in suits brought against chipmakers.
OPPOSITE FINDINGS. Sure, Teitelbaum has strong opinions. But even experts not involved with chipmakers or their former workers seem to echo this sentiment. "I think the data would be a lot easier to gather if it were a government-funded study," says Len Jelinek, an analyst at El Segundo (Calif.)-based chip consultancy iSuppli who, at one point, worked at a fab. "My guess is [this study] will be very superficial."
However, previous studies sponsored by the SIA haven't all been kind to the industry. Back in the early 1990s, one study found a higher-than-average occurrence of miscarriages among female fab workers. Chipmakers responded by removing from their manufacturing processes certain photo-resistant chemicals, called glycol ethers, which are thought to be linked to the condition.
Still, industry- or plaintiff-financed studies usually face scrutiny, and rightly so, since they often offer highly divergent results. Last year, for example, IBM hired researchers at the University of Alabama in Birmingham and Harvard University to analyze records of 126,000 employees who worked at its fabs from 1965 to 1999. The study, completed last November and now undergoing prepublication peer review, unearthed some surprising results: IBMers had apparently 16% fewer cancer incidences than the surrounding population and enjoyed 35% lower mortality rates. "IBMers are healthy," says a company spokesperson.
WHERE'S THE GOVERNMENT? But Richard Clapp, an environmental-health professor and researcher at Boston University who was paid to do a different study by plaintiffs in two cases against IBM, has come to dramatically different conclusions. After analyzing 33,000 deaths of IBM workers that occurred between 1969 and 2001, he found that IBM fab workers had two to four times higher-than-average mortality rates due to brain, breast, kidney, and non-Hodgkin's lymphoma cancers. Clapp's findings were made public in court cases. His paper is currently being edited and is expected to be published in a major scientific journal in the next six months, he says.
So who is right? Most industry insiders say that question can only be answered by an independent, government-sponsored study, which is nowhere in sight. That's in stark contrast to other countries. Officials in Britain are investigating incidences of cancer among employees at a National Semiconductor (NSM) plant in Scotland. They also hope, by September, to have completed a feasibility study concerning a government-led, industry-wide investigation.
In the U.S., it's a different picture. The National Institute for Occupational Safety & Health (NIOSH), for example, has done little work for years relating to chipmaking. "The semiconductor industry is really the golden goose of the technology revolution, and the government has left it alone," says Ted Smith, founder of an environmental-advocacy group, the Silicon Valley Toxics Coalition (SVTC), in San Jose, Calif.. The coalition does everything from staging protests during shareholder meetings to offering free informational workshops.
RECORD GAPS. Meanwhile, semiconductor manufacturing has become "more toxic" over the years, says Teitelbaum. As chipmakers struggled to follow Moore's Law and stuff more transistors onto each chip, they moved to processes using arsenic and phosphorus compounds, as well as radiation. "And they're not doing as much as they need to do to protect the workers," Teitelbaum says.
The semiconductor industry counters this assertion by citing extensive employee training and safety measures, such as requiring staffers at certain jobs to wear goggles and two layers of gloves. "Most companies [also] have the practice of using the least dangerous chemicals possible," says SIA spokesman John Greenagel.
Sorting it all out won't be easy for the scientists conducting the SIA study. Even researchers from Johns Hopkins University who considered the feasibility of doing an industry-wide cancer study acknowledged in their 2004 report that getting complete employee records will be challenging, due to the mergers and acquisitions fever of the past 10 years. Some chip businesses, such as Motorola's (MOT) semiconductor division, got spun off. And, in most cases, the new owners didn't obtain the old employee records, according to the feasibility study.
MED UNITS GONE. Besides, over the past five years, many chipmakers got rid of their own fabs and began, instead, to contract with foundries in Taiwan and China -- whose fabs aren't part of the SIA's U.S.-only study. Thus, a chunk of the semiconductor industry isn't even represented in this survey.
Whatever health records do exist are likely incomplete. Several decades ago, most fabs had on-site medical offices. "All of these companies have eliminated their medical departments in the past 10 years because they just make trouble," says Teitelbaum. Now, most chipmakers simply offer traditional health insurance. And they typically only have notes from the customary preplacement physical, performed when an employee is first hired, as well as any repeat exams, designed simply to ascertain that an employee is fit to work, according to Teitelbaum.
Finally, the industry only began keeping electronic records in the mid-1980s. Paper records from before that time might be hard to find and understand. Moreover, many electronic records might be on 5-inch floppy disks, which many newer computers can't read.
SEEKING VALIDITY. SIA acknowledges these challenges. "Much of this is true, and that was one of the challenges confronting the industry when we first received demands that we do the study," says SIA's Greenagel. "This will be a very challenging study to do, but we have been assured by the people [who did the feasibility study] at Johns Hopkins that a valid study can be done."
The bottom line: The SIA study is unlikely to make it much clearer whether working at a fab is dangerous to workers' health or not. The best it might do is raise more questions for follow-up studies to answer. Perhaps then the government will get involved and conduct a truly independent study. With Beth Carney in London
Kharif is a reporter for BusinessWeek Online in Portland, Ore.