18 Semiconductors Bruce A. Fowler and Mary J. Sexton University of Maryland, Baltimore, Maryland 1. INTRODUCTION The use of elements such as arsenic, antimony, gallium, and indium has long been featured, in the manufacture of semiconductors for computer chips, cellular telephones, and light emitting diodes (LEDs). Over the last 30 years, tens of tons of these elements have been incorporated into these devices either as dopants (1,2) for silicon-based computer chips or in the manufacture of the higher-speed III–V semiconductors such as gallium arsenide or indium arsenide (2). As the demand for higher-speed devices has increased, older devices with slower elec- tronic speeds have been discarded, in the absence of well-established recycling programs, generating a large stockpile of electronic devices containing these ele- ments collectively known as ‘‘e-waste.’’ The magnitude of this growing problem has only recently been appreciated in California and Europe (3) but much about the biological properties of these high-technology materials is not yet known. Experimental animal studies (4–9) have demonstrated that particles of GaAs or InAs are broken down in vivo resulting in the release of both constitutive ele- ments (4–9). This creates a binary chemical mixture situation raising the issue of interactive effects. The situation in the semiconductor manufacturing plants is even more complex since a number of solvents are also present and semicon- ductor workers are exposed to a number of toxic agents in the clean room environ- Copyright © 2002 Marcel Dekker, Inc. ments (10). Epidemiological studies of these workers have shown an increased incidence of miscarriages and there are case reports of brain and testicular cancers among workers employed in a gallium arsenide plant. This chapter will review the known biological effects of arsenic, including arsine gas (11–16), antimony (17), gallium (18), and indium (19), from the per- spective of experimental systems and attempt to relate these data to the findings of epidemiological studies in semiconductor workers. It is hoped that this review will provide a contemporary assessment of the present state of knowledge regard- ing the current uses and biological effects of metals/metalloids utilized in the semiconductor industry. The potential human health and environment effects of these elements either alone or as mixtures will be discussed in relation to needed areas of research. 2. EXPERIMENTAL STUDIES The toxic properties of the elements arsenic and antimony have been appreciated for many years but the use of these elements in the production of semiconductors has occurred only during the last 30 years. The utilization of these elements in increasing quantities has given rise to concern about toxic effects in workers and long-term environmental effects as devices containing these materials are discarded. It is important to note that exposures to these elements via the III–V semiconductors will be as binary mixtures of gallium, indium, and arsenic or antimony as well as arsine gas (AsH 3 ). As discussed below, the combination of these elements is generally more toxic than each of them on an individual basis owing to metal-metal interactions. Exposure of workers is usually via inhalation of particles of GaAs or InAs or via arsine gas. Each of these compounds has its own distinctive toxic properties as discussed below. 2.1 Chemical Disposition and In Vivo Metabolism It has been shown by a number of investigators that particles of GaAs and InAs undergo biological attack in vivo resulting in the dissolution of these particles and the release of both GA/In and As components. The Ga and In components are concentrated in organs such as the kidney while the As component undergoes a metabolism similar to that of As 3ϩ resulting in the urinary elimination of mono- methyl arsonic and dimethyl arsinic acids (4–9). Arsenic internalized from arsine gas exposures seems to follow a similar metabolic pathway. 2.2 Arsenic The toxic properties and metabolism of arsenic via methylation have been re- cently reviewed by the National Research Council (20). Inorganic arsenic is a general cellular poison that produces a variety of effects in different organ sys- Copyright © 2002 Marcel Dekker, Inc. tems (20). These include liver toxicity, renal toxicity, skin dermatoses, peripheral neuropathy, and cancers in a variety of organ systems (20). Urinary excretion of methylated metabolites such as monomethyl arsonic acid and dimethyl arsinic acid is the main route for elimination of arsenic in vivo for both man and a majority of animal species (20). The mitochondrion is a major target organelle for arsenicals (21,22) and the production of reactive oxygen species (ROS) sec- ondary to inhibition of mitochondrial respiratory function is a likely mechanism for initiation of cell injury and arsenical-induced carcinogenesis (20). Arsenical induction of the stress protein response has been observed by a number of investi- gators (23–26) in a variety of cellular systems following both in vivo (23) and in vitro (24–26) exposures to this agent. Prolonged in vivo exposure of laboratory animals (27–29) and humans (30,31) to arsenicals has also been shown to pro- duce a characteristic porphyrinuria pattern characterized by increased excretion of uroporphyrin with lesser amounts of coproporphyrin. Similar porphyrin excretion patterns characteristic of arsenic have been observed in rodents exposed to arsine gas (32), GaAs (12), and InAs (23,29,33). These findings are consistent with the observations noted above regarding the dissolution of GaAs and InAs particles following in vivo exposure releasing the Ga, In, and As moieties. 2.3 Antimony Antimony compounds and stibine gas (SbH 3 ) have received less attention than arsenicals since they are generally used in smaller quantities and have lower toxicities than inorganic arsenic and arsine gas (17). 2.4 Gallium Gallium compounds have been used as antitumor agents and in the manufacture of III–V semiconductor gallium arsenide (GaAs) (2). The kidney is the chief target for gallium toxicity (4,18) after acute administration and is the limiting factor in its usage in cancer chemotherapy. Nephrotoxic effects have also been reported in rodents following intratracheal instillation of GaAs particles (4) or subcutaneous administration of GaAs particles (7–9). GaAs administration has also been shown to markedly inhibit blood ALAD resulting in the increased ex- cretion of ALA in the urine (4). Administration of GaAs particles to rodents in vivo has been shown to cause the induction of a characteristic stress protein response pattern in kidney proximal tubule cells and development of renal tubular proteinuria pattern (34,35). 2.5 Indium Indium is also nephrotoxic (19) and has been shown to be a potent inhibitor of renal ALAD (36). Administration of InAs particles has been shown to produce Copyright © 2002 Marcel Dekker, Inc. a porphyrinuria pattern (29) that has elements of both the In and As moities. Administration of InAs particles to rodents also produces characteristic alter- ations in gene expression patterns in rodent renal proximal tubule cells and a more marked tubular proteinuria pattern than that of GaAs administered at equiv- alent dosage levels (35). These data indicate that the InAs compound is more nephrotoxic than GaAs at equivalent dose levels. The underlying difference seems to be related to the potent inhibitory effects of In on the protective stress protein induction response in these cells. 2.6 Gender Differences in Responsiveness The semiconductor industry has a diversified workforce with a high percentage of women employed in the manufacture of semiconductor chips (37). For this reason, concern has been focused on women of childbearing age and the potential reproductive health effects associated with exposure to chemicals used in this industry as discussed below. On a cellular basis, gender differences in the respon- siveness of renal proximal tubule cells from both hamsters and humans exposed to combinations of Ga, In, or As in vitro have shown marked differences in the responsiveness of alterations in gene expression patterns between cells derived from males and females (34). Since these exposures were conducted in vitro in the absence of hormones, it is suggested that cellular imprinting of the gene regulatory mechanisms may be the most likely explanation for the observed dif- ferences. The potential importance of such findings rests with the fact that gender- based differences in the stress protein response may account for differences in susceptibility to toxicity from these agents. 3. ORGAN SYSTEM EFFECTS 3.1 Immune System Administration of GaAs particles to rodents in vivo or rodent splenocytes in vitro has been shown to produce inhibitory effects on the humoral antibody reponse (38–43). Similar inhibitory effects on both humoral and cellular immune re- sponses have also been observed in rodents following prolonged exposure to tolerated concentrations of arsine gas (16). Data from these studies suggest that the immune system is a target organ system for these commonly used metals/ metalloid compounds and are consistent with finding of increased semiconductor worker absenteeism (44) due to illness relative to other manufacturing sectors. 3.2 Liver Acute administration of indium (44,45) and acute and chronic administration of arsenicals (20) have been shown to produce hepatoxicity. For indium, this is Copyright © 2002 Marcel Dekker, Inc. highly dependent upon the chemical form with ionic indium producing hepatocel- lular toxicity (44,45) and colloidal indium compounds producing marked toxicity to cells of the reticular endothelial system. The endoplasmic reticulum and its attendant metabolic biochemical system in hepatocytes is a major site of intracel- lular toxicity (45). Prolonged exposure to arsenicals produces hepatocellular tox- icity with the mitochondria as a major site of intracellular action. As noted previously, arsenicals including arsine produce a specific porphy- rinuria pattern of apparently primarily hepatic origin in both rodents and humans (20) following prolonged exposures. The potential hepatotoxic effects of gallium have not been studied in detail. 3.3 Kidney The group III elements Ga and In have each been shown to produce toxicity in renal proximal tubule cells (4,18,44). Arsenicals such as arsenate have also been demonstrated to produce toxicity in this renal cell population (46). As in the liver, the mitochondrion appears to be a primary target organelle with inhibition of respiratory function as a primary underlying mechanism of toxicity (20). Alter- ations in renal gene expression patterns and attendant tubular proteinuria are im- portant and marked biochemical and physiological responses of this organ system following exposure to GaAs or InAs (35). 4. INHALATION TOXICOLOGY STUDIES The inhalation toxicology of GaAs particles in rats and mice has been extensively studied by the National Toxicology Program (NTP) and the results published as a peer-reviewed report (47). Data from the 14-week studies confirmed findings from previous intratracheal instillation studies (4) with regard to lung toxicity and demonstrated the testicular toxicity of GaAs in both rats and mice. This latter effect is undoubtedly due to the known toxicity of Ga to the testes (47). In addi- tion, GaAs exposure also produced a microcytic anemia with increased zinc protoporphyrin/heme ratios also confirming biochemical effects on the heme bio- synthetic pathway. In the 2-year exposure study, a significant increase in the incidence of lung tumors was observed in female but not male rats leading to the conclusion that there was clear evidence of carcinogenic activity in female but not male rats following chronic GaAs inhalation exposure. 5. EPIDEMIOLOGICAL STUDIES Uncertainty regarding whether workers in the semiconductor industry experience adverse health effects has led to three kinds of investigations: (a) assessments of frequencies of occupational illnesses and injuries, (b) comparisons of rates of Copyright © 2002 Marcel Dekker, Inc. specific illnesses, such as respiratory functioning, and (c) examinations of repro- ductive health, in particular, spontaneous abortions among female workers. To date, studies have addressed short-term health effects, and none has addressed more chronic conditions, such as cancer. 5.1 Assessments of Incidence of Injury and Illness Annual reports from the Bureau of Labor Statistics (BLS) provide data on occupa- tional injuries and illnesses based on workers’ compensation information that are categorized by the Standard Industrial Classification (SIC) codes for industries in the United States. The information available is limited in scope and can give only suggestive indications of whether semiconductor workers have any excess of occupational injuries or illnesses because the database is not a research one. Robbins et al. (37) compared workers in SIC 367, the code for companies that manufacture all electronic components and accessories, SIC 3674, the code for those that more narrowly manufacture semiconductors and related devices, all manufacturing industries, and all private industries on incidences of injuries and illnesses for the period of 1980–85. The incidence of illness was about the same for workers in SIC 3674 and in SIC 367. The incidence of illness, however, was higher for workers in the 3674 industry when compared with those in private industries for each of the 6 years and higher than that of workers in manufacturing industries overall for 3 of the years and about equal in incidence of illness in the other 3 years. When specific types of illnesses were examined, the semiconductor workers had notably higher rates of respiratory conditions due to toxic agents when compared with workers in each of the other three industries. As the authors indicated, there was no way to establish the probability that these differences arose by chance given the limitations of the available data. However, an examina- tion of the 1998 data indicate that the incidence rate per 10,000 full-time workers for respiratory conditions was 5.9 compared to 3.5 and 2.0 for all manufacturing industries and private industries, respectively (48). Similarly, LaDou (49) more recently reported higher rates of illnesses among semiconductor workers (SIC 3674) based on California BLS data for 1987–91. He found that as a percentage of workloss cases, occupational illnesses were two to three times higher for semiconductor workers than for those in all manufacturing industries and was somewhat higher than for workers in SIC 367. Of the occupational illnesses, about one-third to one-half were categorized as systemic poisoning resulting from exposure to toxic materials. The percentages of systemic poisoning were comparable to those for SIC 367 but were about twice as high as for workers in all manufacturing industries. As LaDou (49) pointed out, changes in the manufacturing process of semi- conductors occur rapidly making it difficult to pinpoint clearly the etiology of any excess health problems that might be present. Despite this difficulty, the Copyright © 2002 Marcel Dekker, Inc. inherent variation of year-to-year rates based on BLS data, and the limitations of the data collected, it appears that injuries are not higher but that the illness rates for workers in SIC 3674 may be higher than for workers in other manufac- turing jobs and were higher in the earlier years of the eighties than for the later years. 5.2 Specific Health Conditions 5.2.1 Nonreproductive Health Conditions There is suggestive evidence that semiconductors have increased respiratory and other symptoms of compromised health. Pastides et al. (50) found that a number of symptoms were reported by manufacturing workers at a higher frequency than reported by nonmanufacturing workers in the semiconductor industry. McCurdy et al. (51) surveyed over 3000 employees in both large and small semiconductor sites. They found a higher prevalence of upper respiratory symptoms among fab- rication workers than among nonfabrication workers. Other symptoms, such as dermatitis and headaches, were reported in subsets of the work groups. Luo et al. (52) conducted a similar study of self-reported health symptoms and of measured pulmonary functioning among workers in a semiconductor plant in Taiwan in 1995. They found that compared to controls and after adjustments were made for smoking and age, males working in either the photolithographic process or the ion-implantation process had about a fourfold risk of having restrictive lung functioning indicated by spirometric measurements, although this was not a sta- tistically significant difference. Significant differences were found between fe- male controls and photolithographic or diffusion workers in airway irritation and eye irritation as well as in other self-reported symptoms. Both groups of investi- gators acknowledge that the sites studied had concentrations of chemicals well below the legally acceptable levels when monitored measurements were re- viewed. However, neither study obtained individual exposure measurements that could be linked to symptoms. The question remains of whether some workers are exposed at levels exceeding acceptable limits or whether there is a group of workers sensitive enough to experience symptoms at very low levels. Further, given the degree of cooperation needed from plant management to support such studies, it is almost surely the case that any adverse health effects will be underes- timated since the more cooperative plants are expected to have the cleanest work conditions. Despite these issues, the results from very different studies, conducted under quite varied conditions, are consistent in finding elevated respiratory symp- toms in semiconductor workers. 5.2.2 Reproductive Outcomes A higher risk of adverse pregnancy outcomes, in particular spontaneous abortions and congenital malformations, from exposure to semiconductor chemicals has Copyright © 2002 Marcel Dekker, Inc. been examined in a handful of studies. No one chemical or metal has been singled out in the semiconductor industry as a more possible causative agent than another. Furthermore, the difficulty of isolating one chemical has been raised along with the realization that the more realistic work environment is one with multiple exposures (53). Most human studies though have addressed solvent exposure because animal studies have established glycol ether as a teratogen (54–57) and because a occu- pational solvent exposure has been linked to adverse pregnancy outcome (58). Spontaneous Abortions. All but one published study on the association of exposure related to the semiconductor industry and reproductive outcomes have focused on exposure at the workplace, and most of these have focused on female workers. However, Wrensch et al. (59) examined pregnancy outcomes in women in two communities in which the drinking water was contaminated in 1980–81 from solvents that leaked from an underground storage tank of a semiconductor firm in California. Anearlier examination (60) of pregnancyoutcomesfor women in one of these communities had found a statistically significant excess of spontaneous abortions (SABs) compared to a control community, but Wrensch et al. (59), in the follow-up investigation, failed to find an association with a long period, 1980–85, and with changes in the study design. The later study included a postcontamination period and estimates of exposure based on hydrogeological modeling that took into account the amount of contaminated water delivered to individual households as well as a second exposed community. Comparisons were made with two control communities. The authors concluded that ‘‘the contaminants probably did not have a measurable impact on adverse pregnancy outcomes in this community given the doses received and the numbers of women exposed’’ (p. 299). Table1providesasummaryofthreekeycohortstudiesthatassessedthe risk of SABs in female employees with potential exposure from their semicon- ductor jobs. Most investigators adjusted for other risk factors known to relate to SAB, but the adjusted risks were not substantially different from the crude esti- mates. In the 1988 Pastides et al. (50) study a statistically significant elevation of about a twofold risk in diffusion workers compared with workers, such as clerks and administrators, who were judged to have minimal or no exposure was reported. Other comparisons were not statistically significant. Because of the small numbers in the Pastides et al. investigation, an investigation, known as the Semiconductor Health Study (SHS), with several components, was conducted. It included a larger number of workers and was designed to examine several types of health outcomes, with one component being an historical cohort study. The results have appeared been published (61–64). A statistically significant finding of an elevated risk of 1.45 for SABs among fabrication workers was reported (65). In the historical cohort a dose-response relationship was shown between level of exposure to solvents and SABs, which provided stronger evi- Copyright © 2002 Marcel Dekker, Inc. T ABLE 1 Features and Results of Historical Cohort Studies to Assess Risk of Spontaneous Abortion in Female Semiconductor Workers Measure of Measure Comparison Exposed Authors Site Exposure of SAB Group (n) Group (n) Rel Risk 95% CI Pastides, et al., 1 Mass facility work hx by inter- self-reports of clerical, admin, photolitho- 1.75 c 0.77, 3.25 1988 view and plant pregnancies engineers graphic (16) 2.18 c 1.11, 3.60 records/ob- (398) diffusion (18) .87 a 0.45, 1.60 servations of clerical, admin, fabrication (15) use of chemi- engineers cals (398) non-fab (288) Beaumont, et al., 14 U.S. facili- work hx by com- self-reports of non-fab (444) fabrication 1.45 c 1.02, 2.05 1995 ties pany records & pregnancies (447) interview Correa, et al., 2 Eastern U.S. work hx by inter- self-reports of no exposure low (125) 1.0 a 0.60, 1.70 1996 facilities view and plant pregnancies (332) medium (74) 1.4 a 0.80, 2.60 records (some con- high (30) 2.8 a 1.40, 5.60 firmed by medical rec- ords) Pinney and 1 Southeastern work hx by inter- self-reports of non-fab, no ex- fab (189) 1.62 a Lemasters, 1996 U.S. facility view exposure pregnancies posure (191) non-fab, expo- 2.00 a categories de- sure (74) veloped by in- dustrial hy- gienists a ϭ adjusted. c ϭ crude. Copyright © 2002 Marcel Dekker, Inc. dence of a causal relationship (64,66). Correa et al. (53) focused on one specific type of chemical, ethylene glycol ethers, and quantified the risk into no-, low-, medium-, and high-exposure categories. They found elevated risks of SABs for female workers in the medium- and in the high-exposure categories, with the latter having a 2.8-fold increased risk of an SAB. They also found a statistically significant trend of a higher risk of SAB related to a higher exposure, thus sug- gesting a dose-response relationship and thereby lending more biological credibil- ity to a causal association. The results from the historical cohort studies, conducted under different conditions and times, are consistent in suggesting an elevated risk of SABs. In three of the investigations a statistically significant increase in risk was observed either overall for women in process jobs that entailed potential exposure to chemi- cals or for women in jobs in a more narrowly defined part of the process. Pastides et al. (50) and Pinney and LeMasters (67) found an increase in SABs for fabrica- tion workers but the difference was not statistically significant. The findings from these cohort studies provide evidence of an increase of at least 50% in the risk that a pregnancy will end in an SAB for process workers in semiconductor jobs. We agree with Correa et al. (53) that the point estimates for the magnitude of risk are almost surely underestimated. These investigators found a smaller propor- tion of women at each level of increased risk. On the one hand, investigations that examine only one overall risk estimate will tend to find it weighted toward a smaller risk since the level of exposure for the largest proportion of the employ- ees is low. On the other hand, the partitioning of the workers into specific work or exposure units will decrease the sample size and thus the power of the study unless these small numbers have been carefully considered in the design of the study. Thus the confidence in the reported findings will be weakened because of the broad statistical confidence limits around the point estimate, but these large confidence intervals are a direct result of a sample under study that was far too small to detect elevated risks of 50–100%. Nevertheless, the cohort studies, to date, indicate an elevation in the risk of SAB in female semiconductor workers. A particular difficulty in studying SABs is the fact that some are not clini- cally recognized. Eskenazi et al. (63), as a component of the SHS, addressed this problem with a study of women who worked in semiconductor production at seven sites. Eligible women gave daily urine samples and kept a diary over a 6- month period. The urine samples were analyzed for concentrations of human chorionic gonadotropin for determination of pregnancy. This prospective cohort was then followed to assess the outcomes of the pregnancies. These investigators reported a relative risk of 1.39 (confidence intervals of 0.84, 2.31) for an SAB in women working in fabrication jobs that put them at higher risk of exposure (n ϭ 152) compared to females working in nonfabrication jobs (n ϭ 251). Though not statistically significant, the risk estimate is comparable to estimates from his- Copyright © 2002 Marcel Dekker, Inc. [...]... studies, the epidemiological studies are consistent in finding increased risks in respiratory symptoms and in adverse reproductive outcomes The confidence in these findings is increased because of variation in the study designs, in the study populations, and in the methodologies applied (66) Moreover, laboratory and animal studies have produced findings consistent with the human health studies that provide insights... and combined effects of each of the constituent elements These interactive effects appear in a number of different organ/biochemical systems including the lung, liver, kidney, hematopoietic, and reproductive systems Taken in combination with the other chemicals known to be present in work environments found in the semiconductor industry, it is not unreasonable to expect the health effects findings of... methodologies to address the issue Other Reproductive Outcomes A few studies have focused on whether the reproductive health of males because of exposure to chemicals in their semiconductor jobs is compromised Investigators have examined fertility in male workers and SABs in women married to men working in semiconductor jobs Findings from other studies support the relationship of compromise in reproduction... findings on fertility between the cross-sectional SHS study and the SHS cohort studies In sum, the human data assembled to date are inconclusive regarding whether there is any risk to the reproductive health of men because the findings have been based on such small numbers, and the results are inconsistent Some investigators have commented on adverse pregnancy outcomes other than SABs, such as low birth... insights into the biological mechanisms (47) While the processes, specifically in the United States with the use of ethylene glycol ethers, have changed since the earliest investigations, associations have continued to be found that suggest adverse health effects 6 CHEMICAL MIXTURE INTERACTIONS It is clear that binary compounds such as GaAs and InAs produce biological effects that reflect the individual... Marcel Dekker, Inc mations (50,59), but have had few data to report Thus, there is no understanding of the long-term health status of liveborn children whose mothers or fathers are exposed to chemicals because of their occupations in the semiconductor industry The studies to date, taken as a whole, are not reassuring with regard to the protection of the semiconductor workers’ health With the exception... established that a substantial proportion of these early fetal losses are chromosomal anomolies (69), it raises the question of whether these unrecognized pregnancy losses occur more frequently in women in semiconductor jobs The findings from this investigation again point up the underestimation of the magnitude of SAB risk when unrecognized pregnancies are not included In general, a case/control study is considered...torical cohort studies The analysis was based on only 19 pregnancies in the fabrication workers and 33 in the nonfabrication workers The results of this small study are consistent with those reported by Gray et al (68) It is worth noting that about three-fourths of the pregnancies detected by the assay that ended in SABs were not clinically recognized nor recognized by the woman herself Since it is well established... (confidence intervals of 0.30, 1.40) for fabrication workers A number of limitations in these case/control studies could account for the lack of consistency in the findings with those from the cohort studies (72,73) Even so, their results underscore the lack of establishing definitively a causal link between health problems among semiconductor workers and specific occupational exposures, and the need for... on the semiconductor workforce The best method of understanding the current relationship of occupational exposure from the manufacturing of semiconductors with the workers’ health is to conduct well-designed studies with sufficient numbers of subjects to insure adequate statistical power that can detect even modest health effects While the assembled evidence provides guidance in the design features, the . ethylene glycol ethers, and quantified the risk into no-, low-, medium-, and high-exposure categories. They found elevated risks of SABs for female workers in the medium- and in the high-exposure categories,. zinc protoporphyrin/heme ratios also confirming biochemical effects on the heme bio- synthetic pathway. In the 2-year exposure study, a significant increase in the incidence of lung tumors was observed in female. attack in vivo resulting in the dissolution of these particles and the release of both GA /In and As components. The Ga and In components are concentrated in organs such as the kidney while the As