Journal of Occupational Medicine and Toxicology New Views on the Hypothesis of Respiratory Cancer Risk from Soluble Nickel Exposure; and Reconsideration of This Risk's Historical Sources in Nickel Refineries

Introduction: While epidemiological methods have grown in sophistication during the 20 th century, their application in historical occupational (and environmental) health research has also led to a corresponding growth in uncertainty in the validity and reliability of the attribution of risk in the resulting studies, particularly where study periods extend back in time to the immediate postwar era (1945–70) when exposure measurements were sporadic, unsystematically collected and primitive in technique; and, more so, to the pre-WWII era (when exposure data were essentially non-existent). These uncertainties propagate with animal studies that are designed to confirm the carcinogenicity by inhalation exposure of a chemical putatively responsible for historical workplace cancers since exact exposure conditions were never well characterized. In this report, we present a weight of scientific evidence examination of the human and toxicological evidence to show that soluble nickel is not carcinogenic; and, furthermore, that the carcinogenic potencies previously assigned by regulators to sulphidic and oxidic nickel compounds for the purposes of developing occupational exposure limits have likely been overestimated.

significant arsenic exposure during the processing step in the Clydach refinery's hydrometallurgy department in the 1902-1934 time period likely accounts for most of the elevated respiratory cancer risk observed at that time. An understanding of the mechanism for nickel carcinogenicity remains an elusive goal of toxicological research; as does its capacity to confirm the human health evidence on this subject with animal studies.

Introduction
While epidemiological methods have grown in sophistication during the 20 th century, their application in historical occupational (and environmental) health research has also led to a corresponding growth in uncertainty in the validity and reliability of the attribution of risk in the resulting studies, particularly where study periods extend back in time to the immediate postwar era  when exposure measurements were sporadic, unsystematically collected and primitive in technique; and, more so, to the pre-WWII era (when exposure data were essentially non-existent). These uncertainties propagate with animal studies that are designed to confirm the carcinogenicity by inhalation exposure of a chemical putatively responsible for historical workplace cancers since the exact historical exposure conditions were never well characterized. In this report, we present human and toxicological evidence to show that soluble nickel is not carcinogenic; and, furthermore, that the carcinogenic potencies previously assigned by regulators to sulphidic and oxidic nickel compounds for the purpose of developing occupational exposure limits have likely been overestimated. [Note to the reader: Nickel-containing substances can be grouped into five main classes based on their physicochemical characteristics: nickel carbonyl (gas), metallic nickel (e.g., elemental nickel, nickel-containing alloys), oxidic nickel (e.g., nickel oxides, hydroxides, silicates, carbonates, complex nickel oxides), sulphidic nickel (e.g., nickel sulphide, nickel subsulphide) and water soluble nickel compounds (e.g., nickel sulphate hexahydrate, nickel chloride hexahydrate). Exposures during nickel refining may contain several of these nickel species depending on the type of process used.] Support for the soluble nickel carcinogenicity hypothesis was found in the epidemiological findings at two refineries, involving high exposure to soluble nickel, i.e. nickel sulphate hexahydrate (1-5 mg/m 3 ), of workers in the electrolysis department at the Kristiansand Nikkelrafferingsverk refinery (KNR) in Norway [1][2][3][4][5][6][7][8] and the hydro-metallurgy department at Clydach Wales [3]. These findings led the International Committee on Nickel Carcinogenesis in Man (ICNCM) to conclude in 1990 that 'soluble nickel exposure increased the risk of these cancers [lung and nasal] and that it may enhance risks associated with exposure to less soluble forms of nickel [i.e. sulphidic and oxidic nickel]' ( [3].pp74). The ICNCM exercised caution and prudence in this conclusion despite available contradictory epidemiological evidence from a nickel refinery study in Port Colborne Ontario (PCNR) that found no increased risk of lung cancer among its electrolysis workers who also had soluble nickel exposures comparable to those in the corresponding KNR department [9,10]. Both refineries (KNR and PCNR) used the Hybinette electrolytic refining process [11,12] and, although PCNR electrolysis workers had somewhat less exposure to airborne soluble nickel than KNR workers, differences were likely due in part to the classification of nickel carbonate as insoluble at PCNR and as soluble at KNR. KNR electrolysis workers reportedly experienced higher levels of insoluble nickel exposures than did PCNR workers, especially before 1967 ( [3].pp20).
The present paper focuses primarily on published KNR human health studies for two reasons: (1) because KNR studies still show lingering respiratory cancer risk after 30 years of epidemiological studies, which, if true, must raise serious occupational and public health concerns for Norwegian health authorities; and (2) because it remains in current production, KNR's evidence provides the gravitas of evidentiary support for soluble nickel's carcinogenicity. The Clydach refinery era of epidemiological interest in this respect extended from 1902 to 1937 after which time the throughput on Clydach's copper extraction (copper plant) and nickel sulphate refining (hydrometallurgy) departments had been considerably reduced. By 1948, the copper leaching step on calcines and the nickel sulphate recycle were eliminated, ending the nickel-copper oxide dust and nickel sulphate spray and mist hazards in the copper plant ( [3].pp15-16).
In its investigations, the ICNCM reported that no measurements of actual nickel concentrations, let alone nickel species, existed in the workplaces of any nickel plant operations before 1950 ( [3].pp11). Very few measurements were available before the early 1970s for the KNR refinery ( [3].pp15- 16), and likely for the Welsh refinery as well. In the absence of real exposure data, therefore, the range and percentage of total airborne nickel (and of nickel species) were estimated on the basis of process knowledge, subjective impressions of relative dustiness, and a few measurements ( [3].pp12-13). KNR historical exposure data were similarly based on the subjective judgements of retired personnel with the distribution of nickel species in airborne dust assumed to be the same as that in the bulk feeds and materials handled ( [3].pp15-16). In their Clydach risk-exposure modeling study, Easton et al. rightly acknowledged the uncertainties in their nickel speciesspecific cancer risk models, which they found to be highly sensitive to small shifts in the historical values imputed to insoluble and soluble nickel exposures [13].
Focusing the human health studies exclusively on nickel without considering exposures from nuisance carcinogens in the mined nickel ore and production steps has also meant that few recorded measurements of these contaminants (viz. arsenic, sulphuric acid mists) are available today to estimate their possible contribution to observed carcinogenic risk. The established human health evidence on nickel has necessarily influenced the interpretation of nickel toxicology studies as well. In this paper, we will demonstrate that epidemiological studies have not proven that soluble nickel is carcinogenic. Indeed, this shift in the human health evidence must change the interpretation of soluble nickel's toxicology, and raise questions for regulatory toxicologists to consider concerning possible overestimation of the carcinogenic potencies previously assigned to sulphidic and oxidic nickel.

Methods
We examined in detail all published reports of occupational cancer in nickel operations around the world with environmental exposures to soluble nickel, including refineries at Kristiansand Norway [1][2][3][4][5][6][7][8], Clydach Wales [3,[14][15][16][17][18][19][20][21], Port Colborne Ontario [9,10], Thompson Manitoba [F1: Roberts RS, Jadon N and Julian JA: A mortality study of the INCO Thompson workforce. McMaster University, 1991. Available from the authors], and Harjavalta Finland [22,23]; and a British nickel-plating company [24]. We also obtained file and archival information from the KNR and PCNR environmental departments. Our examination included: historical production processes, environment and hygiene issues at both refineries; personal files, including a detailed report, filed with the ICNCM, of KNR's building development, process steps and exposure patterns over the 1910-1986 period [F2:  23,1977. Available from Xstrata Nickel], and a translation (from Norwegian) of a publication of KNR's history [25]. We reviewed a published study of historical environmental exposures in KNR's Roasting, Smelting and Calcining (RSC) department that was cited in support of the substantive changes to the original KNR JEM that resulted in the historical exposure dataset for all post-1998 KNR occupational health studies [26]. On the subject of arsenic exposures, we also examined published and file materials and anecdotal evidence on: (1) historical arsenic exposures in nickel refinery process operations arising from arsenic-rich nickel ores mined in the Sudbury basin [27] and putative associated risks [10,28,29]; (2) the presence of arsenic in KNR's purification section, which was connected to its Ni electrolysis department; and on (3) sulphuric acid contaminated with significant concentrations of arsenic that was used for copper extraction at Clydach during the critical time period of high respiratory cancer risk at this refinery   [14,27]. Finally, we examined the toxicological literature related to soluble nickel and related animal studies [30][31][32][33][34][35][36][37][38][39][40][41][42][43].

The effects of topography and building architecture on the presence of insoluble nickel exposures in KNR's electrolysis department and their absence in PCNR's Ni tankhouse
The KNR began operations in 1910 on a Norwegian fjord with a land base of 10 hectares of typical hilly terrain in order to access cheap power and transport by sea [25] ( Figure 1). The PCNR began production in 1918 on 360 acres of a flat and uneventful former lake bed on the shores of Lake Erie, also to access cheap power and marine transport. PCNR's buildings and working areas occupied about 220 acres (89 hectares) of the property, almost 9 times the size of the comparable KNR foot print ( Figure  2). Both plants employed the Hybinette electrolytic process, the final step in nickel refining and source of soluble and metallic nickel exposures in their respective electrolysis departments, which also carried trace level exposures to oxidic nickel but very low exposures to sulphidic nickel compounds. [Note to the reader: For complete accuracy, it is noted that a small portion of the PCNR tankhouse was devoted to electrolytic refining of sulphidic anodes starting in the mid-1950s until the Thompson refinery was commissioned in 1960. Exposure to nickel sulphides in the PCNR tankhouse would have been low and of relatively short duration.] KNR has a unique and eventful history that included partial destruction by fire and cessation of operation in 1918, followed by the refinery's repair and reopening only to face shutdown and bankruptcy during the twenties because of the sharp downturn in global nickel prices. Following its purchase by Falconbridge Nickel Mines Ltd in 1928, it was modernized and resumed operation in February 1930 [25]. The plant was occupied and operated by German forces from April 1940 to the cessation of hostilities in Europe in the summer of 1945. The following chart shows that, except for the shutdown in the twenties and the war period, KNR always operated more intensively (as measured in tons of nickel produced per year per hectare of land base) than PCNR (including 1961 when PCNR's production level fell by over 90%) (Figure 3). PCNR's flat topography and ample land base allowed physical separation of key buildings and horizontal process layouts. Unlike the PCNR facility, KNR's topography and foot print necessitated multi-storied building structures that either abutted each other or were connected by covered tramways linking successive process steps ( Figure  4) ( Figure 5) ( Table 1). The schematics highlight building development, including the evolution of the Hybinette process refining steps over four time periods (i.e. 1910-29, 1930-49, 1950-69, 1970-78) [25], and support our contention of cross-contamination of KNR's electrolysis department environment by known carcinogens (sulphidic and oxidic nickel) originating within its RSC department. For example, Thornhill (1986)  Differences in (1) land topography and footprints led to (2) differences in production intensity and to (3) differences in building architecture at the two refineries (including stacking, abutment and connection of key KNR department environments, and the isolation of PCNR's Ni tankhouse from its LC&S building and insoluble Ni carcinogenic exposures). Coupled with (4) KNR's disruptive production history, these factors all contributed to significant differences in each refinery's environmental hygiene history over the twentieth century and were likely responsible, in our opinion, for the presence of known insoluble nickel carcinogenic exposures (i.e. oxidic and sulphidic nickel) in KNR's historical electrolysis department and their comparative absence in the corresponding PCNR department. KNR researchers have criticized the PCNR study's mortality ascertainment methods, contending that it underestimated the carcinogenic risk of its electrolysis workers. Their critique is addressed fully by the analysis provided in Appendix 1 and accompanying tables (Table  14 and Table 15).

Exposure and worker misclassification issues in the published KNR epidemiology
KNR's epidemiology studies can be grouped for examination into three time periods distinguished by the methodology for assigning person years at risk (PYRs) to exposure categories defined by process department, job type, time period and nickel compound ( Table 2).

KNR studies using rule based allocation of workers to process department
The earliest studies by Pedersen et al. (1973) [1] and Magnus et al. (1982) [2] adopted a rule based procedure to assign a worker's case (if he contracted cancer) and his PYRs to electrolysis, RSC or 'other specified' work processes, depending on which of these three categories he had spent the longest time even if it was less than half of his overall KNR employment experience ( Table 3). The Scale drawing of KNR showing building layouts and process flows by time period process classification rules in both studies made it impossible to distinguish respiratory cancer risk among the key roasting-smelting and electrolysis departments (Table 4); and even assigned nasal cancer risk implausibly to 'other specified processes' and administrative and service areas. Both studies found that cancer risk was elevated throughout the KNR refinery, an unlikely finding that signals the presence of misclassification problems. In retrospect, the Pedersen et al. [1] study was the first human health study to raise the hypothesis of soluble nickel's carcinogenicity in the scientific literature.

KNR studies using ICNCM Job Exposure Matrix developed by protocol
The ICNCM provided the impetus for fresh research on nickel carcinogenicity at KNR. Research was governed by a protocol defining a rule based procedure, followed by a consensus committee of retired personnel, to review employment records and develop a JEM to assign species specific nickel exposures to every KNR worker [F3]. The protocol was developed by a team from Falconbridge KNR and Canada, the Norwegian Cancer Registry (NCR), and the Norwegian Institute of Occupational Health (NIOH) and chaired by one of us (Thornhill) who had specific responsibilities to gather and prepare data on species, specific historical exposures and their quantitative ranges, and to confirm results with KNR and NIOH officials. He recalled warning KNR researchers that the refinery's historical records could not support the elevation in individual worker exposure levels that would result from converting the original JEM's exposure categories from ordinal to continuous values (by averaging range boundaries).
The next table (Table 5) is drawn from the resulting KNR study published in the ICNCM (1990) report [3]. The estimates display the same problem identified in earlier studies, namely that lung cancer risk remained improbably elevated throughout the refinery including administrative and service department areas. This finding underlines the persistence of misclassification problems in KNR's epidemiology.
These problems may be related to the presence of a parttime or seasonal subcohort. We discovered historical KNR employment data filed with the ICNCM that showed enormous annual turnovers in staff, averaging over 50% annually during the 1951-69 period ( Table 6) [F2]. This finding supports the existence of a large part-time workforce of men entering and leaving the refinery every year (since it would have been impossible to train over 600 new job entrants annually). Part time workers may have circulated in more heavily exposed jobs and departments on the principle that seniority was the pathway to better jobs. Their employment records would be less likely to provide reliable documentation of their department and job histories, largely because they would have entered a labour pool where departmental foremen assigned jobs on the basis of daily requirements. Anecdotal reports suggest that these seasonal workers included local farmers and merchant seamen with their own acquired risk histories ( socio-economic status (SES) and heavier smoking behaviour (an ever-smoking prevalence of 82% was found in the historical KNR workforce [2]). No account of this workforce was provided in the published KNR studies, and failure to analyze its epidemiology separately may account for the misclassification issues.

KNR studies using revised Job Exposure Matrix
On the basis of environmental studies conducted in the nineties (discussed later), Grimsrud et al. (2000) revised the original KNR JEM [5]. Revisions included backcasting over the 1910-73 time period and the development of nickel speciation fractions and levels by department and time period ( [5].pp340). We examined the effect of the revisions on the cumulative exposures to nickel species [mg m -3 yr] predicted by the ICNCM and Grimsrud et al. JEMs for a hypothetical KNR worker employed continuously over successive 10 year postwar periods in key categories of work/departments (Table 7) (Table 8). We performed this analysis knowing that correlation and regression analyses examining dose-response relationships between nickel exposure and lung cancer risk would apportion risk for a worker whose job experience fell within a specific category of work and time period according to the absolute and relative values of exposure to each nickel species predicted by the JEM for that time and place. Statistically speaking, the revised absolute and relative exposures would affect estimates of lung cancer carcinogenic potency for the risk in each JEM cell defined by department and time period.
The JEM changes by Grimsrud et al. [5] (shown in Table 7) produced enormous reductions in nickel exposure across all species, categories of work and time periods (e.g. 80-90% reduction in total exposure in the nickel electrolysis category). On the other hand, relative exposure to soluble nickel was increased in 4 of 5 categories of work (copper leaching excepted) by reducing relative exposure to oxidic nickel in those categories. In four departments [roasting (day workers), old smelter building no. 1 (day workers), copper leaching and copper cementation], sulphidic nickel levels increased, dropping only in nickel electrolysis (shown in Table 8).

Plan view of the three floors of KNR's Purification section
The reductions in KNR's historical exposure values had the effect of increasing lung cancer risk (per unit dose) for all nickel species in dose-response modeling studies. The effect of increasing relative soluble nickel exposures and decreasing relative oxidic nickel exposures was to increase soluble nickel's share of the overall risk at the expense of oxidic nickel's share. The absence of a systematic and protocol-driven procedure for these revisions meant that, unlike the original KNR JEM, it was impossible to test the validity and reliability of the resulting exposure dataset's Vertical section through row of KNR cementation tanks shown in Figure 4

Process Flows Description
(2) to (3) Ground matte lifted to roasters @ 25 m elevation using bucket elevators (144 t/day) a (3) to (3) Cooled calcine to air classification in closed circuit regrind @ 35 m elevation (216 t/day) (3) to (6) Calcine to copper leach (205 t/day) (6) to (5) Residue fine fraction to anode smelting (97 t/day) (5) to (9) b Anodes to Ni electrorefining (6) to (4) Residue coarse fraction to Mond reducers before 1953 (hydrogen reduction after) (46 t/day) (4) to (10) Reduced Cu leach residue to copper cementation (38 t/day) (10) to (3) Cement Cu (17 t/day) and dried cement Cu slimes (23 t/day) to roasters c (10) to/from (11) d Cement Cu slimes to drying (40 t/day) before transfer to roasters c (10) to (15) Crude effect on risk estimates in subsequent modeling studies. In the ICNCM JEM, averaging created a systematic upward bias in absolute exposure values, whose effect on risk estimation could have been studied. In our opinion, this is not possible with the latest KNR JEM and obscures the search for the sources of lung cancer risk in the refinery.
Without access to the complete KNR epidemiological database, it is impossible to reach precise conclusions. However, this preliminary examination strongly suggests that the overall effect of KNR JEM changes by Grimsrud et al. [5] was to increase soluble nickel's share of the overall risk of lung cancer in the refinery. This increase came in key departments [i.e. roasting and smelting, and electrolysis] identified in a succession of KNR studies from Pedersen et al. (1973) to Grimsrud et al. (2000) [1][2][3][4][5] as the principal sources of the refinery's lung cancer risk. Furthermore, it appears that the increase in risk attributed to soluble nickel exposures came primarily at the expense of oxidic nickel since this latter species' hypothesized share of carcinogenic risk declined. The rationale provided by Grimsrud et al. (2000) [5] to justify changes to the original ICNCM job exposure matrix and its use of backcasting procedures to fill in the empty portions of the refinery's  exposure history back to the 1910 start date lack a sound scientific basis. Part of this rationale hinges on a key environmental study by Andersen et al. (1998) [26] that is shown in the next section to be scientifically unsound. This finding calls into question the validity of inferences drawn in Grimsrud et al. (2002Grimsrud et al. ( , 2003Grimsrud et al. ( , 2005) that were based on the revised JEM [6][7][8].   the reader can note the mostly non-significantly elevated risk in workers with 1-14 years since first exposure (and upturn in risk for the most recent subcohort yet again), suggesting that these men were entering the workforce with prior lung cancer risk.
In a recent e-letter, Andrews and Heller (2006) published an analysis of the Grimsrud et al. (2002) [6] case control study [44], which used the revised JEM, to demonstrate that smoking and nickel exposure were strongly related in their study, making it impossible to assess the risk from exposure. Appendix 2 lists the SAS ® program code for our analysis (see Appendix 3 for additional explanatory material). The principal author replied by dismissing our concerns [45]. However, the counterintuitive relationship between risk and year of first exposure and the entrenched prior risk in new hires discussed above reinforce the conclusions in our analyses showing smoking and nickel exposure interaction in the most recent study.

KNR environmental studies
Concern about the levels of soluble nickel exposure in KNR's electrolysis and RSC departments was noted in the Preface to the ICNCM report ( [3].pp5-6); and led to a 1998 speciation study at the refinery [26]. Its purposes were: to investigate if workers in the RSC department were exposed to soluble nickel, to demonstrate a speedier method for speciation than the Zatka et al. (1992) indus-try standard [46], and to confirm the presence of soluble nickel compounds by other analytical methods. This study was problematic by its very nature. For example, it assumed the same type of roasting was taking place in KNR's new fluid bed furnaces as in its old multi-hearth Herreshoff furnaces (replaced by 1978). Process feeds and kinetics of roasting for the two furnace technologies are, however, very different. The newer roasting uses a copper sulphide residue after leaching most of the nickel with chlorine [47], which is not at all like the multi-hearth roasting where the feed was a nickel-copper sulphide matte. Not only are the feeds different for the two furnace types; the roasters themselves are very different. The old multi-hearth had a well controlled temperature gradient to prevent caking and sintering as the feed fell in stages from top to bottom. In contrast, the fluid bed is indeed fluidized and, therefore, much more homogeneous in temperature. Therefore, the kinetics and chemistry of the roasting processes in the two furnace types is expected to be significantly different. Furthermore, the amounts of dust leaking out of the older multi-hearth roaster far exceeded dust leakages from a fluid bed roaster. For these reasons, therefore, it made no scientific sense to design a study to collect samples from the four floors and basement of the new roaster building when the old Herreshoff furnaces no longer existed. One could reasonably hypothesize that each floor accessing a different height of a multihearth roaster would have differences in dust reflecting differences in the chemistry and temperatures at each level of the roaster, and this fact would be reflected in aerosol sample differences. However, these conditions would not apply in a modern fluidized bed roaster. The authors gathered data to measure roaster conditions that no longer existed!
The sampling methods were also of concern. Five parallel sets of stationary samples were collected for each floor and the basement for a total of 25 samples using an airflow rate of 20 m 3 d -1 over 3-6 days. This procedure yielded dust samples from each filter weighing 50-100 mg. These sampling methods can be compared with those in the  studies, also conducted at the same refinery and time period, that measured inhalable and total aerosol exposures for four different process areas including roasting/smelting processes [48,49]. The latter studies used personal aerosol samplers mounted on a lapel in the worker's breathing zone for a full work shift, where possible, but for four hours at least at flow rates of 2 L min -1 . The sample measurements gathered from the roasting/smelting process (using 37 mm cassette samplers) averaged 0. 12     volume of air as the former team (1 L min -1 = 1.44 m 3 d -1 ) and would have collected 3.6-7.2 mg of inhalable and 3-6 mg of total aerosol exposures, respectively. The differences in sampling methods in the two studies are also, therefore, of concern.
We asked Dr.  [50] noted that oxygen and water in the air swept through the particles on a filter could cause oxidation and sulphate formation, changing the values estimated for the nickel phases.
The Andersen et al. [26] study samples were separated into two groups so that an external laboratory could apply the speciation method developed by Zatka et al. (1992) [46] as a check on the modified method that was proposed by the authors to provide rapid measurements of two phases only, soluble and insoluble nickel. The speciation results for all floors but one overestimated the soluble nickel percentage, which Zatka attributed to the modified method's reliance on the Blauband ("Blue band") filter, which would have passed some of the finest solid particles through its relatively larger pore size. The most startling result reported in Andersen et al.
(1998) was the Ni:Cu ratio ( [26].pp688). In the feed to the copper-sulphide roasting, the authors reported a Ni:Cu ratio of 0.17. They also reported that workroom air sample ratios ranged from 0.29 to 0.62. The authors provided no explanation for this finding, simply noting that it was an 'interesting result' ( [26].pp688). Nickel dust preferentially exiting the roaster provides one highly improbable explanation. Another more likely explanation is that fugitive nickel-containing aerosols were infiltrating the workroom areas from elsewhere in the plant. This idea is difficult to dismiss because of KNR's unique contiguous and stacked plant layout features and the opportunities it provided for the migration between departments of dust generated by other refinery processes. In fact, the authors raised this idea in their introduction without pursuing it ( [26].pp687): That report [Pedersen et al. (1973)

Other nickel operations with soluble nickel exposures
Estimates of lung and nasal cancer risk from the most recent studies of other nickel operations with environmental exposures to soluble nickel are depicted in the next table (Table 10). Except where noted, risk estimates did not account for the prior risk of lung cancer (from smoking or off site risky work exposures to asbestos, pesticides, etc.) by removing the first 15-20 PYRs since first exposure.
For Clydach workers, lung cancer risks were significantly elevated among men first employed during the operation of the copper plant and hydrometallurgical departments (before 1937), a period coinciding with arsenic contamination in the environment (see next section). However, by the 1930's, risk for this inception cohort had fallen to levels consistent with higher putative smoking prevalence in the workforce (defined by the ICNCM's chair as an SMR with a lower bound on the 95% CI under 150) ( [3].pp6).
Clydach epidemiologists have noted that 'the greatest change in exposure to a known carcinogen that occurred over this period [<1910-1924] was, of course, the increase in ciga-  and oxidic nickel were present. LC&S workers acquired risk with as little as 1 year of exposure (SMR = 183) [10]. None was found in other (Non-LC&S) departments, including electrolysis work with its predominantly soluble nickel exposures. The ICNCM report examined the PCNR cohort data and found no evidence of lung cancer risk that could be attributed to electrolysis work, but did find nasal cancer risk among electrolysis workers with less than 5 years of LC&S work. However, no nasal cancers occurred in a subgroup with 15 or more years since first electrolysis department exposure who had "high" soluble nickel exposure (from the washing of anode scrap and pumping of anode slimes) and less than 5 years of LC&S work ( [3].pp56). This same subgroup's lung cancer risk, however, was elevated with 6 observed and 2.51 expected deaths, 12% and 7%, respectively, of the observed and expected lung cancer deaths among all PCNR workers with less than 5 years in LC&S work. The report noted that only 5.6% (N = 109) of the electrolysis workers had any exposure in these areas, and only 1.3% (N = 25) had more than five years exposure. The ICNCM's authors acknowledged that sintering work in this subgroup weakened their argument for soluble nickel risk.
The latest Harjavalta study [23] found elevated lung cancer risk in nickel smelter and refinery workers with 20+ years since first exposure (SIR = 200 and 338, respectively). However, the risk in smelter operations (not considering latency) was confined to workers with less than 5 years exposure (SIR = 250), and none was found in the 5+ years of exposure group (SIR = 101). Although lung cancer risk in the refinery workers (not considering latency) was elevated in both the <5 years and 5+ years exposure groups (SIR = 375 and 199, respectively), the highest risk was again found in the group with least duration of exposure. This declining gradient of lung cancer risk (with increasing years of exposure) in both smelter and refinery workers suggests employee misclassification, possibly related to: the assignment of men who worked at two or three refinery sites in all categories ( [23].pp246), sulphuric acid mist exposure (see following), and smokingrelated confounding. Nasal (and stomach) cancer risk was found in refinery workers with 20+ years since first exposure (2 and 3 cases, respectively) and with 5+ years of exposure (2 and 4 cases, respectively).
No account was taken of sulphuric acid mist exposure, a Group I carcinogen [51], in the leaching of nickel matte and electrowinning processing at Harjavalta [52]. For the sake of completeness, a mortality study of British nickel platers was reviewed [24]. Workers in the chromium plating or nickel/chromium plating departments were excluded, leaving 284 men who received nickel chloride and nickel sulphate aerosol exposures in the nickel plating departments. Stomach cancer was the only reported diagnosis with elevated risk (8 observed and 2.49 expected deaths).

Arsenic as a source of carcinogenic risk in nickel production
The role of arsenic, a Group I carcinogen [53], as an agent of historic occupational cancer risk in the nickel industry has never been adequately investigated despite case reports as early as 1939 of arsenic induced illness [14]. Arsenic is often found in nickel ore bodies, and where it appears as orcelite, a complex defect structure of Ni After calcining, a large fraction of the copper was leached out with dilute sulphuric acid (~10%) and, after recrystallization, marketed as copper sulphate. This was a large operation involving tens of thousands of tonnes of sulphuric acid per annum. The copper-depleted calcine was transferred to a closed system of sequences of towers for the reduction, carbonylation and decomposition steps for the removal of nickel. The residue comprised some 20% of the original charge, but was still relatively rich in nickel, cobalt and precious metals. This concentrate then underwent sulphidization using gypsum, sand and coke in a batch furnace operation. Several batches were combined, calcined, leached again to remove copper oxide, reduced and carbonylated to enable further nickel recovery. This process was repeated up to seven times to recover more nickel and commercially significant quantities of cobalt and precious metals. What was not appreciated until 1920 was the fact that use of a cheap source of sulphuric acid contaminated with significant quantities of arsenic was a contributor to the inefficiency of nickel extraction. Furthermore, this unwanted arsenic was successively concentrated in the recycling processes, reaching 8-10% in later batches as determined recently in analyses of two samples of Clydach process materials in powder form dating back to 1920 and 1929. Analysis of 12 elements in the samples revealed significant differences only for arsenic and iron. The 1920 sample contained 9.6% of arsenic and 4.4% of iron while the 1929 sample had 1.0% and 0.8% respectively. Both samples contained arsenic in the form of the compound orcelite. It appears to have been formed by interactions occurring, most probably, in the furnacing operations. Draper (1997) remarks that the presence of arsenic in the process materials was well-known and some concern about the medical implications was expressed by the medical staff, because there was some evidence of arsenicism among process workers [14,27]. Also, the sample particles were of respirable size, averaging 2 μm in diameter. The presence of arsenic contamination in Clydach's refining processes during the 1902-1934 period has been hypothesized to account for much, if not all, of the observed respiratory cancer risk during this time. From 1932 to 1936, the entire calcine-leaching-copper sulfate production-concentrate recycling was eliminated and the Bessemer matte feedstock was replaced with low copper, low sulphur feedstocks [27].
To address the Clydach refinery arsenic hypothesis, Draper (1997) reconstructed detailed work histories for the 365 respiratory cancer cases (280 lung and 85 nasal cancers) attributed to exposure during the 1901-1970 period. He found that 81 of the 85 nasal cancer cases and 260 of the 280 lung cancer cases began work during the high risk period before 1928. The work records of 215 lung and 85 nasal cancer cases showed that their critical exposures all occurred either from processes operating until the mid-1930's or from contaminant residues lingering in machinery or buildings from the operations of the first four decades. The seven high risk job designations in these work histories included 4 major sequential process sites in the winning of nickel metal (calciner, copper shed, nickel shed and furnaces), 2 main subsidiary process sites dependent on the main process line (copper sulphate and nickel sulphate), and rigger/fitter/handyman (the skilled technicians and tradesmen that maintained or rebuilt the machinery of the production lines) [27].
In the all-sulphate system in use at KNR up until about 1953, it was necessary to treat the anolyte with reduced matte to neutralize excess acid and precipitate the copper by cementation. The arsenic was also precipitated in this step and was recycled to the roasters with the cement copper. Arsenic was thus allowed to build up in the refinery circuit to as high a concentration as could be tolerated without contamination of the final products, and was bled from the system by the periodic removal of cement copper (containing 10.4% As by weight) (Table 11) ( [3].pp17). Cementation was carried out in 20 mechanically agitated tanks. We noted elsewhere that 'even though the reduced matte was delivered in a moist condition, the feeding operation was reported to be one of the dirtiest in the plant.' KNR electrolysis workers, therefore, were likely exposed to nickel arsenide dust made airborne by passage of heated air through a bed of cement copper slimes located in the Electrolysis Department for drying in preparation for return (by ER personnel) to the roasters. After conversion to a predominantly chloride circuit in 1953, it was advantageous to precipitate iron before cementation, resulting in a reduction in arsenic (in the cement copper step) from 10.4% As by weight before 1953 to 0.3% afterwards (although As was eliminated from the circuit after 1953 primarily in the Fe precipitate step containing 4.0% As) ([F2].pp14).
The issue of lung cancer risk and arsenic exposures at KNR was recently addressed in Grimsrud et al. (2005) but relied on the revised JEM described above for its analysis, one of the several reasons that undermine the study's findings associating excess risk with water soluble nickel exposure [8].
Although arsenic is present in the mined nickel ores in the Sudbury basin, no systematic measurements were ever reported. Nevertheless, some published data exist. The basin's Frood and Garson mines provided arsenic-rich nickel ores for the Coniston sinter plant where significantly elevated lung cancer risk (SMR = 298) was recorded [10]. Until 1934, Coniston's bessemer matte in which the concentration of arsenic as an arsenide was about 0.2%, was Clydach's feedstock [27]. Falconbridge's arsenic-rich nickel mine in the basin (where elevated As levels in soil were recently detected in a risk assessment proceeding under regulatory authority) provided its sinter plant's feedstock where elevated lung cancer risk was reported (SMR = 144) [28]. This plant's nickel matte was shipped to KNR for final processing. In both sinter plants, sintering preceded the smelting step. In contrast, sintering followed the smelting step in both the Copper Cliff (CC) sinter plant and PCNR's LC&S department where lung and nasal cancer risks were significantly elevated (p < 10 -7 ) [10,29]. Before 1956, all nickel sulphide destined for electrolytic refining at CC was sintered to oxide, reduction smelted, and cast into metal anodes. To avoid preferential fusion rather than oxidation, the sulphide feed was diluted with five times its weight of sinter returns [54], undoubtedly magnifying dust levels for all contaminants, including arsenic, enormously and causing the significant respiratory carcinogenic risk reported by Chovil et al. (1981), among others, for this plant before the process was changed in 1962 [29]. This study found 57 cases of lung cancer and 7 of nasal cancer in a cohort of 495 workers, all of whom were employed at CC between 1948 and 1962. All 64 cases were first employed before 1957; and the reported SIRs (standardized incidence ratios) were 10.71 and 1.85 for men first employed in 1948-51 and 1952-62 respectively. We contend that enough published evidence has existed for some time to establish the hypothesis of arsenic's contribution to the overall historical respiratory cancer risk recorded at CC, Coniston, the Falconbridge sinter plant, PCNR's LC&S department, Clydach and key KNR departments including electrolysis; and warrants further investigation.

Review of the toxicology of soluble nickel
The US Environmental Protection Agency (EPA) issued a nickel health advisory document in 1986 to signal speciation as a leading regulatory concern in the determination of nickel's carcinogenic potential [33]. This concern led to the creation of the ICNCM (discussed earlier) whose report in 1990 [3] concluded that more than one form of nickel can give rise to lung and nasal cancer and that much of the respiratory cancer risk seen among nickel refinery workers could be attributed to exposure to a mixture of oxidic and sulphidic nickel at very high concentrations (≥ 10 mg Ni/m 3 ). The ICNCM report also concluded that there was evidence that soluble nickel exposure increased the risks of these cancers and that it may enhance risks associated with exposure to less soluble forms of nickel. It also reported that no evidence was found that metallic nickel was associated with respiratory cancer risks. The ICNCM looked for support for its findings to animal carcinogenesis studies then underway using inhalation as the route of exposure for nickel subsulphide, high temperature ("green") nickel oxide and nickel sulphate hexahydrate. It also looked to future work on the mechanisms of nickel carcinogenesis to help unify and explain its findings and those from animal experimentation.
Although not related directly to respiratory cancer risk, we note in passing that newly published studies using a population based birth and perinatal registry for the Arctic town of Monchegorsk, Russia where a nickel refinery is located found no negative effect of maternal exposure to water-soluble nickel on the risk of delivering a newborn with malformations of the genital organs [55][56][57].
A clearance study by Benson et al. (1994) [42] demonstrated a retention half-life of 4 days for nickel subsulphide and 120 days for green (high temperature) NiO exposure in inhalation studies of F344/N rats. The Ni 3 S 2 study also detected nickel in kidney and other extrarespiratory tract tissue indicating that its clearance was dominated by a dissolution rather than a mechanical clearance pathway. Nickel was not distributed to other extrarespiratory tract tissue in the NiO study. Benson et al. (1995) [43] found that approximately 99% of the inhaled nickel sulphate in rats exposed to the same levels as in the NTP studies (described below) cleared with a half-time of 2 to 3 days. In mice, 80-90% of the inhaled nickel sulphate cleared with a half-time within 5 to 17 days. Nieboer reported a half-life for water soluble Ni(II) salts in the blood stream of 24 hours or less [58].
Two year National Toxicology Program (NTP) inhalation studies of male and female F344/N rats and B6C3F 1 mice found varying strengths in the evidence of carcinogenic activity to two insoluble nickel compounds, nickel subsulphide and nickel oxide, but none at all to nickel sulphate hexahydrate, a soluble nickel compound [30][31][32]. A 2 year inhalation study of carcinogenicity in Wistar rats conducted by WIL Research Laboratories, Inc. showed that exposure to a lifetime dose of respirable sized metallic nickel powder did not cause cancer (Table 12) [35].
The findings in these animal studies raise important questions that are addressed in Hayes' classic textbook on toxicology [37]. Dose selection plays a key issue in the design and interpretation of the animal bioassay. Typical protocols call for animal exposures at the maximum tolerated dose (MTD) and at 2-3 additional dose levels at fractions of the MTD (e.g. 1/2, 1/4, etc.). The MTD is predicted from subchronic toxicity studies as the dose "that causes no more than a 10% weight decrement, as compared to the appropriate control groups, and does not produce mortality, clinical signs of toxicity or pathologic lesions (other than those related to a neoplastic response) that would be predicted [in the long-term bioassay] to shorten an animal's natural lifespan". The MTD is not a nontoxic dose and is expected to produce some level of acceptable toxicity to indicate that the animals were sufficiently challenged by the chemical. The MTD has been justified as a means of increasing the sensitivity of an animal bioassay involving limited numbers of animals so as to be able to predict risks in large numbers of humans. An objection to the use of MTDs has been that metabolic overloading may occur at high-dose levels, leading to an abnormal handling of the test compound; for example, toxic metabolites could be produced as a consequence of saturation of detoxification pathways. Organ toxicity could occur that might not happen at lower concentrations to which humans are typically exposed. Thus, it has been argued that nongenotoxic agents that are determined to be positive in rodent carcinogenicity bioassays may exert their own carcinogenicity via target-organ toxicity and subsequent cell proliferation and should not be assumed to be carcinogenic at low doses [37].
Ames and coworkers [38,39] have suggested that targetorgan toxicity and subsequent mitogenesis are responsible for the fact that over half of all chemicals tested in chronic bioassays at the MTD are determined to be carcinogens in rodents. They observed that both genotoxic and nongenotoxic agents tested at the MTD cause increased rates of mitogenesis, thus increasing the rate of mutation.
For several chemicals, induction of tumors was more strongly correlated with cell division than with DNA adducts or mutagenic activity. Others have reported that cancer potency and MTD are inversely correlated and that, consequently, the potency estimate is simply an artefact of the experimental design. Goodman and Wilson [40] found that cancer potency and the MTD were more strongly related for nonmutagens than for mutagens in rat bioassays, indicating that the carcinogenic effect and toxicity were more closely associated for nonmutagens than for mutagens; however, they noted that even for most mutagens, their findings suggested that at high doses carcinogenicity is induced via mechanisms associated with toxicity [37].
Gaylor [41] noted that, given sufficient animals (e.g. ~200 per group), it is estimated that about 92% of all chemicals tested would, if tested at the MTD, yield a positive response at one or more tumor sites in rats or mice. Gaylor observed that "this MTD bioassay screen is not distinguishing between true carcinogens and noncarcinogens." The author further suggests a common mechanistic explanation for this result; that is, for nongenotoxic carcinogens in particular, the mode of action involves cytotoxicity followed by regenerative hyperplasia. Thus, the relevant question is not so much whether a chemical causes cancer at the MTD (i.e., is a chemical a carcinogen?), but what is the dose at which the chemicals induce cancer [37]? We have drawn on this toxicology literature to highlight uncertainties around the interpretation of the findings in these animal bioassays (Table 13). Cytotoxicity at the target organ (lung), i.e., chronic active inflammation and/or macrophage hyperplasia, was observed in all animals (rats and mice) exposed at all levels to all nickel groups with only a few exceptions at the lowest nickel sulphate exposure levels for both species. Yet, alveolar/bronchiolar adenomas or carcinomas were found only in rats chronically exposed to sulphidic and oxidic nickel but not to nickel sulphate and metallic nickel. And they were found only in female mice exposed to oxidic nickel (a finding rated by the NTP as 'equivocal evidence'). Cell proliferation (macrophage hyperplasia) was found in both species exposed to nickel subsulphide and nickel sulphate but not to oxidic and metallic nickel. But it was associated with lung neoplasms found only in rats exposed to nickel subsulphide and nickel oxide (for which exposure female mice also showed carcinomas) but not to nickel sulphate and metallic nickel. While the NTP exercised prudence in their conclusions, we must ask in view of the opinions noted above the following questions: (1) Were the observed cancers caused by target-organ (lung) toxicity and subsequent cell proliferation in the face of MTD levels of exposure?
(2) Are these cancers likely to occur at low levels of human exposure; and (3) Were they caused by the chemical itself as carcinogen or by the dose at which the chemical(s) induce cancer? Table 13: Selected neoplastic and non-neoplastic lung effects in 2 year inhalation studies of male and female F344/N rats and B6C3F 1 mice exposed to nickel subsulphide, nickel oxide and nickel sulphate hexahydrate [30][31][32]; and Wistar rats exposed to nickel metal powder [35] F344/N rats B6C3F 1 mice Ni species, dose and lung effects

Male Female Male Female
The NTP soluble nickel report's authors noted in vitro evidence [36] that water-soluble nickel (i.e. nickel chloride) enhanced the cytotoxicity and mutagenicity of DNA-damaging agents by inhibiting nucleotide excision repair in mammalian cells and repair of ultraviolet-induced photoproducts [36]. They also cited ICNCM epidemiological findings [3] involving high exposure to nickel sulphate hexahydrate (1-5 mg/m 3 ) of refinery workers in the electrolysis department at Kristiansand, Norway and the hydrometallurgy department at Clydach, Wales as providing 'evidence that exposure to soluble nickel increased the risk of lung cancer in workers also exposed to oxidic, sulfidic, and/ or metallic nickel'. On the basis of the evidence to date, Oller et al. (2008) have concluded that the exact direct or indirect effects of Ni(II) ions needed for the generation of respiratory tumors are still the subject of much research. They suggest that the bioavailability of these ions at nuclear sites of target epithelial cells may determine the carcinogenic potential of Ni-containing substances. This bioavailability will depend on several factors: respiratory toxicity; deposition; clearance; target cell uptake; and intracellular dissolution (solubility) [35].

Concluding remarks
We have adopted a weight of scientific evidence standard [63] to examine the support for the soluble nickel cancer hypothesis; and have presented new findings and new analyses of existing findings of the human health and toxicological evidence for this compound that refute or seriously weaken the proposition. Sharp contrasts in the architecture, topography, industrial hygiene, intensity of use and histories of the KNR and PCNR plants point to the likelihood of mixed insoluble nickel exposures, including arsenic, as the most probable cause of the respiratory cancer risk observed in KNR's electrolysis department; and their absence in the same environment at PCNR as the likely reason for the normal risk observed there. We have shown that misclassification problems in KNR's epidemiology are the likely cause of implausible findings of elevated respiratory cancer risk in the plant's administrative and service areas that is comparable to the observed risk in its electrolysis and RSC departments. These may be related to the existence of a part-time or seasonal workforce unacknowledged and likely, therefore, unaddressed in KNR human health studies.
We have also identified unsupported changes to historical worker exposures in the most recent KNR epidemiological studies that cast serious doubt on the validity and reliability of inferences drawn from them. Unlike the protocol driven KNR JEM developed for the ICNCM (1990) report [3], the revised JEM for all post-1998 KNR epidemiological studies lacked a systematic rationale, thereby preventing review through sensitivity analyses of the validity and reliability of the JEM changes on overall and nickel species-specific risk exposure modeling estimates. We suggest, however, that the effect of the changes would have been to increase lung cancer unit dose risk estimates for all nickel species, and to transfer risk previously attributed to oxidic nickel to soluble nickel. We also demonstrated statistically that smoking and nickel exposures were strongly related in recent KNR respiratory cancer risk studies, making it impossible to draw valid inferences on carcinogenic risk from specific nickel compounds.
The long term (2 year) NTP animal inhalation studies of soluble nickel found no evidence of carcinogenic risk. Nor has in vivo toxicological evidence supporting a promotional carcinogenic effect been demonstrated. In concert, the evidence from the animal bioassays for all the nickel compounds has raised several questions: (1) Were the observed cancers caused by target-organ (lung) toxicity and subsequent cell proliferation in the face of MTD levels of exposure? (2) Are these cancers likely to occur at low levels of human exposure; and (3) Were they caused by the chemical itself as carcinogen or by the dose at which the chemical(s) induce cancer?
For all these reasons, therefore, we argue that, while KNR's epidemiology has determined the overall level of historical respiratory cancer risk in the refinery, it has failed to identify accurately its causes. We suggest that close scrutiny of the Clydach epidemiological database would lead to similar conclusions. Furthermore, the era of risk attributed to soluble nickel at the Welsh refinery extended from 1902 to 1937, an era not only when rudimentary industrial hygiene practices meant that mixed exposures throughout the workplace were far more likely and when smoking behaviour was embedded in the working culture through such inducements as free cigarettes for every WWI British soldier, but also when arsenic exposures were present and likely contributed to the lung and nasal cancer risks attributed to this workplace. There is consistent historical evidence across the Clydach, KNR and all Ontario nickel processing facilities of respiratory cancer risk in the likely, but systematically unrecorded, presence of arsenic exposures. Where none of these conditions applied (viz. PCNR's electrolysis department), no evidence to support the soluble nickel-cancer risk hypothesis was found.
We know that arsenic in the ore mined in the Sudbury basin has likely found its way into all related nickel processing plants. We also know that other sources of arsenic entered processing at Clydach. We suggest that arsenic was responsible for an indeterminate proportion of the respiratory carcinogenic risk previously ascribed to one or more nickel species in refinery studies. The animal bioassay evidence was developed to address the respiratory carcinogenic potential of each of the four nickel compounds found in nickel processing environments. However, the human health evidence involved exposures to a complex mixture of nickel species and, likely, to arsenic compounds (not to ignore the contributions of other offsite risky exposures and smoking as well), and to sulphuric acid mist exposures in the Harjavalta refinery. The animal studies provide evidence for pure substance exposure conditions never found in historical refineries; and, therefore, cannot directly support propositions on nickel carcinogenicity arising from human health studies. A similar problem was identified earlier that resulted from efforts to compare the respiratory cancer risks in the KNR and PCNR electrolytic departments. Since their environmental exposures were different, their epidemiology must have differed as well. Unless the animal studies could duplicate the complex mixture of exposures found in KNR's RSC or its electrolysis department, or in the PCNR's LC&S or its electrolysis department, it could not inform the related human health evidence. This is a fundamental problem with all observational studies, and is one argument in favour of randomized clinical trials (RCTs) for epidemiological evaluations. Obviously, RCTs cannot be used to develop historical environmental and occupational health evidence, but inferences drawn from those studies must be approached with great caution.
In the absence of human health and animal evidence supporting soluble nickel's carcinogenicity, we argue that this hypothesis lacks a sound scientific basis and should be reconsidered. At the very least, an independent review should be conducted of the KNR epidemiological database to locate the source(s) of respiratory cancer risk in the refinery, whether occupational or public health or both in nature. Secondly, we argue that appropriate regulatory agencies should reconsider their recommendations concerning this nickel compound. We also note in passing that our arguments raise fresh difficulties for regulatory toxicologists dealing with the development of occupational exposure standards for all nickel compounds, particularly for the still remaining carcinogens, sulphidic and oxidic nickel.
remained untraceable despite 'herculean' efforts by the researchers; and 44 had left the country of which partial information revealed that 13 were known to be alive and 4 dead. Roberts et al. reasoned that, if the mortality rate in the 4.4% who had left the country was similar to the unknown group as a whole, then one would expect record linkage to miss the corresponding 4.4% of deaths that occurred outside of Canada. The remaining 31 men were younger and had shorter durations of employment, on average, with likely higher rates of mobility therefore, making their trace more difficult but also rendering the assumption that their mortality rate was similar to that for the unknown group as a whole a conservative one.
With these tests, Roberts et al. reasoned that the record linkage procedure failed to detect 8% to 15% of the deaths in the unknown status group. Taking the most conservative figure, they estimated that the 2,455 deaths found by record linkage should really have been 2,455/0.85 or 2,888 deaths of which 433 would have been missed as a result. This represented a loss of 5.1% of all deaths [433/ 8,387] and they judged that a 95% ascertainment rate was methodologically acceptable by epidemiologic standards.

SAS ® code for the logistic model calculations in Andrews and Heller (2006) [44]
options nocenter nonotes linesize = 80; title1 'Logistic model analysis for Table three    proc logistic nosimple; model r/n = cell2 cell3 cell4 cell5 cell6; run; title2 'This step estimates logistic model Ni exposure odds ratios by smoking level'; proc logistic nosimple; model r/n = exp; by smo; run; title2 'The next two steps check the statistical significance of an exposure*smoking'; title3 'interaction term in the logistic model approximation'; title4 'The first run is an additive model with the interaction term'; proc logistic nosimple; model r/n = exp smo2 smo3 smoexpint; run; title4 'The second run is an additive model without the interaction term'; proc logistic nosimple; model r/n = exp smo2 smo3; run;

Appendix 3
The linear logistic model refers to the logit transform of the probability of disease from exposure (P), expressed as a linear function of regression variables (x) whose values correspond to the levels of exposure to the risk factors (total nickel exposure and smoking status). The model is defined by logit P(x) = log [P/(1-P)] = α + βx. Table seven of Grimsrud et al. (2002) [6] shows case and control counts for 6 cells defined by 2 levels of nickel exposure, < 0.75 or ≥ -.75 mg m-3 yr, (E) and 3 levels of smoking -Never/Former, Light/Medium and Heavy (S). Those counts were estimated using a conditional logistic regression model since each case in that study was matched to one or more controls. To duplicate the counts and corresponding odds ratios would require that study's complete dataset. Since the exposure variables are not finely stratified, however, it becomes possible to approximate the Table seven results with a logistic regression model. The SAS ® code in Appendix 2 applies a logistic model to obtain a close approximation to the original study's odds ratio estimates (Table ten). The code also estimates the statistical significance of a nickel exposure-smoking interaction term (E*S) when added to the logistic model approximation. The models with and without E*S had χ 2 likelihood ratios, respectively, of 99.1857 [4 degrees of freedom (df)] and 93.4049 [3 df]. The addition of the interaction term led to an increase in the model likelihood ratio of 5.5808, a result that is statistically significant at the 5% level since Pr {χ 2 > 5.5808} = 0.016 [1 df]. Note: SAS ® software is licensed by the SAS Institute Inc., Cary, NC.