(a) Lithography

Lithography is the most widespread of the printing processes and is probably used by more individual companies than all the other processes combined.

The litho process depends upon the ability to make the non-image areas of the planar printing plate hydrophilic, and thus readily wetted by water, and to make the image areas oleophilic, and thus readily wetted by oil-based inks. The plate is then wetted with a water-based fountain (damping) solution and inked with the oil-based ink. The development of the modern litho process has been made possible by the introduction of ‘offsetting’, in which the ink film is transferred from the printing plate onto a rubber blanket and then offset onto the substrate. This offsetting process prolongs plate life, particularly when printing on metal, and also improves the quality of print on substrates of poor smoothness.

The printing plate most commonly consists of a thin flat sheet of aluminium that has been treated to make the surface hydrophilic. Onto this is placed a light-sensitive polymer coating that is exposed to UV radiation through a film (which is positive or negative depending on the coating). The non-image areas are then developed away to expose the hydrophilic surface. Modern processing is usually in self-contained equipment that restricts any exposure of the operator to the developing solutions involved. Aqueous developers are replacing solvent-based systems. The metal plate is replaced by paper or polyester materials for some short-run small-format printing.

The paper or other substrate being printed can be fed into the press as individual sheets and delivered as a pile of sheets (sheet-fed litho) or as a continuous reel (web offset) that is converted to the finished publication after printing. Web offset printing can be further subdivided into cold-set web offset (with inks drying by penetration) and heatset web offset (with inks drying by evaporation in gas-heated ovens). The sheet-fed process is generally used for printing a wide variety of commercial/advertising print and cartons whereas web offset is used for newspapers and magazines.

The inks used are high-viscosity products based upon low-volatility resins and oils. The printing press is cleaned after use with solvent washes using volatile organic solvents, usually hydrocarbons, although there is now a move towards vegetable oil-based cleansers and even water-based cleansers (Searle, 1993).

A major segment of the lithographic printing industry is the cold-set web offset printing of newspapers using lower-viscosity inks that run on high-speed presses and are prone to misting, but less so than with rotary letterpress.

(b) Letterpress

Letterpress is a ‘relief’ printing process — that is, the image area is raised above the non-image area. The raised image area is inked before being brought into direct or indirect contact with the substrate to be printed. The more specialized applications that have developed over the last two decades have made greater use of offset letterpress, where the image is transferred via a rubber blanket.

Traditionally, the raised image was a lead alloy type or etched metal plates, but these have largely been replaced by photopolymer plates that are exposed to either negative or positive film and developed in dedicated equipment. Aqueous developers are replacing solvent-based systems.

Sheet-fed letterpress has declined to such a degree that it is of little more than historic interest today in industrialized countries. The same is true of the use of rotary letterpress in the printing of newspapers. The one sector in which letterpress is still of major importance is the printing of reel-fed self-adhesive labels, but even there its use is now in decline. The inks most widely used for rotary label printing are UV-cured, although there is still some use of heat-set inks. The inks used for letterpress printing are very similar, and sometimes identical, to those used for lithography.

(c) Flexography

Although the newest of the major print processes, flexography (or flexo) is now the most important process for the production of flexible packaging. A number of other applications include corrugated cases, tickets, forms, directories, paperback books, comics and newspapers.

The process is similar to letterpress in that the image area is raised above the non-image area. An inking roller is positioned in an ink duct and runs in contact with an engraved metering roller known as the ‘anilox’. This controls the amount (film weight) of ink that is transferred to the surface of the printing plate. At one time, the two processes were easily differentiated because letterpress used metal type plates and flexo used rubber stereos. Now both may use photopolymer plates. Although the distinction between letterpress and flexo is becoming blurred in some areas, this is not the case in the most important sector, namely flexible packaging, where the use of letterpress is insignificant. The dry ink film weight deposited in flexo is similar to that of litho and thus generally lower than that of letterpress.

The majority of flexo printing is web-fed with the main exception being the printing of corrugated cartons that are fed in as sheets.

The inks used for flexo are of low viscosity and are based upon a wide range of synthetic resins dissolved in low-boiling solvents. The drying process requires the removal of these solvents, usually by evaporation. However, there is a rapidly increasing interest in UV-cured (Atkinson, 1995) and water-based inks (Williams, 1994a^,b) in an effort to reduce the emissions of volatile organic compounds.

(d) Gravure

The printing image for gravure consists of recessed cells engraved in a metal cylinder. This image cylinder runs directly in the ink duct so that both the cells and the surface of the cylinder are flooded with ink. A reciprocating metal doctor blade held against the rotating cylinder scrapes the surplus ink from the non-image surface of the cylinder. The substrate, usually in web form, is held against the cylinder by an impression roller so that transfer of the ink takes place. The printed substrate then passes through a heated drying oven to remove the solvent by evaporation.

Until the mid 1980s, chemical etching of the copper cylinder through a gelatine image resist with ferric chloride solution was the usual process. Because of the difficulty in maintaining consistent quality and the potential hazard of ferric chloride, this has now been largely replaced by electromechanical or laser engraving.

The cells in the image vary in surface area according to the screen ruling being used. However, with a typical 60-lines/cm screening, the cells will be about 125 µm across and the depth will vary from 2 µm in the highlights to 40 µm or more in the shadows. The result is an ink film that varies in thickness according to the density of colour and can be up to three times as thick as that deposited by litho or flexo.

There are two sectors in the gravure printing industry: production of flexible packaging similar to that of flexo, and publication gravure that produces long-run magazines. With flexible packaging, gravure is generally preferred to flexo, since higher print quality is required and longer runs are used, and the high set-up costs can be recovered. In publication, gravure is again preferred to web offset for the long runs where it is more cost-effective.

The inks used are similar to those of flexo printing, although hydrocarbon solvents can be used as there are no rubber parts in the printing unit.

(e) Screen process

Unlike those used in other mainline printing processes, screen printing (sometimes called silk screen or screen process) inks do not have to transfer from an image surface to the substrate but actually pass through the image, which is the stencil on the screen.

The screen consists of a fine mesh, usually nylon or polyester, but sometimes stainless steel, stretched over a metal frame. The stencil is produced by coating the mesh with a photosensitive coating, exposing it to UV radiation through a positive film and developing away the unexposed image areas. The open areas of the screen around the stencil are blocked with screen filler, using volatile solvents such as xylene (see IARC, 1989a) or dichloromethane (see IARC, 1987a).

The inks used are intermediate in viscosity between those used for litho and those used for flexo or gravure. Evaporation, oxidation, chemical curing and UV curing methods are all used for drying in screen printing. As with the other printing processes, there is growing use of UV- and water-based inks (Pfirrmann, 1994).

Screen printing is the most versatile process and can cover much work that is impossible by other means — it can print on curved surfaces, even those that are not totally regular or symmetrical; it can deposit a heavy ink film that is very much more durable than those applied by other processes; and it can be used on a wide range of surfaces including fabrics, plastics, metals, glass and printed circuit laminates as well as on paper and board. Above all, origination costs and set-up times can be very low so that short-run work is economically viable. As a result, the importance of the process has grown steadily over recent years.

(a) Inorganic pigments

The only black pigment used to an appreciable extent is carbon black (Pigment Black 7), mainly furnace black and thermal black (see separate monograph, this volume). It is used in newsprinting and in publication, commercial and packaging printing, and is therefore used in large quantities. Different grades of carbon black, varying in particle size, surface condition and structure, are required depending on the application.

Titanium dioxide (Pigment White 6) (see IARC, 1989b) is by far the most commonly used opaque white pigment. Other opaque whites used in printing inks include zinc oxide, zinc sulfide and lithopones (mixtures of zinc sulfide and barium sulfate).

Transparent white pigments, in order of decreasing transparency, include alumina hydrate, magnesium carbonate, calcium carbonate (chalk), blanc fixe (precipitated barium sulfate), talc and clays. Transparent white pigments are also known as extenders, because they are used to reduce the colour strength of inks, as well as to modify ink dispersion and flow characteristics.

Inorganic colour pigments are generally complex mixtures of inorganic salts or minerals, manufactured to meet colour, rheological and other specifications for the particular application. In printing inks, some of these pigments include chrome yellow and chrome orange (mixtures of lead chromate (see IARC, 1990) and other lead compounds (see IARC, 1987b)), iron blue (a series of precipitated complex ferri-ferrocyanide compounds) and several iron oxide (see IARC, 1987c) and iron silicate pigments (Pigment Red 101 and 102, Pigment Brown 6 and 7). Cadmium (see IARC, 1993a) yellow, orange and red (selenides) and cadmium-mercury red have been used in printing applications where their high stability to light and alkali are advantageous, although their use has reportedly been almost completely eliminated except where extreme chemical resistance properties are required (National Association of Printing Ink Manufacturers, 1988).

Metallic powders (bronzes) also find some use in producing ‘silver’ (aluminium powder) and ‘gold’ (copper-zinc alloy powder) inks for packaging and advertising.

(b) Organic pigments

The printing ink industry uses more organic pigments than any other industry. Approximately 100 organic pigments are suitable for use in printing inks. Most of these pigments are prepared from azo, anthraquinone and triarylmethane dyes, phthalocyanines and vat dyes. Organic pigments are characterized by their high colouring strength, pure shades and transparency. However, as with the inorganic pigments, they are seldom pure chemicals and are classified by colour rather than by chemical composition. Organic pigments known as ‘lakes’ may be insoluble organic salts or may be produced by depositing organic dyes on one of the transparent inorganic white pigments, usually alumina hydrate.

Yellows are mainly diarylide yellows, Hansa yellows and lake yellows. These yellows are frequently used in place of the chrome yellows to avoid the use of lead and chromate compounds. Diarylide yellows are the principal yellow pigments currently used in printing inks. Pigment Yellow 12, one of the most common diarylide yellows, is a bis-azo dye based on 3,3′-dichlorobenzidine (see IARC, 1987d). Hansa yellows are also azo pigments based on toluidine and β-naphthol (Lewis, 1993a).

The most common orange pigment in printing inks is Pigment Orange 34, a diarylide orange. Others include dianisidine orange and Persian orange lakes.

There are more red pigments used in inks than any other organic colour. Some of the more common are the para reds (lakes based on the azo dye from para-nitroaniline and β-naphthol), lithol reds (e.g. Pigment Red 57:1, a calcium salt of an azo dye, which is the most important magenta colourant in inks), rhodamine reds (e.g. Pigment Red 81, which is rhodamine 6G laked by precipitation with phosphotungstic and phosphomolybdic acids) and Red Lake C (Pigment Red 53:1 (see IARC, 1993b), which is the barium salt of the azo dye obtained from 2-amino-4-methyl-5-chlorobenzenesulfonic acid and β-naphthol).

Phthalocyanine blues, such as Pigment Blue 15, are the most common organic blue pigments in inks. Peacock blues (Lewis, 1993b) (salts of certain sulfonated triarylmethane dyes, sometimes laked on alumina hydrate) were used in the past. Other organic blue pigments include the victoria blues and the alkali blues.

Phthalocyanines also form green pigments (for example, Pigment Green 7, a chlorinated phthalocyanine). Triarylmethane dyes, the salts of which are used as pigments, include malachite green and methyl violet. Pigment Violet 19, a quinacridone structure, is an example of the anthraquinone-type pigments used in printing inks.

(a) Lithographic (offset) inks

The broad spectrum of applications for lithography requires a wide variety of inks to serve the needs of the offset printing industry (National Association of Printing Ink Manufacturers, 1988; Kübler, 1993; Bassemir et al., 1995; Pardoen, 1995).

Newspaper offset (cold-set) printing inks are typically very simple. Blacks are carbon black in a high-boiling mineral oil with asphaltic material (bitumen (see IARC, 1987e), gilsonite). Such inks could consist of up to 70% mineral oil (see IARC, 1987f) which, in the past, might have contained up to 15% aromatic hydrocarbons. During the last decade, these mineral oils have been replaced with grades that have about 5% aromatics, of which less than 0.1% are polycyclic aromatic hydrocarbons (PAHs) (see IARC, 1983). Coloured inks generally have a soya bean oil vehicle instead of mineral oil For recycling of newsprint, the ink must be removable and therefore these inks do not contain binders that undergo significant oxidative cross-linking.

Heat-set web offset inks are designed to produce high-gloss printed images (magazines, books). They contain lower-boiling mineral oils that are removed (within 1 sec) as the printed roll (web) passes through a hot air oven. A typical formulation might be: organic pigments (15–25 wt %), hard resins (25–35 wt %), soft resins and drying oils (5–15 wt %), mineral oil (b.p. 240–260 °C; 25–40 wt %) and additives (5–10 wt %) (Williams, 1992; Kübler, 1993).

Sheet-fed offset inks are used in commercial litho presses for printing, for example, advertising brochures, business papers and packaging, on individual sheets rather than long rolls. Inks are based on phenolic or maleic acid-modified rosin ester and alkyd resins in vegetable drying oils (linseed, soya, tung) diluted with mineral oil. Inks dry by quick-setting, i.e. by absorption and oxidation. A typical formulation would be: organic pigments (12–20 wt %), hard resins (20–25 wt %), soft resins and drying oils (20–30 wt %), mineral oil (b.p. 250–300 °C; 20–30 wt %) and additives (5–10 wt %) (Kübler, 1993).

Radiation-cured offset inks, which are based on acrylate or vinyl/ether monomers, are becoming very important in both sheet-fed and web offset processes. The printed substrate is exposed to UV radiation or an electron beam at the end of the press, and the ink sets within a fraction of a second.

To exclude the oil-based ink from the hydrophilic areas of the printing plate, all offset litho printing processes require water-based fountain or dampening solutions. These solutions are typically slightly acidic aqueous solutions (pH 3.5–5.5) containing small amounts of buffers, alcohols, surfactants, hydrophilic polymers (gum arabic or cellulose derivatives), complexing agents (EDTA, ethylenediamine tetraacetic acid) and preservatives.

(b) Letterpress inks

While letterpress is being replaced by other printing processes, it is still used to a limited extent to produce newspapers, magazines, self-adhesive labels, packaging and other printed products (National Association of Printing Ink Manufacturers, 1988; Williams, 1992; Kübler, 1993; Bassemir et al., 1995).

Letterpress news ink is similar to web offset inks used to print newspapers.

Moisture-set inks have been used for food packaging printing and contain maleic or fumaric acid-modified rosin products or modified phenolics as binders in glycol solvents. The printed surface is treated with steam or a fine mist of water, and the water-insoluble acidic binders precipitate, setting the ink.

Water-miscible inks maintain the stability of the ink through an organic base that evaporates or is neutralized to induce drying.

A variety of other ink types have been used in letterpress printing, including heat-set, quick-set, water-washable and high-gloss inks (National Association of Printing Ink Manufacturers, 1988).

(c) Flexographic inks

Flexographic inks are liquid inks, rather than pastes, and are designed to dry quickly primarily by evaporation. Both solvent- and water-based ink systems are used extensively in flexography.

Common solvents in the solvent-based inks include the lower alcohols (ethyl, n-propyl, isopropyl (see IARC, 1987g)), usually mixed with esters and sometimes small amounts of higher glycol ethers or aliphatic hydrocarbons to obtain optimum resin solubility, viscosity and drying speed. A wide variety of resins are used in solvent-based flexo inks, such as nitrocellulose, polyamides, cellulose esters, acrylics and various modified rosins. Although pigments are the most common colourants, there is some use of both basic and metal-complex dyes in flexographic inks. Because much of the flexographic printing is on non-absorbent flexible packaging (polyethylene (see IARC, 1987h), polypropylene (see IARC, 1987i), poly(vinyl chloride) (see IARC, 1987j)), ink additives may include plasticizers to promote formation of a flexible ink film and waxes to add rub resistance (National Association of Printing Ink Manufacturers, 1988; Kübler, 1993; Bassemir et al., 1995).

For environmental reasons, the solvents in ink are being increasingly replaced by water. Approximately 50% of all flexographic inks have water as their primary solvent (Bassemir et al., 1995). Resins in these water-based formulations are generally acidic acrylates or fumaric acid-modified rosin or shellac, neutralized with ammonia or volatile amines, which evaporate from the printed substrate and thereby set the ink film. A typical water-based flexographic ink for paper or paperboard contains: organic pigments (12–15%), resins (10–25%) and additives (5–7%); the remainder is water. For printing on plastics, the ink usually contains a small amount of alcohol (2–5%) and more additives (6–10%) (Kübler, 1993).

Water-based flexographic printing also finds some application in newspaper printing, but it still represents only a small segment of the market (6–7% in the United States in 1992) (Bassemir et al., 1995), and removal of the ink from newsprint before recycling the paper remains a problem.

UV-cured inks are also beginning to be used in flexographic printing. Their composition is similar to the UV-cured offset litho inks although they are less viscous (Kübler, 1993; Atkinson, 1995).

(d) Gravure inks

Gravure inks are similar to flexographic inks except that ketones and aromatic hydrocarbons can be used as solvents, providing a much greater latitude in the selection of binders (Bassemir et al., 1995). In the United States, gravure inks are divided into 10 categories according to the type of binder or solvent. For example, aliphatic hydrocarbons (hexane, VM&P (varnish makers' and painters') naphtha, mineral spirits) are used mainly in type A, B and D inks; aromatic hydrocarbons (toluene (see IARC, 1989c), xylene) are used in types B, D, M and T; and ketones (acetone, methyl ethyl ketone) and esters (ethyl, isopropyl, n-propyl and butyl acetates) are required for type C inks (Bassemir et al., 1995).

Gravure inks also are classified according to the printed product. ‘Publication gravure’ (magazines, catalogues) utilizes hard resins dissolved in toluene and/or aliphatic solvents. Resins include maleic acid-modified rosin and phenolic resins, calcium and zinc resinates, hydrocarbon resins and others. Polyethylene-based waxes are often added to improve abrasion resistance. A typical publication gravure ink might contain: pigments (8–15 wt %), resins (15–20 wt %), solvent (60–70 wt %) and additives (0.5–5 wt %) (Kübler, 1993).

‘Packaging gravure’ does not use hydrocarbon solvents but rather uses esters and alcohols (type C inks in the United States nomenclature). For various packaging substrates, resins may include cellulose nitrate, maleic resins, acrylate resins, polyurethane resins and polyamide resins, or mixed polymers of vinyl chloride/vinyl acetate/vinyl alcohol. Plasticizers (phthalates, citrates, adipates) also may be required, especially with cellulose nitrate. Basic dyes are used occasionally, in addition to pigments, in gravure inks. More recently, water-based gravure inks (type W), with formulations very similar to the water-based flexographic inks, are finding increasing use in packaging gravure (National Association of Printing Ink Manufacturers, 1988; Kübler, 1993; Bassemir et al., 1995).

A special application of gravure (as well as other processes) is in printing with transfer inks. Aqueous inks containing selected textile-disperse dyes are printed and dried on special papers. The printed image can then be transferred by sublimation to the textile materials (e.g. polyester fabric) by pressing at approximately 200 °C (Kübler, 1993).

Intaglio is another specialized process using a higher-viscosity ink in which a high-quality image is engraved on a steel plate. Inks often contain special high-durability pigments, fillers to increase viscosity, drying oil vehicle and a number of additives. This process is used in printing paper currency (bank-notes), postage stamps, stock certificates and similar products (Kübler, 1993; Bassemir et al., 1995).

(e) Screen process inks

Screen printing is a highly versatile process that can apply a thicker film of ink to the substrate than other printing processes. A very wide range of ink formulations is available, depending on the substrate and the requirements for the printed product. Drying may be by evaporation, oxidation, radiation-curing or other processes. Any of the resin types found in litho, flexo and gravure inks may be used, and the solvents can be of almost any type as long as they evaporate at a suitable rate, which is slower than that required for flexo and gravure. Solvents (propylene glycol ethers, aromatic and aliphatic hydrocarbons and cyclohexanone) typically have somewhat higher boiling-points than those used in gravure printing, and inks are more viscous (Kübler, 1993; Leach & Pierce, 1993; Bassemir et al., 1995).

Although newly introduced product ranges are usually lead-free, screen printing is the one process that still makes significant use of lead chromate pigments. The use of N-vinyl pyrrolidone (see IARC, 1987k) in UV-cured inks has declined.

(a) Cancer at multiple sites

Cancer cases (17 sites) and patients with non-neoplastic diseases admitted to the Roswell Park Memorial Institute in Buffalo, New York, United States, from 1956 to 1965 were compared with regard to occupations associated with either inhalation of combustion products or chemicals (Viadana et al., 1976). Information on lifetime occupational history, educational level and smoking habits was collected at the time of hospital admission. A total of 11 591 white male patients (cancer cases and non-cancer cases) were included in the analysis. Workers ever employed in specific occupations were compared with those in an unexposed clerical group. Eleven cases of cancer of the buccal cavity and pharynx among printers yielded a statistically significant age-adjusted odds ratio of 2.6 (p < 0.05), while the odds ratio adjusted for smoking was 2.1 (p > 0.05). A nonsignificant (p > 0.05) increased age-adjusted odds ratio of 1.5 was found for lung cancer (7 cases).

As part of the United States Third National Cancer Survey, 7518 incident cancer cases, representing a 57% response rate, were interviewed on lifestyle factors and main lifetime industry and occupation (Williams et al., 1977). Respondents and non-respondents were comparable in age, race, sex, marital status, method of diagnosis, country of birth and cancer site distributions. The proportions of specific main-lifetime industries and occupations among patients with cancer at one site were compared (separately for men and women) to those of patients having cancer at other sites combined, controlling for age, race, education, tobacco use and alcohol consumption. The total numbers of cancer cases in printing, publishing and allied products industries were 59 among men and 33 among women. The only significantly increased relative risk was found for cancer of the oral cavity in males (odds ratio, 4.5; p < 0.05; 7 cases). Nonsignificant excesses were observed for cancer of the pancreas both in men (odds ratio, 7.0; 3 cases) and women (odds ratio, 8.2; 2 cases). Among men, there were four urinary bladder cancer cases (odds ratio, 2.0) and two cases of melanoma (odds ratio, 3.0). [The Working Group noted the low response and the resultant possibility of bias.]

In a population-based case–control study of cancer at 19 sites (Siemiatycki, 1991), described in detail in the monograph on Carbon Black in this volume, pp. 188–189, analyses were carried out relating to the printing and publishing industry, employment as a printer and employment as a printing press worker. For the printing and publishing industry, the odds ratio for lung cancer was 2.0 ([95% CI, 1.2–3.5]; 35 cases); no significant excess was noted for any other site. In printers, significant excesses were found for lung cancer (odds ratio for any exposure, 2.1 [95% CI, 1.1–4.1]; 26 cases; odds ratio for over 10 years' exposure, 1.7 [95% CI, 0.7–4.1]; 13 cases) and pancreatic cancer (odds ratio for over 10 years' exposure, 3.7 [95% CI, 1.2–11.2]; 4 cases); for French printers, a significant excess for renal cancer was observed (odds ratio for any exposure, 34 [95% CI, 1.2–9.5]; 5 cases; odds ratio for over 10 years' exposure, 3.6 [95% CI, 1.0–10.6]; 3 cases). In the more narrowly defined category of printing press workers, the odds ratio for lung cancer was 3.1 ([95% CI, 1.1–8.7]; 15 cases), and, when this association was further analysed by histological type of lung cancer, the odds ratio for adenocarcinoma of the lung was 7.0 ([95% CI, 1.8–27.9]; 6 cases).

One of the substances assessed by the occupational hygienists in the Montréal (Canada) study (Siemiatycki, 1991) was inks, including printing inks. Nearly 4% of the total study sample exposure was attributed to inks in one or another of their jobs. The most common job titles in which it was attributed were: typesetters and printers, draftsmen and business machine repairmen. In analyses using cancer controls adjusting for age, smoking, ethnic groups and socioeconomic status, the following sites of cancer were not associated with exposure to inks: oesophagus (odds ratio, 0.9; 3 cases), stomach (odds ratio, 0.9; 7 cases), colon (odds ratio, 0.5; 9 cases), rectum (odds ratio, 1.2; 12 cases), pancreas (odds ratio, 1.3; 5 cases), prostate (odds ratio, 1.3; 13 cases), urinary bladder (odds ratio, 1.2 [95% CI, 0.7–2.1]; 18 cases), melanoma of the skin (odds ratio, 1.0; 5 cases) and lymphoma (odds ratio, 0.8; 6 cases). For lung cancer, there was an indication of excess among all those considered to be exposed (odds ratio, 1.6 [95% CI, 1.0–2.7]; 37 cases) but no greater risk among the subset considered to be substantially exposed (odds ratio, 1.5 [95% CI, 0.7–3.1]; 18 cases). For renal cancer, there was an indication of excess risk for those considered to be substantially exposed (odds ratio, 2.5 [95% CI, 1.0–5.4]; 7 cases).

In a reanalysis of the above data focusing on urinary bladder cancer (Siemiatycki et al. 1994), having worked in the printing and publishing industry for less than 10 years yielded an odds ratio of 0.3 (95% CI, 0.1–1.2; 2 cases), while for 10 or more years in the industry, the odds ratio was 1.9 (95% CI, 0.9–3.9; 11 cases). An elevated odds ratio of 3.0 (95% CI, 0.9–10.1; 4 cases) was found for substantial exposure to photographic products among photoengravers and photographic processors.

(b) Urinary bladder cancer (see Table 9)

Wynder et al. (1963) examined occupation and other risk factors associated with urinary bladder cancer risk in 300 male and 70 female cases and an equal number of controls from seven New York hospitals in 1957–61. Controls, matched by age and sex, were selected among patients of the same hospitals excluding subjects with cancers of the respiratory system and of the upper alimentary tract and those with myocardial infarction. Information on possible risk factors was collected through interviews. Three cases had ever worked as printers, engravers or lithographers [crude odds ratio, 2.2; 95% CI, 0.4–13.2].

A case–control study was conducted in eastern Massachusetts, United States (Cole et al., 1972), on 510 male and female histologically confirmed cancers of the lower urinary tract (94% were urinary bladder cancer) diagnosed between January 1967 and June 1968 (Cole et al., 1971). Each case was matched by sex and age to a general population control; 92% of the cases and 91% of the controls were interviewed and usable occupational history was collected for 461 cases (356 men and 105 women) and 485 controls (374 men and 111 women). The age-adjusted odds ratio for men whose ‘usual’ occupation was printing was 0.8 (95% CI, 0.3–1.9; 5 cases). Men ever employed in printing had an age- and smoking-adjusted odds ratio of 1.3 (95% CI, 0.7–2.5; 14 cases).

Sixty-five male and 10 female white patients with histologically confirmed urinary bladder cancer, admitted in 1978 to two community hospitals in northern New Jersey, United States, and 142 (123 men, 19 women) controls matched by age, race, place of birth, sex, hospital and residence were investigated by Najem et al. (1982). Controls were selected from patients treated for other conditions in the same hospitals, excluding those with a history of neoplasms or tobacco-related heart diseases. Lifetime occupational histories and a variety of other personal factors were collected through interviews. Among subjects employed in the printing industry for at least one year, the crude odds ratio was 2.7 (95% CI, 0.8–9.6; 7 cases). [The Working Group noted that these were prevalent cases and that this was a hospital-based series.]

Cartwright (1982) conducted a case–control study on urinary bladder cancer in three hospital districts of West Yorkshire, United Kingdom. A total of 625 prevalent cases were identified in October 1978 (472 men, 153 women) and 366 incident cases (272 men, 94 women) were identified between 1978 and 1980. Prevalent cases were matched to one hospital control and incident cases with two controls. Controls were selected randomly among patients without malignant diseases and matched by age, sex and health district. Information on a variety of personal habits and occupational history was derived from interviews. A significant odds ratio, adjusted by sex and type of case (incident or prevalent), of 3.1 (95% CI, 1.4–6.8; 18 cases) was detected for subjects working for at least six months as printers (defined as those exposed to ink-fly from high-speed presses).

Four papers are available giving results on urinary bladder cancer risks in printers based on various subsets of the United States National Bladder Cancer Study (Silverman et al., 1983; Schoenberg et al., 1984; Silverman et al., 1989, 1990). This population-based case–control study was carried out in 10 areas of the United States during 1977–78. Age frequency-matched population controls were selected for each geographical area. Employment was considered as ‘ever’ or ‘usual’ in each occupation or industry; subjects never employed in the industry served as referents.

In the Detroit, MI, United States, area, out of 420 male subjects with carcinomas, or papillomas not specified as benign, of the lower urinary tract (95% being urinary bladder cancer) were identified and 339 (81%) were interviewed (Silverman et al., 1983). Analysis was restricted to 303 white male cases and 296 white male population controls. Subjects ever employed in the printing industry did not show an increased risk (crude odds ratio, 1.1; 95% CI, 0.7–1.7; 50 cases). A nonsignificant elevation was seen for subjects who ever worked as printers (odds ratio, 3.0; 95% CI, 0.6–14.8; 6 cases).

In the New Jersey area, United States, 658 incident male cases of urinary bladder cancer and 1258 general population male controls were studied by Schoenberg et al. (1984). Age- and smoking-adjusted odds ratios were presented for each employment/-material category in the multiple regression model. The findings suggested no increased risk for men ever employed in the printing industry (odds ratio, 0.9; 95% CI, 0.5–1.5; 20 cases). However, self-reported exposure to printing inks yielded an odds ratio of 1.6 (95% CI, 1.0–2.5; 42 cases).

In the latter two of this group of four papers, Silverman et al. (1989) reported on 2100 white male cases with urinary bladder cancer and 3874 population controls. The smoking-adjusted odds ratio was 0.8 for white men who had ever worked as printers (95% CI, 0.5–1.2; 37 cases). There were 652 cases of urinary bladder cancer and 1266 controls among white women (Silverman et al., 1990). The smoking-adjusted odds ratio for women who had ever worked as printers was 0.2 (95% CI, < 0.1–1.4; 1 case).

A death certificate-based case–control study included 291 urinary bladder cancer deaths occurring in England and Wales in the period 1975–79 among males under the age of 50 years (Coggon et al., 1984). For each case, two controls were selected among men who had died from any other causes, matched by sex, year of birth, year of death and residence. Based on occupation recorded on the death certificate, exposure to four known and five putative carcinogens was evaluated and graded (high, low, none), applying a job–exposure matrix. No increased risk was found for those occupations involving exposure to printing inks (odds ratio, 1.1; 95% CI, 0.5–2,3; 12 cases). However, the risk was elevated for occupations with high exposure to printing inks (odds ratio, 5.0; 95% CI, 1.0–25.8; 5 cases). [The Working Group noted the young age of the study subjects and that occupational status was derived from death certificates.]

Vineis and Magnani (1985) conducted a hospital-based study of 512 male cases of urinary bladder cancer diagnosed between 1978 and 1983, and 596 male hospital controls, matched by age, having benign urological (other than haematuria, cystitis and benign polyps of the urinary tract; Vineis et al., 1984) and surgical conditions. All cases and controls lived in the province of Turin, Italy. Information on lifelong smoking habits and a complete occupational history were collected through interviews. The age-adjusted odds ratio for subjects ever employed in the broad category of the printing and publishing industry for at least six months was 1.8 (95% CI, 0.8–4.0; 17 cases). Adjustment for smoking (and not for age) resulted in a significant odds ratio of 1.7. When incident cases only were considered, the age-adjusted odds ratio was 1.2 (95% CI, 0.5–3.3) [number of incident cases not given].

Baxter and McDowall (1986) examined the association between occupation and urinary bladder cancer risk using death certificate information in the six London boroughs that had the highest mortality from this disease in England and Wales. Cases comprised 1080 deaths from urinary bladder cancer between 1968 and 1978 among male residents of the six boroughs. Two controls for each case were chosen: one selected randomly from male deaths from all other cancers and another from male deaths from all causes. Both were matched to the cases by residence, year of death and age. Odds ratios for printers were 1.5 (controls from all other cancers) and 1.2 (controls from all other causes including cancers) (p > 0.05; 21 cases). [The Working Group noted that the number of exposed controls was not given, that the method used to calculate the odds ratios was not clearly stated, and that occupational status was derived from death certificates.]

Using data derived from a large case–control study of smoking and cancer among hospitalized patients in six cities in the United States, Kabat et al. (1986) examined the potential risk factors for urinary bladder cancer for reported lifetime nonsmokers in a series of 76 male and 76 female cases and 238 male and 254 female controls. Controls, matched by age, sex, race, hospital and year of interview, were selected among patients with non-tobacco-related cancers (67% in men, 59% in women) or non-neoplastic diseases (33% in men, 41% in women). All subjects were interviewed between 1976 and 1983 and the usual (longest-held) occupation recorded. The study failed to detect an increased risk for male printers: two cases and six controls yielded a nonsignificant odds ratio of 1.1. [The Working Group noted the small number of exposed cases in this study.]

The Missouri Cancer Registry has been collecting data on cancer cases since 1972. Data on smoking, alcohol consumption and occupation (the longest-held job) are routinely derived from a standardized form administered to each cancer patient. Brownson et al. (1987) selected 823 histologically confirmed urinary bladder cancers diagnosed among white males between 1984 and 1986. For each case, three white male controls, frequency matched by age, were chosen among other patients in the registry, excluding smoking-related cancers. Analysis by occupational history was performed using low-risk employment categories as reference. The low-risk occupations included professionals, managers and sales and clerical workers. Printing machine operators had a statistically significant age-, smoking- and alcohol-adjusted odds ratio of 3.1 (95% CI, 1.1–8.9; 7 cases).

Iscovich et al. (1987) reported on a case–control study of 117 histologically confirmed urinary bladder cancer cases diagnosed between 1983 and 1985 in the La Plata area, Argentina. Cases were individually matched by age and sex to one neighbourhood control and one non-neoplastic hospital control. Information on smoking, demographic, socioeconomic and medical variables and occupational history for the three occupations of longest duration and the most recent occupation were collected from subjects by questionnaire. Three cases and two controls had ever been employed in printing industries, which yielded a nonsignificant age- and smoking-adjusted odds ratio of 2.7. Seven subjects (5 cases, 2 controls) working as printers either in the printing industry or elsewhere gave a statistically significant age-adjusted odds ratio of 5.4; after adjustment for smoking, the odds ratio was 5.6 but no longer significant.

As part of a methodological investigation of data sources for occupational studies, Steenland et al. (1987) conducted a case–control study of male urinary bladder cancer deaths in Hamilton County, OH, United States. A total of 731 urinary bladder cancer cases, deceased during 1960–82, were matched on age, sex, race, date of death and residence at death to two controls per case. Controls were chosen from among all other deaths with the exclusion of deaths from urinary tract tumours and pneumonia. Commercial city directories report yearly occupation and employer for all residents in the county over the age of 18, making it possible to reconstruct occupational lifetime histories. Occupational data from city directories were available for 648 cases and 1275 controls. Odds ratios, conditional on matched sets, were calculated by occupation, industry and employer. The odds ratio for having ever been employed as a press machine operator was 5.0 ([95% CI, 1.3–19]; 5 cases). For employment over 20 years in the printing industry, the odds ratio was 1.3 (p > 0.05; 12 cases).

A case–control study of occupation and urinary bladder cancer was carried out by González et al. (1989) in five provinces of Spain. Cases of histologically confirmed urinary bladder cancer were recruited from the registers of 12 hospitals between 1985 and 1986. A total of 497 (438 men and 59 women) cases, alive at the time of interview, resident in the province where the hospital was located and below the age of 79, were interviewed (71 were not traced or refused interview). The proportion of incident cases was 51%. For each case, two controls, matched by sex and age, were selected. The first was selected from the same hospital as the case, excluding patients with diagnoses possibly associated with the risk factors under study. Among hospital controls, 583 (62 women) were interviewed; 159 were not traced or refused interview. A second control was randomly selected from the same section of the census or municipality register as that of the case. Of the population controls, 530 (65 women) were interviewed; 200 were not traced or refused interview. A questionnaire on lifestyle and occupational history was administered to each subject. The age- and sex-adjusted odds ratio was 2.1 for male printers (95% CI, 1.0–43; 15 cases). The odds ratio was slightly lower (1.8; 95% CI, 0.8–4.1) after further adjustment for tobacco consumption and employment in other high-risk occupations. In the analysis by occupational subgroups, typesetters and linotypists were associated with an odds ratio of 2.0 (95% CI, 0.5–7.5; 9 cases) after adjusting for smoking and employment in other high-risk occupations.

Using data derived from an on-going case–control study conducted in the area of Milan, Italy, La Vecchia et al. (1990) examined the relationship between occupation and urinary bladder cancer. The case series included 263 (219 men, 44 women) histologically confirmed invasive urinary bladder cancers diagnosed between 1985 and 1988. The controls were 287 patients (210 men, 77 women) with acute medical and surgical conditions and non-neoplastic and non-urinary tract diseases. The proportion of cases ever employed in the printing industry (2.3%) was similar to that among controls (2.4%) [crude odds ratio, 0.9; 95% CI, 0.3–2.8].

Steineck et al. (1990) examined the relationship between urothelial cancer and exposure to benzene and exhaust in a population-based case–control study in Sweden. The study was based on men born between 1911 and 1945 and living in the county of Stockholm for all or part of the observation period of September 1985 to November 1987. Incident cases of urothelial cancers and/or squamous-cell carcinoma in the lower urinary tract were identified from the regional Cancer Registry and the urological departments (320 subjects). Controls (363 subjects) were selected by stratified (gender and year of birth) random sampling during the observation period from a computerized register covering the population of Stockholm. Information on exposure was collected by a postal questionnaire and all subjects were contacted at their homes. An industrial hygienist classified the subjects as having been exposed or unexposed to 38 agents and groups of substances. The adjusted odds ratio among men classified as having been exposed to carbon black (including printing inks) was 2.0 (95% CI, 0.8–4.9; 14 cases). The odds ratios were 3.2 (95% CI, 0.4–27.1) for low exposure to printing inks, 0,5 (95% CI, 0.1–2.1) for moderate exposure and 3.6 (95% CI, 0.8–12.1) for high exposure.

A population-based case–control study was conducted in the United States to examine the relationship between urinary bladder cancer, usual occupation and industry, and cigarette smoking (Burns & Swanson, 1991a). Cases and controls were selected through the Metropolitan Detroit Cancer Surveillance System, which had collected population-based cancer data for the three-county metropolitan Detroit area since 1973. Incident cases occurring among white and black subjects of each sex between the ages 40 and 84 were enrolled into the study. Subjects, or their surrogate respondents, were interviewed by telephone; the response rate for cases and controls was 94% and 95%, respectively. A total of 2160 urinary bladder cancer cases and 3979 controls with cancers of the colon and rectum, with complete histories of occupation and tobacco use, were included in the analysis. The analysis by usual occupation did not reveal an increased risk for printers (adjusted odds ratio, 0.9; 95% CI, 0.4–1.9; 12 cases). Nor was there an excess risk found for the printing industry (adjusted odds ratio, 0.7; 95% CI, 0.4–1.2; 22 cases). Case series in this study partly overlapped with the United States National Bladder Cancer Study (see Silverman et al., 1983, 1989, 1990). [The Working Group was concerned that only colonic and rectal cancers were used as controls.]

A hospital-based case–control study of 531 male matched pairs was conducted in South Lower Saxony, Germany (Claude et al., 1988; Kunze et al., 1992). Of the cases, 93% had histologically confirmed benign or malignant epithelial tumours of the urinary bladder; the remainder had tumours of the lower urinary tract. The cases were ascertained between 1977 and 1985 from three hospitals. Controls were patients with non-neoplastic diseases of the lower urinary tract from the same hospitals, individually matched by age to the cases. Among controls, 64% had hyperplasia of the prostate and 19% infection of the lower urinary tract. All subjects were interviewed on lifestyle, familial occurrence of urinary bladder cancer and occupational history. Cases and controls were compared according to whether they were ever employed in specific industries or occupations or ever exposed to selected agents. A significantly increased crude odds ratio of 5.0 (95% CI, 1.3–19.6; 11 cases) was observed among men ever employed in the printing industry. Analysis by occupation showed an elevated, although statistically nonsignificant, risk for printing workers (odds ratio, 3.0; 95% CI, 0.7–13.8; 7 cases). Analysis by length of employment was not performed due to the small numbers. [The Working Group noted that the definition of printing workers was unclear.]

Urinary bladder cancer and occupational exposures were investigated in a multicentre hospital-based case–control study in France between 1984 and 1987 (Cordier et al., 1993). Incident and prevalent cases, diagnosed after 1982, of histologically verified urinary bladder cancers in patients under 80 years of age were collected in urology departments of seven university hospitals in five regions of France. There was a total of 658 male cases. For each case, one hospital control was selected; controls were matched for age, ethnic origin and place of residence and had been admitted to other departments of the same hospital for diseases other than cancer, respiratory diseases or symptoms suggestive of urinary bladder cancer. Detailed interview data were collected for jobs held for at least six months. Occupational industrial titles were classified by a team of industrial hygiene experts. An unconditional logistic regression model was employed, controlling for age, hospital, place of residence and smoking status. The adjusted odds ratio for men ever employed in the printing and publishing industry for at least six months was 0.9 (95% CI, 0.5–1.5; 26 cases). The odds ratio for men who, based on job title, were classified as printers was 1.5 (95% CI, 0.6–3.5; 14 cases); the subgroup classified as printing pressmen had an odds ratio of 4.2 (95% CI, 0.8–20.8; 8 cases).

(c) Lung cancer (see Table 10)

Coggon et al. (1984), in a case–control study described on pp. 73–74, examined the relationship between lung cancer and occupation, based on information derived from death certificates. The study included 598 lung cancer deaths occurring in the period 1975–79 among males under the age of 40 in England and Wales. For each case, two controls were selected among men who had died from any other causes (matched by year of birth, year of death and residence). Information on occupation was obtained as described previously. A nonsignificant excess was found in occupations involving exposure to printing inks (odds ratio, 1.6; 95% CI, 0.9–2.7; 28 cases). The odds ratio for occupations with high exposure was 2.0 (95% CI, 0.8–5.0; 9 cases). [The Working Group noted the young age of the study subjects and that occupational status was derived from death certificates.]

Occupational risk factors for lung cancer were investigated in a case–control study in France (Benhamou et al., 1985, 1988). The data were collected during the period 1976–80 in 16 hospitals, 13 of which were in Paris. A total of 1260 male cases with histoogically confirmed lung cancer and 2084 matched (by age at diagnosis, hospital of admission, and interviewer) controls hospitalized for non-tobacco-related diseases were included. Complete occupational histories were recorded in personal interviews. The smoking-adjusted matched odds ratio for printers and related workers was 1.2 (95% CI, 0.7–1.9; 32 cases). There was no evidence of an increase in risk with duration of exposure.

A case–control study of lung cancer in Florence, Italy, was conducted to investigate occupational risk factors (Buiatti et al., 1985). All histologically confirmed cases of primary lung cancer admitted during 1981–83 to the regional general hospital (referral centre for all lung cancers) were selected for the study. For each case, one or two controls were selected from the same hospital by sex, age (plus or minus five years), date of admission and smoking status. All cases and controls were interviewed after their first admission to the hospital; in total, 592 cases (547 men, 45 women) and 1036 controls (955 men, 81 women) were interviewed. After restricting both series to residents of metropolitan Florence, 376 cases (340 men, 36 women) and 892 controls (817 men, 75 women) were available for the analysis. The odds ratio (adjusted for age, smoking and place of birth) for men having ever worked in paper or printing occupations was 1.6 (95% CI, 0.7–3.5; 11 cases). In women, two cases and four controls had ever worked in paper and printing occupations.

The association between occupation and lung cancer risk was examined in a population-based case–control study in New Mexico, United States (Lerchen et al., 1987). Cases included Hispanic white and other white residents of New Mexico, aged 25–84 years, with primary lung cancer, other than bronchiolar/alveolar carcinoma, diagnosed between January 1980 and December 1982. Cases were identified through the New Mexico Tumor Registry. Controls were selected by random-digit dialling for persons aged 65 and older from the Health Care Financing Administration's roster of Medicare participants. The controls were frequency matched to cases for sex, ethnicity and 10-year age category. A personal interview was used to obtain lifetime occupational and smoking histories and self-reported history of exposure to specific agents. The overall case interview rate was 89%, half of which were conducted with next-of-kin. The analysis was based on data from 333 male cases and 499 male controls. The odds ratio adjusted for age, ethnicity and smoking was 0.8 (95% CI, 0.4–1.8; 11 cases) for workers in the printing industry. Five cases had ever been employed as printers (odds ratio adjusted for age, ethnicity and smoking, 0.8; 95% CI, 0.3–2.6).

The association between lung cancer risk and occupation was examined in a case–control study in six areas of New Jersey, United States (Schoenberg et al., 1987). The study included 763 histologically confirmed incident lung cancers in white males and 900 general population white male controls between September 1980 and October 1981. The response rates were 70% among cases and 64% among controls. A total of 429 cases and 334 next-of-kin were interviewed. For self-respondent cases, controls were selected from State Drivers License files and frequency matched by race, age and residence. For deceased or incapacitated cases, controls, individually matched by race, age, residence and closest date of death or diagnosis, were selected from State mortality files, excluding deaths from lung cancer and respiratory diseases. Among controls, 564 were self-respondents. The smoking-adjusted odds ratio for subjects ever employed as printing workers (printing pressmen, compositors, typesetters, photoengravers, lithographers, printing engravers, printing trade apprentices) was 2.5 (95% CI, 1.0–6.1; 20 cases). After adjustment for additional variables (age, area, respondent type, education, vegetable consumption), the odds ratio was 2.3 (95% CI, 1.0–5.3). A significant crude odds ratio of 8.4 (7 cases) was found for duration of employment of 10 or more years as a printing worker.

The association between lung cancer risk and occupation was studied in a case–control study in Missouri, United States (Hoar Zahm et al., 1989). Study subjects were identified through the Missouri Cancer Registry. Cases were all white male Missouri residents who had been diagnosed with histologically confirmed lung cancer from January 1980 through to November 1985; controls were white male Missouri residents diagnosed with cancer, excluding cancers of the lip, oral cavity, oesophagus, lung, urinary bladder, ill-defined sites and unknown sites. The study involved 4431 cases and 11 326 controls. Occupation at the time of diagnosis was abstracted from questionnaires administered to the cancer patients at the hospitals. Forty-eight percent of lung cancer cases and 55% of controls had unknown occupation on the records. The age- and smoking-adjusted odds ratio for printing occupations was 1.1 (95% CI, 0.6–1.9; 21 cases). An excess of adenocarcinoma was found for printing occupations (odds ratio, 1.8; 95% CI, 0.7–4.2; 7 cases). [The Working Group noted that many cases and controls had no known occupation.]

Occupational risk factors for lung cancer were assessed in a population-based case–control study in the Detroit Metropolitan Area, United States (Burns & Swanson, 1991b). Incident cancers occurring among white and black men and women between the ages of 40 and 84 were enrolled into the study through a rapid reporting system within two to six weeks of diagnosis. There were 5935 lung and bronchial cancer patients in the case group and 3956 colonic and rectal cancer patients in the control group. Subjects or their surrogate respondents were interviewed by telephone. Interviews with surrogates for deceased subjects were conducted with 39% of lung cancers, compared to 13% of the colonic and rectal cancer controls. The overall response rate was 93%. No elevated risk was found for printers (odds ratio adjusted for age at diagnosis, race, smoking and gender, 0.8; 95% CI, 0.4–1.7; 18 cases). The adjusted odds ratio for the printing industry was 0.7 (95% CI, 0.4–1.1; 54 cases). [The Working Group was concerned that only colonic and rectal cancers were used as controls and noted that the series of controls used in this study is probably the same as that in Burns & Swanson (1991a).]

Morabia et al. (1992) reported on a multicentre case–control study involving 1793 histologically confirmed male lung cancer cases at 24 hospitals in nine metropolitan areas in the United States during 1980–89. Information on sociodemographic characteristics, cigarette smoking and occupational history (usual job title and 44 specific exposures) were collected through a standardized questionnaire. For each case, one or two controls were selected from among patients without lung cancer, individually matched by race, age, hospital and cigarette smoking habits. Out of 3228 controls, 69% had a diagnosis of cancer other than of the lung and 31% had a diagnosis of nonneoplastic diseases. Unconditional logistic regression was used to compute odds ratios adjusted for matching variables. A statistically significantly increased risk was found for bookbinders and related printing trade workers (odds ratio, 3.3; 95% CI, 1.2–8.9; 11 cases).

(d) Lymphohaematopoietic neoplasms (see Table 11)

Occupational risks of leukaemia were investigated using the data from the Tri-State Leukemia Survey collected in New York, Baltimore and Minnesota, United States, during the period 1959–62 (Viadana & Bross, 1972). A subsample of 1345 white adult leukaemia cases and 1237 controls randomly selected from households in the same area as the cases were interviewed and included in the analysis. No relationship was found between leukaemia and occupation among women. Men reporting employment as printers in any of the five most recent jobs prior to interviews had a nonsignificant excess when compared to either non-printers (age-adjusted odds ratio, 1.5 [95% CI, 0.6–3.7]; 17 cases) or to clerks (age-adjusted odds ratio, 1.9 [95% CI, 0.7–5.0]).

Blair et al. (1993) examined the relationship between non-Hodgkin's lymphoma and occupation in a population-based case–control study in Iowa and Minnesota, United States. Cases were 715 white men with histologically confirmed non-Hodgkin's lymphoma reported to the Iowa State Health Registry during 1981–83 and patients identified through a surveillance network of hospitals in Minnesota during 1980–82. Controls were selected by random-digit dialling from the medical files of the Health Care Finance Administration and state vital records. All controls were frequency matched by state, age (15-year categories) and by year of death for deceased cases. A total of 622 cases and 1245 controls were interviewed on lifestyle and occupational history. The proportion of next-of-kin interviews was 30% for cases and 34% for controls. Subjects employed only as farmers were excluded, which left 546 cases and 1087 controls in the study. Odds ratios adjusted for age, state, smoking, type of respondent, family history of lymphoma and agricultural exposure to pesticides were computed. Six cases employed in printing press occupations yielded an odds ratio of 1.5 (95% CI, 0.4–5.1). A significant excess was found among workers with a duration of employment in the printing and publishing industry longer than 10 years (odds ratio, 2.5; 95% CI, 1.1–5.7); a decreased odds ratio of 0.5 was found for duration of employment less than 10 years (95% CI, 0.2–1.3).

(e) Oropharyngeal cancer (see Table 11)

Huebner et al. (1992) conducted a population-based case–control study of oropharyngeal cancers in four areas of the United States (Atlanta, New Jersey, Los Angeles and San Francisco). A total of 1114 black and white incident cases (762 males and 352 females) of primary histologically confirmed cancers of the oral cavity and pharynx (salivary glands and nasopharynx excluded) were identified from population-based cancer registries during 1984–85. A total of 1268 controls were obtained through random-digit dialling (aged 18–64 years) and from Health Care Financing Administration files (aged 65–79). These were frequency matched to the cases by race, sex, five-year age group and study area. Interviews on personal habits and occupation were conducted with the subjects or with next of kin for 75% of cases and 76% of controls. Surrogate respondents constituted 22% of cases and 2% of controls. The odds ratio adjusted for smoking, alcohol consumption, study location, age and race for men ever employed in the printing industry was 0.8 (95% CI, 0.5–1.5; 28 cases) and that for women was 2.9 (95% CI, 0.7–11.6; 8 cases). Analysis by job category revealed an odds ratio of 0.7 (95% CI, 0.3–1.4; 16 cases) for male printers/pressmen.

(f) Testicular cancer (see Table 11)

Coldman et al. (1982) examined the relationship of testicular seminoma to several risk factors in a case–control study of 128 histologically confirmed cases diagnosed between 1970 and 1977 in a regional treatment centre in Vancouver, Canada. One-hundred-and-twenty-eight controls were matched by age and year of diagnosis, and selected from other patients with skin cancers or Hodgkin's disease. A detailed occupational history was obtained for 89% of cases and 88% of controls. Seven cases had ever worked in the printing industry (odds ratio, 7.2; 95% CI, 0.9–162.3). No significant differences between cases and controls for exposure to inks were noted [actual results not reported].

The risk for testicular cancer associated with socioeconomic status and occupation was examined by Swerdlow et al. (1991) in England. A total of 259 cases of primary testicular cancer, incident between January 1977 and February 1981 and aged at least 10 years, were enrolled in six radiotherapy and oncology centres. Two control groups were selected from the same centres: 238 patients with other cancers (35% represented Hodgkin's disease, 13% non-Hodgkin's lymphoma, 10% brain cancer, 8% urinary bladder cancer) and, in other departments, 251 hospital inpatients with a wide range of non-malignant conditions. The subjects were interviewed on education, lifetime occupational history and father's occupation. Odds ratios adjusted for age and region of residence were estimated using conditional logistic regression models. When the results of the analyses for each control group were similar, the two sets of controls were combined. The risk was examined for occupation at age 20, at age 30, for the longest held occupation, for ever held occupation and for that most recently held. The only substantially increased odds ratio was reported for subjects ever employed in paper and printing (odds ratio, 2.1; 95% CI, 0.8–5.0; 10 cases).

(g) Other cancer sites (see Table 11)

The Missouri Cancer Registry identified 1993 white male cases of histologically confirmed colonic cancer and 9965 age-matched cancer controls, diagnosed between 1984 and 1987 (Brownson et al., 1989). Two control groups were selected. The first contained all other cancers, the second excluded cancer sites known to be related to occupations (leukaemia, peritoneum, nasal cavity, lung, pleura, urinary bladder). Data on occupation and industry were collected routinely from hospital medical records at the time of diagnosis. Age-adjusted odds ratios were computed. Since odds ratios were generally of similar magnitude using each of the two control groups, all other cancers were used as controls. Excesses were found for printing machine operators (odds ratio, 1.9; 95% CI, 1.0–3.3; 18 cases) and workers employed in the printing and publishing industry (odds ratio, 1.8; 95% CI, 1.2–2.7; 33 cases).

Using the above-mentioned Missouri Cancer Registry, Brownson et al. (1990) evaluated the risk of brain cancer in relation to employment history in white males in a study of 312 histologically confirmed brain and other central nervous system cancers (ICD codes 191–192) and 1248 other cancer controls, frequency matched by age. Controls excluded cancers of ill-defined or unknown primary sites. The study covered the period 1984–88. Data on occupation and tobacco smoking habits were collected as in Brownson et al. (1989). Of the eligible subjects, 34% of cases and 38% of controls were excluded due to missing occupational information. [The Working Group was concerned by this high proportion of missing information.] The age- and smoking-adjusted odds ratio for workers in the printing and publishing industry was 2.8 (95% CI, 1.0–8.3; 7 cases). The elevated risk was equally distributed between astrocytic- and ‘other’-cell type cancers [odds ratios not reported].

The role of occupational risk factors in the occurrence of pancreatic cancer was investigated in a case–control study conducted in France (Pietri et al., 1990), with 171 cases (105 men, 66 women) enrolled in seven public hospitals in Paris between 1982 and 1985. For each case, two controls, matched by sex, age at interview (within five years), hospital and interviewer, were selected: one among patients with cancers other than neoplasms of the biliary tract, liver, stomach and oesophagus and the other among patients with non-neoplastic diseases. The total number of controls was 317 (196 men, 121 women). All subjects agreed to be interviewed. Odds ratios were derived from unconditional logistic regression models adjusting for age, sex, foreign origin, education, coffee, cigarette and alcohol consumption using the two sets of controls combined. Cases and controls did not significantly differ in having been employed for at least one year in the printing industry (odds ratio, 1.3; 95% CI, 0.5–3.4; 8 cases). Among manual workers, three cases had ever been employed in printing occupations (odds ratio, 0.8; 95% CI, 0.2–3.4).

A case–control study of renal-cell cancer was conducted in Finland (Partanen et al., 1991). A total of 672 primary incident cases of renal-cell adenocarcinoma were identified through the Finnish Cancer Registry in 1977–78. Two population controls were matched to each case for year of birth, gender and survival status at the time of data collection; they were randomly selected from the population register, and the total number of controls was 1344 (280 alive, 1064 deceased). Data were collected by questionnaire, which had response rates for self-respondents of 76% for cases and 79% for controls; for next-of-kin respondents, the rates were 67% for cases and 65% for controls. The analyses were based on 338 matched sets of cases and controls, 192 cases having one control and 146 cases having two controls. An ad-hoc scoring system (three levels) was developed based on exposure level (background, low, high), the proportion of annual work time spent in the higher level of exposure and the period and duration of exposure for nine selected exposures. For men and women, conditional logistic models were used to calculate separately odds ratio estimates adjusted for smoking status, coffee consumption and obesity. The odds ratio associated with employment for at least five years in the printing and publishing industry was 4.6 (95% CI, 0.9–23.5; 6 cases) in men and 8.0 (95% CI, 0.9–69.8; 5 cases) in women. Analysis by occupation revealed an increased risk among male printers who had worked for at least five years (odds ratio, 6.0, 95% CI, 0.7–54.5; 4 cases). Out of 10 cases with a high level of exposure to gasoline (odds ratio, 7.4; 95% CI, 1.6–35), three worked as operators of a relief-printing press and one as a rotogravure pressman who cleaned the press cylinders with gasoline; one of the two controls with high exposure was a press operator and foreman in a relief-printing shop.

A population-based multicentric international case–control study investigated the relationship between renal-cell adenocarcinoma and occupation (men and women) (Mandel et al., 1995). Study centres in Australia, Denmark, Germany, Sweden and the United States interviewed 1732 incident cases and 2309 population controls (about 60% were men). The elicited occupational data were coded as ever/never in an occupation and industry. Ever employment in the printing or graphic industry was associated with an odds ratio of 1.3 (95% CI, 0.8–2.0; 39 cases) for men, adjusted for age, smoking status, body mass index, education and study centre.

(a) Cohort study

Kristensen et al. (1993) investigated the offspring of a cohort of 10 992 men who had been members of the Oslo (Norway) unions of compositors, lithographers and bookbinders between 1 January 1930 and 31 December 1974, and who were alive at the time of the 1960 census. A total of 6251 infants born during the period 1967–86 was identified by linking records from the printers' unions and the Norwegian Medical Births Registry. The reference group was all 118 403 births to married Oslo couples during the period 1967–86. The morbidity ratio, standardized for maternal age, year of birth, sex and birth order for total birth defects (ICD-8 codes 740–759.9) was 0.9 (95% CI, 0.8–1.0). When specific congenital anomalies were considered, the highest morbidity ratio was 1.6 (95% CI, 1.0–2.5) for cleft lip and/or palate. Further analyses in this study related to exposure to lead and solvents on the basis of job classification. On the basis of 69 job codes in the union records, each worker was categorized a priori into one of four exposure groups: lead only, solvents only, lead and solvents or ‘other’ exposure. The father's job category one year prior to the child's birth was considered in the allocation of exposures, as the a-priori hypothesis was that exposure acts on the developing paternal germ cells, that is up to three months prior to conception. None of the paternal exposure categories was associated with the risk of total birth defects. However, boys whose fathers had been exposed to lead were at increased risk of cleft lip (using the reference population of total births to married couples in Oslo, the SMR was 4.1; 95% CI, 1.8–8.1). The authors noted that they did not have an a-priori hypothesis about cleft lip, and the possibility of a chance cluster could not be excluded.

In addition to congenital anomalies, the authors considered fetal growth and birth weight, gestational age and perinatal death. A higher proportion (67/1000 births) of printers' children than that of total births in Oslo (54/1000 births) was small for gestational age; there was no consistent association with exposure category between internal analysis and analysis using the external reference group. There was no association with birth weight, and the overall distribution of gestational age was similar between printers' children and total births in Oslo. The proportion of printers' children that were delivered as early pre-term births was 6.5/1000, as compared to 8/1000 for total births to married couples in Oslo. Compared with ‘other’ exposures, the RR for early pre-term births (16–27 weeks) was 5.4 (95% CI, 1.7–17.4) for infants with paternal exposure to solvents and 8.6 (95% CI, 2.7–27.3) for infants with paternal exposure to lead and solvents; these RRs were adjusted for fathers' occupational status, year of birth, twin births, maternal age, birth order and sex. Compared with the external reference group, the elevated risk associated with external exposure to lead and solvents was still apparent (SMR, 2.1; 95% CI, 1.2–3.3). The corresponding SMR for exposure to solvents only was 1.3 (95% CI, 0.6–2.3). The rate of late abortion (still-born infants with gestational age of less than 28 completed weeks) was 2.7/1000 births to printers, compared with 5.3/1000 for the reference population. The rate of still births and early neonatal deaths was 16.0/1000 infants born to printers, as compared to 12.6/1000 for the reference population. The RR for total perinatal deaths was 2.4 (95% CI, 1.2–4.9) for children with fathers exposed to lead and 1.9 (95% CI, 0.96–3.7) for children with fathers exposed to lead and solvents. The association with lead only was not apparent in the comparison with the external reference group, but the SMR for paternal exposure to lead and solvents was 1.6 (95% CI, 1.2–2.2) (Kristensen et al., 1993). [The Working Group noted that the parents' smoking behaviour was unknown.]

(b) Case–control studies of congenital anomalies

In the Metropolitan Atlanta Congenital Defects Programme (United States), parents of infants with a variety of anomalies have been interviewed since 1970n about a range of exposures three to six months after the birth of their children (Edmonds et al., 1981). The occupational distribution of parents of babies with one malformation was compared with that of parents with babies with all other malformations (Erickson et al., 1979). [The period of study was not specified]. In general, the mother was the sole informant; some 85% of eligible mothers were interviewed. Information adequate for coding the industry of paternal employment at the time of conception was obtained from 76% of the total 989 women interviewed. Eleven (1.6%) of the 705 fathers for whom this information was available worked in the printing and publishing industry at the time of conception; this proportion is similar to that of all employed persons aged 16 or more as determined in the 1970 census. Analyses were made by industry, occupation and occupation–industry cross classification. Results were presented if the RR was greater than or equal to 2, associated with a p value of less than 0.05 and the exposure of at least two fathers. The only category involving paternal employment in printing for which a result was presented was for cleft palate; two fathers of affected births were clerical personnel in the printing industry. Both were mailers and handled freshly printed material. None of the fathers of infants with other types of malformations had this category of employment. In regard to maternal occupation, the reference period during which exposure was considered was the first trimester and during the period up to and including the time of conception. Two mothers of infants with Down's syndrome were managers in printing firms; none of the mothers of infants with other types of malformations were so employed. Three mothers of babies with omphalocele and gastroschisis were also employed in the printing industry (RR, 7.7; p < 0.05); two operated printing presses while one worked both as a binder and a press operator. Another mother employed in the printing industry operated a press and her baby was affected by microcephaly, cleft palate and limb defects.

In a large study based on the national congenital malformation system in England and Wales during the period 1974–79, information on the occupations of both mothers and fathers of births with congenital anomalies was obtained from the notification form which is completed by Area Health Authority staff (Office of Population Consensus and Surveys, 1982). About 60% of births with malformations notified had a father's and mother's occupation stated. In a total of 47 352 births with notified malformations, the father's occupation was stated; 20 778 births with malformations had a mother's occupation stated. In the analysis of paternal occupation, the analysis was performed for two separate periods — 1974–76 and 1977–79 — and births with malformations notified in those periods were compared with total births in the middle year of each period (i.e. 1975 and 1978, respectively) to produce prevalence rates at birth by father's occupation. The distribution of paternal occupation for total births was obtained from birth certificate data, so the sources of information for cases and total births differ. The malformation ratio (observed cases/expected cases) associated with paternal employment in the paper and printing industries was 0.92 in 1974–76 and 0.86 in 1977–79. Data were presented on three groups of malformations — neural tube defects, cleft lip and/or palate and Down's syndrome. No increase in the malformation ratio was apparent for any of these groups. When analysis was made according to a smaller subdivision of types of occupations, the malformation ratio associated with the father being a printer was 1.6 in the first period and 1.4 in the second period. These excesses were not statistically significant. This information was not presented for the three selected groups of malformation. As noted by the authors, the differences in sources of information for cases and total births are a potentially serious source of bias as, at birth registration, details of occupation are collected by the Registrar with the aid of several questions whereas, at notification, details of occupation are derived either from a record obtained by health workers during pregnancy or from an enquiry made to the mother by a health visitor. Thus, the information obtained from birth registration tends to be more precise than that obtained from notification, and this may cause certain occupations to be classified differently between the sources, with a differential effect on numerator and denominator. This bias may account for the nonsignificant excess of malformations in the offspring of printers. As there are no comparable data on maternal occupation from birth registration, the analysis by maternal occupation was limited to a study of the proportions of different malformations in births delivered to mothers of certain occupational groups, compared to the overall distribution of total births with malformations. For neural tube defects, the proportionate malformation ratio associated with the mother being a paper or printing worker was 1.4 during the period 1974–76, and 1.2 during the period 1977–79. For cleft lip and/or palate, the corresponding ratios were 1.2 and 1.7; for Down's syndrome, the ratios were 1.0 and 1.7, respectively. These ratios were not statistically significant.

A study of similar design was made for births with malformations notified in the period 1980–82 (McDowall, 1985). Notification may be made of malformations recorded in still-births and in live-births during the first week of life. Of a total of 40 346 births with malformation notified in England and Wales during this period, 62% (24 922) had a father's occupation stated and 28% (11 115) a mother's occupation. Standardized congenital malformation ratios for paternal occupations were calculated as the ratio of observed notifications to expected notifications, the former being adjusted for each specific type of malformation to account for the under-reporting of paternal occupation at notification. The effect of this adjustment was to make the standardized congenital malformation ratio equal to 1 rather than about 0.6 for a paternal occupation that had the same malformation notification rate as all cases for which an occupation was recorded. The number of expected notifications was derived by applying maternal age-specific rates of notified malformations to the number of births in each paternal occupation unit as determined from live- and still-birth registration. A significant positive association (standardized congenital malformation ratio, 1.4 based on 159 exposed cases) between total malformations and paternal employment as a printer was found, but the author noted that this would have been affected by the bias arising from the use of different sources of information for the numerator and the denominator. With regard to anomalies of specific types, a significant excess was found for rectal and anal atresia and stenosis (standardized congenital malformation ratio, 3.5 based on five exposed cases). No excess was found for related occupations; the standardized congenital malformation ratio was 0.4 for ‘compositors’, based on 11 exposed cases, 0.4 for ‘electrotypers, stereotypes, printing plate and cylinder preparers’ based on four exposed cases, 0.6 for ‘printing machine minders and assistants’ based on 26 exposed cases and 0.8 for 'screen and block printers' based on six exposed cases. As in the earlier analysis (Office of Population Censuses and Surveys, 1982), only standardized proportionate congenital malformation ratios could be calculated for maternal occupation. A significant excess was found for reduction deformities associated with maternal employment as a printer (standardized proportionate congenital malformation ratio, 10.1 based on three exposed cases). No excess was apparent for related occupations.

A large survey assessing the effect of occupational factors on pregnancy outcome in employed workers was conducted in Montreal, Canada, during the period 1982–84. A sample of 56 067 women who had just given birth or been treated for a spontaneous abortion in 11 hospitals, in which about 90% of deliveries in the city take place, were interviewed about their index pregnancies and all previous pregnancies (McDonald et al., 1987, 1988). In the main analyses relating to the women's occupation, 60 occupational groups were considered; none of the groups was related specifically to printing and related occupations. However, noting the findings of Erickson et al. (1979) relating to omphalocele or gastroschisis in employees in the printing industry, McDonald et al. (1988) noted that there was no evidence of such an association in the Montréal data. The only malformations in printing workers were two cases of Fallot's tetralogy. The women in the Montreal study were also asked about the husband's or partner's employment at the time of the first missed menstrual period for both current and past pregnancies (McDonald et al., 1989). For spontaneous abortion, the analysis was limited to those in which the women were employed for 30 h a week or more at the time of conception; there were 47 326 such pregnancies — 24 711 index and 22 615 previous. The RR of spontaneous abortion associated with the father working in ‘printing operations’ was 0.9 ([95% CI, 0.7–1.1]; 81 exposed cases). The analysis relating to congenital anomalies was based on 47 822 pregnancies (27 472 index and 30 350 prior) of women employed 15 h a week or more at the time of conception. This total excluded spontaneous abortions of less than 20 weeks gestation but included therapeutic abortions following prenatal diagnosis of fetal abnormality. These RRs were adjusted for maternal age, gravidity, previous miscarriage, ethnic group, educational level, smoking and alcohol consumption. Two pregnancies associated with chromosomal anomalies were observed where the father was employed in printing operations (RR, 1.8 [95% CI, 0.3–9.9]). The RR of developmental defects, defined to include neural tube defects, cleft lip and/or palate and anomalies of the heart and of the respiratory, digestive and urinary tracts, was 0.6 ([95% CI, 0.2–1.8]; 4 exposed cases). In the analysis relating to congenital anomalies, no adjustment was made for confounding variables. No excess was apparent for specific types of congenital anomaly considered and, in particular, none of the fathers of cases with cleft lip and/or palate was a printer.

Olshan et al. (1991) reported a study based on 14 415 live births with birth defects recorded in the population-based registry in British Columbia, Canada, in the period 1952–73. Two controls were matched to each case on month, year and hospital of birth. Information on paternal occupation was obtained from the birth-registration record. Potential confounding by parental age, ethnic group and outcome of previous pregnancy was considered. Twenty categories of congenital anomalies were considered in relation to 58 occupational categories. Of the fathers of infants with congenital anomalies, 1.7% (245/14 415) had been employed as printers at the time of birth of the child. This occupation was associated with RRs greater than or equal to 1.5 for five groups of anomalies, and RRs of 0.7 or less for six groups of anomaly. RRs of 2 or more were observed for obstructive renal defects (RR, 2.0; 95% CI, 0.4–9.9; 3 exposed cases), atresia of urethra (RR, 4.5; 95% CI, 1.0–20.8; 7 exposed cases) and clubfoot (RR, 2.2; 95% CI, 1.2–4.1; 19 exposed cases). The RR for cleft palate was 1.6 (95% CI, 0.3–7.1; 3 exposed cases), that for cleft lip was 1.0 (95% CI, 0.1–11.1; 1 exposed case) and that for cleft palate with cleft lip was 1.2 (95% CI, 0.3–4.5; 3 exposed cases).

Fedrick (1976) determined the RRs, defined as the rate in specific categories relative to the rate in a total population, for anencephalus associated with specific paternal occupational orders at the time of birth in a population-based study in Oxfordshire and West Berkshire, United Kingdom, during the period 1965–72. Paternal occupation could be classified to an ‘order’ for 88% (92 083/104 854) of births in 1965–71 (data on total births in 1972 were not included) and 85% (151/177) of the births with anencephalus. There was a statistically significant increase in risk associated with work in the paper and printing industries (RR, 1.9; 5 exposed cases). All five case fathers employed in the paper and printing industries were printers; the RR for printers was 6.7. [The Working Group noted that although a p value of less than 0.001 was reported, detailed subgroup analysis was not reported for the other occupational groups.]

Using data from a case–control study of neural tube defects in upstate New York, United States, during the period 1968–74 which was aimed primarily at assessing a possible association with oral contraceptive use and included 201 mothers of cases, Polednak and Janerich (1983) analysed parental occupations as reported on the birth records of 171 case–control pairs. The RR associated with paternal employment in the printing industry (other than as a salesman) was 0.5 (not statistically significant, based on three discordant pairs).

Ericson et al. (1988) examined the association between neural tube defects and maternal occupation in Sweden, using three sets of data. In all three data sets, the defects were ascertained from the records of the Swedish Medical Birth Registry. In the first data set, the maternal occupation recorded in the 1975 census was determined for 158 infants with neural tube defects identified during the period 1976–77 and 316 controls matched on year of birth, maternal age and parity. In the second data set, the maternal occupation recorded at the 1980 census was determined for 103 infants with neural tube defects born in 1980 and 1981 and 206 randomly selected controls; the age and parity distributions did not differ significantly between cases and controls. In the third data set, maternal occupation recorded at the first antenatal clinic visit was determined for 87 infants with spina bifida born during the period 1982–84 and 174 controls matched on year of birth, maternal age and parity. Anencephalus was not considered in this last period since few cases were born as a consequence of intensive prenatal screening. No statistically significant association with any type of maternal occupation was found in any of the three data sets or in a pooled analysis of these. The RR associated with maternal employment in the ‘graphic industry’ was 1.0 (1 exposed case and 2 exposed controls in pooled analysis).

Brender and Suarez (1990) examined the association between paternal occupations thought to involve exposures to solvents and anencephalus in Texas, United States, during the period 1981–86. A total of 727 cases were identified by death certificates, live-birth records and records of fetal death. Of these, 528 (72.6%) were included in the analysis, exclusions having to be made either because the record from which the case was ascertained could not be matched with a live birth record from which the information of paternal occupation was obtained, or because the recorded occupation could not be classified. Similarly, only 1160 (79.2%) of 1464 randomly selected live-born controls, frequency matched on ethnic group and year of birth, were included. The RR associated with the father being employed as a printer was 1.6 (95% CI, 0.4–5.5; 4 exposed cases).

Thus, the only group of congenital anomalies associated with printing-related work in more than one study was cleft lip and/or cleft palate (Erickson et al., 1979; Kristensen et al., 1993). [The Working Group noted that cleft lip with or without associated cleft palate and cleft palate without associated cleft lip are considered to be etiologically distinct.]

(c) Case–control study of infertility

Rachootin and Olsen (1983) examined the association between infertility and occupation in a case–control study using data collected from medical records and mailed questionnaires. A total of 1069 infertile case couples were identified from records of the Odense University Hospital in Denmark during the period 1977–80 and were compared with 4305 fertile control couples who had a healthy child born at the same hospital during the period 1977–79. Information on occupation was obtained from 927 case and 3728 control couples, representing a response rate of about 87% for each group. The analyses focused on three subgroups: (1) 258 couples in which the male partner was diagnosed as having abnormalities of sperm density, mortality or morphology; (2) 305 couples in which the female partner was diagnosed as having amenorrhoea, anovulation, luteal insufficiency or other endocrine malfunction; and (3) 129 couples with idiopathic infertility. The first two subgroups were not mutually exclusive as 48 couples were diagnosed as having both of these reproductive disorders. Analysis in relation to occupation was made restricting the control group to couples who reported that the index child had been conceived within a year of the decision to have the child and where the female partner was 20 years of age or older at the time the questionnaire was sent out. Statistically significant positive associations were found between sperm abnormalities and the male partner being employed as a typesetter (a) in the year prior to hospital admission of spouse for infertility investigation and (b) as the longest-held occupation prior to hospital admission. Odds ratios were not presented as there were fewer than 20 controls in this occupational category. There was no association with idiopathic infertility or with delayed conception, defined as conceiving more than a year after the decision was taken to have a child. In addition, there were significant positive associations between the female member of the pair having been employed as a typesetter both in the year prior to hospital admission and for whom this was the longest-held occupation prior to hospital admission. There was no association between female employment in typesetting and idiopathic infertility or delayed conception. (Again no odds ratio was presented as fewer than 20 controls were employed in this category.) A total of 55 categories of male employment and 43 categories of female employment were analysed.

(d) Case – control study of mental retardation

A case–control study of parental employment and mental retardation of unknown etiology was carried out in children referred to the paediatric or child neurology departments of the Nijmegen University Hospital (The Netherlands) or to rehabilitation centres in the vicinity during the period 1979–87 (Roeleveld et al., 1993). A total of 340 cases with mental and psychomotor retardation (ICD codes 317–319) with an IQ of less than 80 according to the Dutch School Criteria for the Mentally Retarded were compared with 362 children with other congenital handicaps. The main diagnoses of the control children were familial neuromuscular and metabolic disorders and cerebral palsy. Data on parental occupation were obtained by interview with 306 parents of cases (participation rate, 90%) and 322 parents of controls (participation rate, 89%). Using a job–exposure matrix developed for use in England and Wales, the RR, adjusted for primigravidity, prematurity, alcohol consumption and leisure-time activities, associated with maternal occupational exposure to printing inks during the last trimester of pregnancy was 1.6 [95% CI, 0.8–3.0]. The corresponding RR associated with paternal exposure during the three months prior to conception was 0.7 [95% CI, 0.4–1.3]. When the father was asked directly about exposure to printing inks, the associated RR was 0.7 [95% CI, 0.2–2.8]. For 57% of the cases and 47% of the controls, the father provided information about his own exposure; for the remaining subjects, the mother supplied information about the father and the appropriate exposure checklists were left for him to complete. None of the mothers reported exposure to printing inks during the third trimester in response to a direct question about this. The lag time between the critical period and the interview varied from two to 25 years, with an average lag time of 10.7 years. A number of exposures that could not be identified by the job–exposure matrix approach were studied. Among these, maternal exposure in the last trimester to copying machines was associated with a RR of 3.0 [95% CI, 1.1–8.3] and exposure to correction fluid was associated with a RR of 1.7 [95% CI, 0.7–4.1]; these RRs were adjusted for primigravidity, prematurity, alcohol consumption and leisure time activities. Paternal preconceptional exposure to copying machines was associated with a RR of 1.5 [95% CI, 0.8–2.8].

(a) Urinary mutagenicity

Crebelli et al. (1985) studied urinary mutagenicity in workers employed in the use of an IBM Printing System 3800, which implies exposure to small amounts of 2,4,7-trinitro-9-fluorenone. Urine samples were collected in the morning after at least three workdays from 22 exposed white men (11 smokers, 11 nonsmokers) and 18 white men (7 smokers, 11 nonsmokers) working in the same bank, but only occasionally in contact with the printing system. Analytical measurements showed the presence of 2,4,7-trinitro-9-fluorenone residues in both printed and photocopied sheets (0.4–4 µg in 21 × 27 cm sheets) and in the surrounding air (0–0.4 µg/m³). Urine samples were concentrated by adsorption onto XAD-2 resin and elution with acetone. Mutagenicity was studied both by the plate incorporation assay and the microtitre fluctuation assay using Salmonella typhimurium TA98. Pure 2,4,7-trinitro-9-fluorenone was highly mutagenic, but urine concentrates of both exposed and presumed unexposed workers were mostly negative. A few concentrates were significantly mutagenic but with no relationship to exposure (one among exposed and four among controls). In the presence of an exogenous metabolic activation system, five of eight concentrates from both unexposed and exposed smokers were mutagenic, whereas the urine of nonsmokers was completely inactive.

(b) Cytogenetic damage in lymphocytes

Forni et al. (1971) studied chromosomal aberrations in peripheral lymphocytes in 34 male workers employed in a rotogravure plant, 10 of whom were exposed to both benzene (for 1–22 years) and toluene (for 12–14 years) and 24 to toluene only (for 3–15 years). Each worker was matched with a healthy control of the same sex and approximately the same age with no history of exposure to benzene or its homologues or to therapeutic or occupational X-irradiation. There was no control for tobacco smoking or alcohol consumption, but people having had a recent viral disease or vaccination were excluded. The concentrations of benzene and homologues of benzene expressed as concentrations of benzene were 131 ppm [419 mg/m³] in the centre of the rotary machine room, 125 ppm [400 mg/m³] near the windows, 363 ppm [1162 mg/m³] near a folding machine and 532 ppm [1702 mg/m³] in the auxiliary room. After a benzene-poisoning epidemic, only toluene was used as an ink diluent. Concentrations of toluene in different parts of the working department varied between 56 and 824 ppm [211 and 3106 mg/m³] up to 1966, when the plant was moved and better ventilation was obtained (156 ppm [588 mg/m³] near folding machines and 265 ppm [999 mg/m³] between machine elements). In the benzene-toluene group, a statistically significant increase in the incidence of unstable chromosomal aberrations was found (1.66% compared with 0.61% in the controls). This was not found in the toluene-exposed group.

Funes-Cravioto et al. (1977) examined eight rotoprinting factory workers exposed to benzene during the 1940s and to toluene from 1950 onwards for periods ranging from two to 26 years. Air concentrations of benzene were not measured. Air concentrations of toluene showed time-weighted average values of 100–200 ppm [377–754 mg/m³] with occasional rises up to 500–700 ppm [1885–2639 mg/m³]. The frequency of chromosome breaks was 11.9% (5.6% in controls). No increase in sister chromatid exchange frequency was found in the four workers examined in this group. In 14 other workers exposed to toluene only (1.5–26 years), the frequency of chromosome breaks was 8.9%. [The Working Group noted that statistical evaluations were not made on the separate groups.]

Mäki-Paakkanenet al. (1980) studied 32 men working in two different rotoprinting factories and 15 control men from a research institute. Toluene was the only solvent used in the two factories. Concentrations of benzene, which is a contaminant, had been checked regularly since 1962 and had always been below 0.05% (average, 0.006%) as measured by gas chromatography. The personal toluene exposures of the printers were calculated as time-weighted averages for an 8-h work day and were between 7 and 112 ppm [26–420 mg/m³]. The cytogenetic study showed little difference between the workers and the controls (chromatid type, 1.0 versus 0.7%; chromosome type, 0.5 versus 0.9%; sister chromatid exchange, 8.5 versus 8.9 per cell). A higher frequency of sister chromatid exchange was observed among smokers both in the group of workers and in the controls. No effect of smoking was observed as regards chromosomal aberrations.

Bauchinger et al. (1982) examined 20 men working for more than 16 years on a rotogravure machine. The continuously measured toluene concentration in the air of the rotary machine was between 200 and 300 ppm [754 and 1131 mg/m³]. The toluene used contained < 0.3% benzene. From the same plant, a group of 24 unexposed workers [sex unspecified] was used as controls. There were 12 smokers and eight nonsmokers in the exposed group, and nine smokers and 15 nonsmokers in the control group. In the exposed group, a significantly larger number of peripheral lymphocytes with structural chromosomal changes were found: chromatid breaks (0.0036 versus 0.0019 per cell) and chromatid exchanges (0.0015 versus 0.0004 per cell), but not acentrics or dicentrics. The yield of gaps was also significantly increased (0.0248 versus 0.019 per cell). Significantly higher levels of sister chromatid exchange were observed in the nonsmoking rotogravure workers compared with nonsmoking controls (8.55 versus 7.75 per cell). This was also the case for toluene-exposed smokers compared with control smokers (10.33 versus 8.89 per cell). In both groups, smokers had significantly higher sister chromatid exchange values than nonsmokers.

A second study of chromosomal changes in peripheral lymphocytes was made on 27 workers from the same factory who had not been exposed to toluene for four months to five years (Schmid et al., 1985). These 27 workers were subdivided into two groups: those without exposure for up to two years, and those without exposure for more than two years. They were compared with 26 controls. The exposure conditions had been the same as in the first study. The group of 13 workers without exposure for up to two years still showed a significantly larger number of cells with structural chromosomal aberrations (S-cells; 0.69 versus 0.49%) and chromatid-type aberrations (but not gaps) than the controls (0.39 versus 0.20/100 cells). No significant difference was observed in the group of 14 workers without exposure for more than two years compared with the controls. In all groups, smokers showed a significantly higher incidence of sister chromatid exchange than nonsmokers, but the two post-exposure groups did not differ from the controls as regards incidence of sister chromatid exchange.

Chromosomal aberrations in peripheral lymphocytes of rotogravure printing plant workers [sex unspecified] were studied in two groups of exposed persons by Pelclová et al. (1990): 42 persons (37 smokers, 5 nonsmokers) who had been exposed to highly purified toluene (104–1170 ppm; 390–4380 mg/m³) and to printing dyes for an average of 12 years; 28 office and technical employees (17 smokers, 11 nonsmokers) of the same printing plant exposed to low levels of toluene (2.1–4.3 ppm; 8–16 mg/m³) and working on average 2 h per day in the rotary machine room. A control group of 32 persons (17 smokers, 15 nonsmokers) was selected from offices in a nearby brewery and dairy. Exposure was assessed by measuring toluene in blood at the end of the working shift (124.0 and 10.3 µmol/L in the two exposed groups, respectively; not measured in the control group). The percentages of aberrant cells were 3.6 and 3.3 in the exposed groups and 2.1 in the control group. There were significantly higher proportions of chromatid breaks per cell in both exposed groups and gaps and chromosomal exchanges in the low toluene-exposed group only, whereas the proportions of chromosomal breaks per cell were not different between the three groups. A small difference in the percentage of cells with aberrations was seen between smokers and nonsmokers (3.4 versus 2.4%; p < 0.05). Smoking printers had 3.8% aberrant cells, smoking office and technical employees had 3.4% and smoking controls had 2.5%. The difference between smoking printers and smoking controls was significant (p < 0.05).

Micronuclei and chromosomal aberrations in lymphocytes have been studied in 21 men employed as rotogravure printers by Niseet al. (1991). As controls, 21 workers were selected from a margarine factory with no exposure to solvents or other components of the printing trade. The median time of exposure was 25 years; median weekly individual exposure measured in 1986 was 150 mg/m³ when the median blood toluene content was 1.6 µmol/L (measured between 11.00–14.00 h in the middle of a working week). For lymphocytes stimulated with pokeweed mitogen, there was a significantly increased incidence of micronuclei in the printers compared with the controls, when allowance was made for age and smoking (2.8 versus 1.5/1000 cells). There was also a significant difference in the incidence of small micronuclei for which the size ratio micronucleus:main nucleus was ≥ 0.03 (1.0 versus 0.3/1000 cells) and, in the exposed group, an association between blood toluene and small micronuclei (0.17/1000 cells/-µmol toluene/L blood; p = 0.0005). No difference was found with lymphocytes stimulated with phytohaemagglutinin (2.8 versus 2.5/1000 cells). There was no significant difference in chromosome breaks between the exposed and the control group, but there was an association between chromosome breaks and earlier benzene exposure. [The Working Group noted that there is considerable variation in micronucleus frequency in pokeweed mitogen-stimulated lymphocytes. The relevant historical database from the same laboratory gives frequencies of 1.5 (this study), 3.4, 6.0, 7.0 and 2/1000 pokeweed mitogen-stimulated cells (Högstedt et al., 1988; Hagmar et al., 1989; Högstedt et al., 1991; Nise et al., 1991; Holmén et al., 1994).]

Lead concentrations in urine and the frequency of chromosomal aberrations and sister chromatid exchange in lymphocytes were studied in 84 printers exposed to lead (Chen et al., 1992). A linear correlation was found between urinary lead concentrations and the frequency of chromosomal aberrations and sister chromatid exchange.

Richer et al. (1993) exposed five male volunteers to 50 ppm [187.5 mg/m³] toluene, which is a prominent solvent in the printing industry, in a controlled exposure chamber for 7 h per day for three days on three occasions at two-weekly intervals. Blood samples were taken before and after exposure. No effect upon sister chromatid exchange frequencies was observed in peripheral lymphocytes.