Stress, the physical and/or psychological departure from a person’s stable equilibrium, can result from a large number of stressors, those stimuli that produce stress. For a good general view of stress and the most common job stressors, Levi’s discussion in this chapter of job stress theories is recommended.
In addressing the question of whether job stress can and does affect the epidemiology of cancer, we face limitations: a search of the literature located only one study on actual job stress and cancer in urban bus drivers (Michaels and Zoloth 1991) (and there are only few studies in which the question is considered more generally). We cannot accept the findings of that study, because the authors did not take into account either the effects of high density exhaust fumes or smoking. Further, one cannot carry over the findings from other diseases to cancer because the disease mechanisms are so vastly different.
Nevertheless, it is possible to describe what is known about the connections between more general life stressors and cancer, and further, one might reasonably apply those findings to the job situation. We differentiate relationships of stress to two outcomes: cancer incidence and cancer prognosis. The term incidence evidently means the occurrence of cancer. However, incidence is established either by the doctor’s clinical diagnosis or at autopsy. Since tumour growth is slow—1 to 20 years may elapse from the malignant mutation of one cell to the detection of the tumour mass—incidence studies include both initiation and growth. The second question, whether stress can affect prognosis, can be answered only in studies of cancer patients after diagnosis.
We distinguish cohort studies from case-control studies. This discussion focuses on cohort studies, where a factor of interest, in this case stress, is measured on a cohort of healthy persons, and cancer incidence or mortality is determined after a number of years. For several reasons, little emphasis is given to case-control studies, those which compare reports of stress, either current or before diagnosis, in cancer patients (cases) and persons without cancer (controls). First, one can never be sure that the control group is well-matched to the case group with respect to other factors that can influence the comparison. Secondly, cancer can and does produce physical, psychological and attitudinal changes, mostly negative, that can bias conclusions. Thirdly, these changes are known to result in an increase in the number of reports of stressful events (or of their severity) compared to reports by controls, thus leading to biased conclusions that patients experienced more, or more severe, stressful events than did controls (Watson and Pennebaker 1989).
Stress and Cancer Incidence
Most studies on stress and cancer incidence have been of the case-control sort, and we find a wild mix of results. Because, in varying degrees, these studies have failed to control contaminating factors, we don’t know which ones to trust, and they are ignored here. Among cohort studies, the number of studies showing that persons under greater stress did not experience more cancer than those under lesser stress exceeded by a large margin the number showing the reverse (Fox 1995). The results for several stressed groups are given.
- Bereaved spouses. In a Finnish study of 95,647 widowed persons their cancer death rate differed by only 3% from the rate of an age-equivalent non-widowed population over a period of five years. A study of causes of death during the 12 years following bereavement in 4,032 widowed persons in the state of Maryland showed no more cancer deaths among the widowed than among those still married—in fact, there were slightly fewer deaths than in the married. In England and Wales, the Office of Population Censuses and Surveys showed little evidence of an increase in cancer incidence after death of a spouse, and only a slight, non-significant increase in cancer mortality.
- Depressed mood. One study showed, but four studies did not, an excess of cancer mortality in the years following the measurement of a depressed mood (Fox 1989). This must be distinguished from hospitalizable depression, on which no well-controlled large-scale cohort studies have been done, and which clearly involves pathological depression, not applicable to the healthy working population. Even among this group of clinically depressed patients, however, most properly analysed smaller studies show no excess of cancer.
- A group of 2,020 men, aged 35 to 55, working in an electrical products factory in Chicago, was followed for 17 years after being tested. Those whose highest score on a variety of personality scales was reported on the depressed mood scale showed a cancer death rate 2.3 times that of men whose highest score was not referable to depressed mood. The researcher’s colleague followed the surviving cohort for another three years; the cancer death rate in the whole high-depressed-mood group had dropped to 1.3 times that of the control group. A second study of 6,801 adults in Alameda County, California, showed no excess cancer mortality among those with depressed mood when followed for 17 years. In a third study of 2,501 people with depressed mood in Washington County, Maryland, non-smokers showed no excess cancer mortality over 13 years compared to non-smoking controls, but there was an excess mortality among smokers. The results for smokers were later shown to be wrong, the error arising from a contaminating factor overlooked by the researchers. A fourth study, of 8,932 women at the Kaiser-Permanente Medical Center in Walnut Creek, California showed no excess of deaths due to breast cancer over 11 to 14 years among women with depressed mood at the time of measurement. A fifth study, done on a randomized national sample of 2,586 people in the National Health and Nutrition Examination Survey in the United States, showed no excess of cancer mortality among those showing depressed mood when measured on either of two independent mood scales. The combined findings of studies on 22,351 persons made up of disparate groups weigh heavily against the contrary findings of the one study on 2,020 persons.
- Other stressors. A study of 4,581 Hawaiian men of Japanese descent found no greater cancer incidence over a period of 10 years among those reporting high levels of stressful life events at the start of the study than those reporting lower levels. A study was carried out on 9,160 soldiers in the US Army who had been prisoners of war in the Pacific and European theatres in the Second World War and in Korea during the Korean conflict. The cancer death rate from 1946 to 1975 was either less than or no different from that found among soldiers matched by combat zone and combat activity who were not prisoners of war. In a study of 9,813 US Army personnel separated from the army during the year 1944 for “psychoneurosis”, a prima facie state of chronic stress, their cancer death rate over the period 1946 to 1969 was compared with that of a matched group not so diagnosed. The psychoneurotics’ rate was no greater than that of matched controls, and was, in fact, slightly lower, although not significantly so.
- Lowered levels of stress. There is evidence in some studies, but not in others, that higher levels of social support and social connections are associated with less cancer risk in the future. There are so few studies on this topic and the observed differences so unconvincing that the most a prudent reviewer can reasonably do is suggest the possibility of a true relationship. We need more solid evidence than that offered by the contradictory studies that have already been carried out.
Stress and cancer prognosis
This topic is of lesser interest because so few people of working age get cancer. Nevertheless, it ought to be mentioned that while survival differences have been found in some studies with regard to reported pre-diagnosis stress, other studies have shown no differences. One should, in judging these findings, recall the parallel ones showing that not only cancer patients, but also those with other ills, report more past stressful events than well people to a substantial degree because of the psychological changes brought about by the disease itself and, further, by the knowledge that one has the disease. With respect to prognosis, several studies have shown increased survival among those with good social support as against those with less social support. Perhaps more social support produces less stress, and vice versa. As regards both incidence and prognosis, however, the extant studies are at best only suggestive (Fox 1995).
It might be instructive to see what effects stress has had in experiments with animals. The results among well-conducted studies are much clearer, but not decisive. It was found that stressed animals with viral tumours show faster tumour growth and die earlier than unstressed animals. But the reverse is true of non-viral tumours, that is, those produced in the laboratory by chemical carcinogens. For these, stressed animals have fewer tumours and longer survival after the start of cancer than unstressed animals (Justice 1985). In industrial nations, however, only 3 to 4% of human malignancies are viral. All the rest are due to chemical or physical stimuli—smoking, x rays, industrial chemicals, nuclear radiation (e.g., that due to radon), excessive sunlight and so on. Thus, if one were to extrapolate from the findings for animals, one would conclude that stress is beneficial both to cancer incidence and survival. For a number of reasons one should not draw such an inference (Justice 1985; Fox 1981). Results with animals can be used to generate hypotheses relating to data describing humans, but cannot be the basis for conclusions about them.
In view of the variety of stressors that has been examined in the literature—long-term, short-term, more severe, less severe, of many types—and the preponderance of results suggesting little or no effect on later cancer incidence, it is reasonable to suggest that the same results apply in the work situation. As for cancer prognosis, too few studies have been done to draw any conclusions, even tentative ones, about stressors. It is, however, possible that strong social support may decrease incidence a little, and perhaps increase survival.