Association between incidence of non-Hodgkin's lymphoma and solar ultraviolet radiation in England and Wales
============================================================================================================

* Graham Bentham

## Abstract

**Objectives**: To examine whether the incidence of non-Hodgkin's lymphoma in different areas of England and Wales is associated with levels of solar ultraviolet radiation.

**Design**: Geographically based study examining the association between incidence of non-Hodgkin's lymphoma and estimated levels of solar ultraviolet radiation, controlling for social class and employment in agriculture.

**Setting**: 59 counties in England and Wales.

**Subjects**: All registered cases of non-Hodgkin's lymphoma during the period 1968-85.

**Main outcome measure**: Age and sex adjusted odds ratio for non-Hodgkin's lymphoma in each county.

**Results**: Incidence of non-Hodgkin's lymphoma was significantly associated with solar ultraviolet radiation levels (P<0.001), even after social class and employment in agriculture were controlled for (P=0.004). In a comparison of counties in the highest and lowest quarters of solar ultraviolet radiation, the relative risk of non-Hodgkin's lymphoma was 1.27 (95% confidence interval 1.24 to 1.29), rising to 1.34 (1.32 to 1.37) after adjustment for social class and employment in agriculture.

**Conclusions**: The incidence of non-Hodgkin's lymphoma in different areas of England and Wales is positively associated with levels of solar ultraviolet radiation. These results are consistent with the hypothesis that exposure to solar ultraviolet radiation increases the risk of non-Hodgkin's lymphoma.

#### Key messages

*   Key messages

*   It has been hypothesised that greater exposure of the population to sunlight, which can cause systemic immune suppression, may have contributed to the increase in non-Hodgkin's lymphoma

*   The geographical pattern of non-Hodgkin's lymphoma in England and Wales tends to match that of solar ultraviolet radiation

*   Exposure to sunlight may increase the risk of non-Hodgkin's lymphoma

*   Risks might be reduced by adoption of a more cautious approach to sun exposure

## Introduction

The incidence of non-Hodgkin's lymphoma has increased greatly in many countries,1 including England and Wales.2 An established risk factor is chronic immunosuppression3 4—as in individuals with congenital immunodeficiency syndromes,5 patients treated with immunosuppressive drugs,6 and those infected with HIV.7 Particular types of non-Hodgkin's lymphoma are also associated with infection with the Epstein-Barr virus and with HTLV-I.8 Risk is also raised in a number of occupations,9 10 notably in agriculture, where exposure to pesticides may be a factor.11 However, trends in known risk factors and in diagnostic practices can account for only a small part of the observed increase in non-Hodgkin's lymphoma.12 13 Part of the rise may have been the result of increased population exposure to some other widespread risk factor.

It has been hypothesised that exposure to solar ultraviolet radiation may be a cause of non-Hodgkin's lymphoma.14 This is an important cause of skin cancers,15 which have been increasing even more rapidly than non-Hodgkin's lymphoma,16 probably as a result of the adoption of lifestyles involving increased exposure to the sun.15 Increased exposure to sunlight might also have contributed to the increase in non-Hodgkin's lymphoma. Cartwright et al note that immunosuppression increases the risk of non-Hodgkin's lymphoma14 and that ultraviolet B radiation has been shown to cause systemic immune suppression in mice and in humans.17 Support for their hypothesis is provided by significant positive associations between the incidence, and change over time of incidence, of non-Hodgkin's lymphoma and non-melanoma skin cancers for a sample of cancer registries.14 This has been strengthened by the results of a study showing that non-Hodgkin's lymphoma patients had an increased risk of subsequently developing squamous cell skin cancer and malignant melanoma. Similarly, there was a subsequent increase in non-Hodgkin's lymphoma among patients with squamous cell skin cancers and malignant melanoma.18 In view of the importance of exposure to the sun for skin cancers, the authors take this to support the hypothesis of a role for ultraviolet radiation in the increasing incidence of non-Hodgkin's lymphoma.18 However, it has been pointed out that this relation was examined only indirectly.19 Also, the increased risk of skin cancer after non-Hodgkin's lymphoma could arise from immunodeficiency as a result of the disease itself or the drugs used to treat it,20 21 although similar factors are unlikely to account for the excess of non-Hodgkin's lymphoma after skin cancers.22 These doubts point to the need for further investigation using a different study design.

Ultraviolet radiation varies from place to place depending on factors such as latitude and cloud cover. Other things being equal, the doses received by the populations of different areas are likely to vary in a similar way.23 The existence of geographical variations in ultraviolet radiation provides an opportunity to test the hypothesis that exposure to sunlight is a cause of non-Hodgkin's lymphoma. If the hypothesis is correct the geographical patterns of non-Hodgkin's lymphoma and ultraviolet radiation will be positively correlated. This study examines this question by using data on non-Hodgkin's lymphoma incidence and estimated solar ultraviolet radiation levels in different areas of England and Wales.

## Methods

### DATA

The Atlas of Cancer Incidence in England and Wales provides information on non-Hodgkin's lymphoma registrations for 59 counties for the period 1968-85.24 In the atlas, age-standardised odds ratios were preferred to incidence rates because they were considered less likely to be affected by variations in the completeness of cancer registration. The odds ratios were obtained by considering the registration data as a series of case-control analyses, comparing risk of non-Hodgkin's lymphoma in a particular county with its risk in all other counties. Cases were the registrations for non-Hodgkin's lymphoma, and controls were a weighted sample of all other cancers.

The number of cases of non-Hodgkin's lymphoma and age-adjusted odds ratios for male and female patients for each county were taken from the atlas, and expected values for each sex were calculated by dividing the observed number of cases by the odds ratio. Data for males and females were combined and an age-sex adjusted odds ratio was calculated.

For the same 59 counties, given the lack of long runs of measurements, estimates of annual erythema producing ultraviolet radiation (in 105J/m2) were calculated from a model that used data on latitude and cloud cover.25 The latitude of the centre of each of 403 local authority districts was taken from official maps. Estimates of cloud cover at these locations for the period 1961-90 were derived from information on hours of bright sunshine.26 District ultraviolet radiation values calculated from the model were aggregated to produce a population weighted estimate of ultraviolet radiation for each county. In recent years depletion of stratospheric ozone may have increased ultraviolet radiation,27 but any such effect would be similar throughout England and Wales and would not affect the comparison between counties.

### ANALYSIS

The possibility of an association between ultraviolet radiation levels and incidence of non-Hodgkin's lymphoma was examined by means of regression analysis of data for individual counties. The dependent variable was the age-sex adjusted odds ratio and the main explanatory variable of interest was the estimate of ultraviolet radiation. Information was also collected on two other explanatory variables which might confound the association. As the incidence of low grade non-Hodgkin's lymphoma in different areas of England and Wales is positively associated with socioeconomic status,28 the percentage of employed men in social classes I, II, and IIIN (the non-manual social classes) was included as an explanatory variable. In view of evidence of an excess risk of non-Hodgkin's lymphoma for agricultural workers,9 10 the percentage of employed people working in agriculture was also used. Both were based on data from the 1971 census.29 Since the odds ratios for non-Hodgkin's lymphoma were based on substantial numbers of cases in each county (minimum 33), regression models were calculated using the ordinary least squares method.

The association was examined further by aggregating data for individual counties into four groups divided at the quartiles of ultraviolet radiation. Observed and expected numbers of non-Hodgkin's lymphoma cases were aggregated for each group and the ratios of observed to expected cases and their 95% confidence interval were calculated, with and without adjustment for social class and employment in agriculture.

## Results

Incidence of non-Hodgkin's lymphoma was significantly positively associated with ultraviolet radiation and social class (table 1), but the association with employment in agriculture was negative and non-significant. The strongest association was with ultraviolet radiation (fig 1). The multiple regression analysis showed a highly significant positive association with ultraviolet radiation. The association with social class remained positive but was not significant, and that with employment in agriculture was negative and non-significant.

View this table:
[Table 1](http://www.bmj.com/content/312/7039/1128/T1)

Table 1 
Regression analysis of relation between odds ratio for non-Hodgkin's lymphoma and estimated ultravio- let radiation, social class, and employment in agriculture

![Fig 1](http://www.bmj.com/https://www.bmj.com/content/bmj/312/7039/1128/F1.medium.gif)

[Fig 1](http://www.bmj.com/content/312/7039/1128/F1)

Fig 1 
Age and sex adjusted odds ratios for non-Hodgkin's lymphoma and estimated annual ultraviolet radiation (105J/m2, erythemally weighted equivalent) in 59 counties of England and Wales, 1968-85

Table 2 gives the ratios of observed to expected non-Hodgkin's lymphoma cases and their 95% confidence intervals for counties aggregated into ultraviolet radiation groups. It shows a clear tendency for the incidence of non-Hodgkin's lymphoma to increase with level of ultraviolet radiation. For the unadjusted data the relative risk of non-Hodgkin's lymphoma for the highest versus lowest ultraviolet radiation group was 1.27 (95% confidence interval 1.24 to 1.29). On adjustment for social class and employment in agriculture, using the coefficients for these variables from the multiple regression equation in table 1, this rose to 1.34 (1.32 to 1.37).

View this table:
[Table 2](http://www.bmj.com/content/312/7039/1128/T2)

Table 2 
Ratio of observed to expected cases of non-Hodgkin's lymphoma 1968-85 by ultraviolet radiation group

## Discussion

This study shows that the incidence of non-Hodgkin's lymphoma in the counties of England and Wales was positively associated with levels of solar ultraviolet radiation and that this was not the result of confounding by social class or employment in agriculture. A study using data from the United States found a negative association between non-Hodgkin's lymphoma and ultraviolet radiation, but this was based on observations for only six areas over a five year period and no account was taken of other factors.19 Though the present study uses cancer registrations over a longer period of time and for a larger number of areas, it has its own limitations.

### LIMITATIONS

Being based on ecological correlation, the results of this study do not guarantee the same association for individuals. The overall explanatory power of the multiple regression model was not high, with the three explanatory variables accounting for only 31% of the variance in the dependent variable. Therefore, confounding of the association between ultraviolet radiation levels and incidence of non-Hodgkin's lymphoma by some other factor cannot be excluded.

Another problem is that the study is based on geographical variations in ultraviolet radiation levels rather than on the doses received by the population. Such doses will depend on variations in the patterns of behaviour of the population as well as on the ultraviolet radiation levels in different areas. Doses are likely to be greater for more affluent populations and where a high proportion of the population is involved in outdoor occupations such as farming.

### EFFECT OF SOCIAL CLASS AND AGRICULTURAL WORK

The results of the multiple regression analysis showed that there was a significant positive association between incidence of non-Hodgkin's lymphoma and ultraviolet radiation levels after social class and employment in agriculture were taken into account.

The simple regression analysis showed a significant positive association between incidence of non-Hodgkin's lymphoma and social class, as in a previous study.28 However, this became non-significant in the multiple regression model, in which ultraviolet radiation levels were the strongest explanatory variable, suggesting that some of the apparent association with social class may be the result of generally higher ultraviolet radiation levels in more affluent southern counties.

The negative association between incidence of non-Hodgkin's lymphoma and employment in agriculture is contrary to expectations based on several case-control studies that have shown an excess risk among farm workers.9 10 However, the observed negative ecological correlation does not necessarily imply lower risk among individuals working in agriculture. It could be that other features of agricultural areas and their population are associated with a lower risk of non-Hodgkin's lymphoma.

### IMPLICATIONS OF THE FINDINGS

In view of the contradictory findings for the United States19 and the limitations of this and previous studies,14 18 19 there is a need for further research before a firm conclusion is possible on whether or not solar ultraviolet radiation is a factor in the aetiology of non-Hodgkin's lymphoma. Nevertheless, the present study adds to the evidence from elsewhere14 18 that supports such a possibility. An aetiological role for ultraviolet radiation could also explain some well established features of the epidemiology of non-Hodgkin's lymphoma. As is the case for malignant melanoma and other skin cancers,15 some of the unexplained rise in the incidence of non-Hodgkin's lymphoma might be the result of increased population exposure to ultraviolet radiation. Higher exposure among more affluent groups might also contribute to the tendency for the incidence of non-Hodgkin's lymphoma to increase with social status.28 This has been observed for malignant melanoma,30 31 where the greater involvement of affluent social groups in outdoor leisure activities may increase risk.32 33

Greater exposure to sunlight could also contribute to the excess risk of non-Hodgkin's lymphoma in several outdoor occupations such as farming, forestry, and construction,9 10 although other factors, including chemical exposures, might also be important.11 Enhanced exposure to ultraviolet radiation could also be a factor in the raised incidence of non-Hodgkin's lymphoma that has been observed in Vietnam War veterans in the United States. It has been hypothesised that this might be the result of exposure to the herbicide Agent Orange, but risk did not differ by level of exposure to the suspected chemicals, although it was associated with length of service in Vietnam.34 Since the whole of Vietnam is within the tropics, service personnel based there could have been exposed to higher levels of ultraviolet radiation than if they had remained in the United States.

If exposure to solar ultraviolet radiation is a cause of non-Hodgkin's lymphoma it might be possible to reduce risk by encouraging people to limit their exposure to the sun. However, increases in the amount of ultraviolet B radiation reaching the surface as a result of stratospheric ozone depletion27 could add to risk.

The climatology data used to estimate ultraviolet levels were supplied by the Climate Impacts Link Project (Department of the Environment Contract PECD7-12-96).

## Footnotes

*   Funding This study was carried out in the Centre for Social and Economic Research on the Global Environment, which is funded by the Economic and Social Research Council.

*   Conflict of interest None.

## References

1.  1.Devesa SS, Fears T. Non-Hodgkin's lymphoma time trends: United States and international data. Cancer Res 1992;52(suppl):5432–40S.
    
    

2.  2.Cartwright R. Changes in the descriptive epidemiology of non-Hodgkin's lymphoma in Great Britain. Cancer Res 1992;52(suppl):5441–2S.
    
    

3.  3.Greiner TC. Epidemiology of non-Hodgkin's lymphoma: a review. Diagn Oncol 1994; 4: 26–32.
    
    

4.  4.Weisenberger DD. Epidemiology of non-Hodgkin's lymphoma: recent findings regarding an emerging epidemic. Ann Oncol 1994:5(suppl 1):S19–24.
    
    

5.  5.Filipovich AH, Mathur A, Kamat D, Shapiro RS. Primary immunodeficiencies: genetic risk factors for non-Hodgkin's lymphoma. Cancer Res 1992;52(suppl):5465–7S.
    
    

6.  6.Kinlen L. Immunosuppressive therapy and acquired immunological disorders. Cancer Res 1992;52(suppl):5474–5S.
    
    

7.  7.Levine AM, Shibata D, Sullivan-Halley J, Nathwani B, Brynes R, Slovak ML, et al. Epidemiological and biological study of acquired immunodeficiency syndrome-related lymphoma in the county of Los Angeles: preliminary findings. Cancer Res 1992;52(suppl):5482–4S.
    
    

8.  8.Mueller NE, Mohar A, Evans A. Viruses other than HIV and non-Hodgkin's lymphoma. Cancer Res 1992;52(suppl):5479–81S.
    
    

9.  9.Scherr PA, Hutchison GB, Neiman RS. Non-Hodgkin's lymphoma and occupational exposure. Cancer Res 1992;52(suppl):5503–9S.
    
    

10. 10.Pearce N, Bethwaite P. Increasing incidence of non-Hodgkin's lymphoma: occupational and environmental factors. Cancer Res 1992; 52(suppl):5496–500S.
    
    

11. 11.Zahm SH, Blair A. Pesticides and non-Hodgkin's lymphoma. Cancer Res 1992;52(suppl):5485–8S.
    
    

12. 12.Holford TR, Zheng T, Mayne ST, McKay LA. Time trends of non-Hodgkin's lymphoma: are they real? What do they mean? Cancer Res 1992;52(suppl):5443–6S.
    
    

13. 13.Hartge P, Devesa SS. Quantification of the impact of known risk factors on time trends in non-Hodgkin's lymphoma incidence. Cancer Res 1992;52(suppl):5466–9S.
    
    

14. 14.Cartwright R, McNally R, Staines A. The increasing incidence of non-Hodgkin's lymphoma (NHL): the possible role of sunlight. Leuk Lymphoma 1994; 14: 387–94.
    
    [PubMed](http://www.bmj.com/lookup/external-ref?access_num=7812197&link_type=MED&atom=%2Fbmj%2F312%2F7039%2F1128.atom) 
    
    [Web of Science](http://www.bmj.com/lookup/external-ref?access_num=A1994PF15800005&link_type=ISI) 

15. 15.Glass AG, Hoover RN. The emerging epidemic of melanoma and squamous cell skin cancer. JAMA 1989; 262: 2097–3000.
    
    [CrossRef](http://www.bmj.com/lookup/external-ref?access_num=10.1001/jama.1989.03430150065027&link_type=DOI) 
    
    [PubMed](http://www.bmj.com/lookup/external-ref?access_num=2795783&link_type=MED&atom=%2Fbmj%2F312%2F7039%2F1128.atom) 
    
    [Web of Science](http://www.bmj.com/lookup/external-ref?access_num=A1989AU86000020&link_type=ISI) 

16. 16.Doll R. Progress against cancer: an epidemiological assessment. Am J Epidemiol 1991; 134: 675–88.
    
    [FREE Full Text](http://www.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6MzoiUERGIjtzOjExOiJqb3VybmFsQ29kZSI7czo3OiJhbWplcGlkIjtzOjU6InJlc2lkIjtzOjk6IjEzNC83LzY3NSI7czo0OiJhdG9tIjtzOjIzOiIvYm1qLzMxMi83MDM5LzExMjguYXRvbSI7fXM6ODoiZnJhZ21lbnQiO3M6MDoiIjt9) 

17. 17.Goettsch W, Garssen J, de Gruijl FR, van Loveren H. UV-B and the immune system. Thymus 1993; 21: 93–114.
    
    [PubMed](http://www.bmj.com/lookup/external-ref?access_num=8337707&link_type=MED&atom=%2Fbmj%2F312%2F7039%2F1128.atom) 
    
    [Web of Science](http://www.bmj.com/lookup/external-ref?access_num=A1993LF66500003&link_type=ISI) 

18. 18.Adami J, Frisch M, Yuen J, Glimelius B, Melbye M. Evidence of an association between non-Hodgkin's lymphoma and skin cancer. BMJ 1995; 310: 1491–5.
    
    [Abstract/FREE Full Text](http://www.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6MzoiYm1qIjtzOjU6InJlc2lkIjtzOjEzOiIzMTAvNjk5My8xNDkxIjtzOjQ6ImF0b20iO3M6MjM6Ii9ibWovMzEyLzcwMzkvMTEyOC5hdG9tIjt9czo4OiJmcmFnbWVudCI7czowOiIiO30=) 

19. 19.Newton R. American data refute ultraviolet hypothesis. BMJ 1995; 311: 750.
    
    [FREE Full Text](http://www.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6MzoiYm1qIjtzOjU6InJlc2lkIjtzOjEyOiIzMTEvNzAwNy83NTAiO3M6NDoiYXRvbSI7czoyMzoiL2Jtai8zMTIvNzAzOS8xMTI4LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

20. 20.Sasieni P, Bataille V. Ultraviolet light is unlikely explanation for association. BMJ 1995; 311: 749.
    
    [FREE Full Text](http://www.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiRlVMTCI7czoxMToiam91cm5hbENvZGUiO3M6MzoiYm1qIjtzOjU6InJlc2lkIjtzOjEyOiIzMTEvNzAwNy83NDkiO3M6NDoiYXRvbSI7czoyMzoiL2Jtai8zMTIvNzAzOS8xMTI4LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

21. 21.Vickers M. Association may be iatrogenic. BMJ 1995; 311: 750.
    
    

22. 22.Adami J. Author's reply. BMJ 1995; 311: 750–1.
    
    

23. 23.World Health Organisation. Ultraviolet radiation. Environmental Health Criteria 160. Geneva: WHO, 1994.
    
    

24. 24.Swerdlow A, dos Santos Silva I. Atlas of cancer incidence in England and Wales 1968-85. Oxford: Oxford University Press, 1993.
    
    

25. 25.Mo T, Green AES. A climatology of solar erythema dose. Photochem Photobiol 1974; 20: 483–96.
    
    [PubMed](http://www.bmj.com/lookup/external-ref?access_num=4457924&link_type=MED&atom=%2Fbmj%2F312%2F7039%2F1128.atom) 
    
    [Web of Science](http://www.bmj.com/lookup/external-ref?access_num=A1974V328300002&link_type=ISI) 

26. 26.Barrow E, Hulme M, Jiang T. A 1961-90 baseline climatology and future climate change scenarios for Great Britain and Europe. Part 1: a Great Britain baseline climatology. Norwich: Climatic Research Unit, 1993.
    
    

27. 27.Madronich S, McKenzie RL, Caldwell MM, Bjorn LO. Changes in ultraviolet radiation reaching the Earth's surface. Ambio 1995; 24: 143–52.
    
    [Web of Science](http://www.bmj.com/lookup/external-ref?access_num=A1995RB82400002&link_type=ISI) 

28. 28.McKinney PA, Alexander FE, Cartwright RA. Epidemiology of non-Hodgkin's lymphoma in parts of England and Wales (1984-88). Leuk Lymphoma 1991; 4: 123–30.
    
    

29. 29.Office of Population Censuses and Surveys. Census 1971. England and Wales. Economic Activity County Leaflets. London: HMSO, 1975.
    
    

30. 30.Kirkpatrick CS, Lee JAH, White E. Melanoma risk by age and socio-economic status. Int J Cancer 1990; 46: 1–4.
    
    [PubMed](http://www.bmj.com/lookup/external-ref?access_num=2365492&link_type=MED&atom=%2Fbmj%2F312%2F7039%2F1128.atom) 
    
    [Web of Science](http://www.bmj.com/lookup/external-ref?access_num=A1990DN99100001&link_type=ISI) 

31. 31.Elwood JM, Whitehead SM, Davison J, Stewart M, Galt M. Malignant melanoma in England: risks associated with naevi, freckles, social class, hair colour, and sunburn. Int J Epidemiol 1990; 19: 801–10.
    
    [Abstract/FREE Full Text](http://www.bmj.com/lookup/ijlink/YTozOntzOjQ6InBhdGgiO3M6MTQ6Ii9sb29rdXAvaWpsaW5rIjtzOjU6InF1ZXJ5IjthOjQ6e3M6ODoibGlua1R5cGUiO3M6NDoiQUJTVCI7czoxMToiam91cm5hbENvZGUiO3M6ODoiaW50amVwaWQiO3M6NToicmVzaWQiO3M6ODoiMTkvNC84MDEiO3M6NDoiYXRvbSI7czoyMzoiL2Jtai8zMTIvNzAzOS8xMTI4LmF0b20iO31zOjg6ImZyYWdtZW50IjtzOjA6IiI7fQ==) 

32. 32.Osterlind A, Tucker MA, Stone BJ, Jensen OM. The Danish case-control study of cutaneous malignant melanoma. II. Importance of UV-light exposure. Int J Cancer 1988;42: 319–24.
    
    

33. 33.Elwood JM, Gallagher RP, Hill GB, Pearson JCG. Cutaneous melanoma in relation to intermittent and constant sun exposure: the western Canada melanoma study. Int J Epidemiol 1985; 35: 427–33.
    
    

34. 34.Selected Cancers Cooperative Study Group. The association of selected cancers with service in the US military in Vietnam. Arch Intern Med 1990; 150: 2473–83.
    
    [CrossRef](http://www.bmj.com/lookup/external-ref?access_num=10.1001/archinte.1990.00390230033005&link_type=DOI) 
    
    [PubMed](http://www.bmj.com/lookup/external-ref?access_num=2244764&link_type=MED&atom=%2Fbmj%2F312%2F7039%2F1128.atom) 
    
    [Web of Science](http://www.bmj.com/lookup/external-ref?access_num=A1990EM10800005&link_type=ISI)