UNEMPLOYMENT AND MORTALITY IN THE OPCS LONGITUDINAL STUDY

1324

received municipal water, the risk was doubled if the family shared a toilet with at least one other family. A mass vaccination programme with OPV was undertaken by health authorities. Initially, the programme was directed at children under 5 years of age and, subsequently, at those up to 15 years of age. Following this mass vaccination programme assessments of immunisation levels in the six areas indicated that more than 50% of 12 to 35 month old children had received at least one dose of OPV during the control programme. The mass vaccination programme resulted in an overall coverage of 91 to 99% for two or more doses of OPV. An average of 59% of the children surveyed who had not had poliovaccine previously received at least one

Department; Ms L. C. Hsu, National Institute of Preventive Medicine; Prof C. Y. Lee, National Taiwan University College of Medicine; Dr C. L. Chen, Taipei Municipal Women and Children Hospital; Dr H. C. Wang, Veteran General Hospital, Taipei; Dr F. Y. Huang, Mackay Hospital, Taipei; Dr S. L. Chao, Cheng Hsin Rehabilitation Center, Taipei; Rev Georges Massin, St Joseph’s Hospital, Lumpel, Yun Lin County; and Dr R. P. Beasley and Dr Lu Yu Huang, University of Washington Medical Research Unit, Taipei. We also thank the many other individuals and institutions in the districts, counties and cities throughout Taiwan who contributed to the collection and analysing of data on clinical cases and vaccine coverage assessments; and Mr J. R. Lilley, Mr S. Brooks, and Ms S. M. Chalmers and their colleagues at the American Institute in Taiwan for their support during our stay in Taiwan; and Ms Connie Keith for preparing the typescript. Correspondence should be addressed to K. J. B., Surveillance, Investigations and Research Branch, Division of Immunisation, Center for Prevention Services, Centers for Disease Control, Atlanta, Georgia 30333,

dose

USA.

during the mass campaign.

REFERENCES

DISCUSSION

1. Health Statistics I. General Health Statistics 1980. National Health

Failure to vaccinate rather than vaccine failure was the most important risk factor for paralytic poliomyelitis. Ensuring that children are vaccinated at the earliest recommended age offers the greatest chance for protection. The efficacy of OPV has usually been assessed by serum antibody response to vaccination, a method that does not give a measure of intestinal immunity. Theoretically, such intestinal immunity may be present even in the absence of systemic antibody response or vice versa. In addition, immune persons may have low levels of systemic antibody that may not be detected by the laboratory method used, and delayed seroconversion may be missed if the serum was collected too early. The best measure of effectiveness of any vaccine is its ability to prevent clinical disease. The clinical efficacy of the vaccine against paralytic disease due to type 1 poliovirus during this outbreak is consistent with levels of seroconversion after vaccination in the United States (92- 100 and 97-100% for 2 and 3 doses, respectively) and is considerably higher than serological estimations of efficacy in the developing world.bMeasurement of clinical efficacy in tropical countries may be a useful adjunct to serological assessment and can be done without sophisticated laboratory support. Clinical efficacy can help in the interpretation of

determining

serological findings. This outbreak shows that major epidemics can occur in places that have been practically free of poliomyelitis for many years and that have high overall community vaccination levels. Overall community vaccination levels can mask the fact that there may be pockets where vaccination levels are considerably lower. The cluster-sampling method may not detect groups of susceptible subjects living within the geographic area being assessed. These groups of unrecognised susceptible subj ects may be sufficient to sustain transmission of wild poliovirus in a community. Such susceptible subgroups were responsible for the last two poliomyelitis outbreaks in the United States in 1972 and 1979 and in the Netherlands and Canada in 1978 among religious 7

groups declining vaccination. The EPI vaccine coverage assessment is a valuable tool for examining immunity levels in a community. When an assessment shows that vaccination levels are low, then additional vaccination efforts are indicated. If, however, the assessment shows that levels are high, it may be appropriate to do additional investigations ofsubpopulations suspected to be at high risk, especially if the population is heterogeneous. We thank the following for their help: Dr T. C. Hsu, Dr T. Y. Lee, Mr K. H. Hsu, Dr C. 1. Ma, Mr Su Mei Hsu, Mr Shiu Loung Lin, and Miss Fei Fung Lin, Department of Health; Dr T. Y. Kuan, and Dr Y. C. Ko, Taiwan Province Health Department; Dr C. T. Wu, Kao-Hsiung City Health Department; Dr T. H. Wei and Dr C. H. Lee, Taipei City Health

2.

3.

4 5. 6

7

Administration,

Taiwan Provincial Health Department, Taipei City Health Department, KaoHsiung City Health Department, 1981: 316 Chen CL, Chu W, Lee KS. Survey on the incidence of poliomyelitis in Taichung City, Maternal and Child Health in Taiwan, Taiwan Provincial Maternal and Child Health Institute, 1963: 66-73 Henderson RH, Sundaresan T. Cluster sampling to assess immunization coverage a review of experience with a simplified sampling method. Bull WHO 1982, 60: 253-60. Cox DR. Analysis of binary data. London. Chapman and Hall, 1970 1-142. Bishop Y, Fiemberg S, Holland T Discrete multivariate analysis Cambridge. MIT Press, 1975 401-33 Katona P, Jones TS. Operational aspects of the use of oral poliovirus vaccine in developing countries. In: Recent advances in immunization. Pan American Health Organisation, 1983; 18-29. Centers for Disease Control. Poliomyelitis surveillance Summary 1979 1981; 14-19

Community Health UNEMPLOYMENT AND MORTALITY IN THE OPCS LONGITUDINAL STUDY K. A. MOSER D. R. JONES Social Statistics Research Unit,

A. J. Fox

City University,

London EC1V 0HB

The mortality of men aged 15-64 who were seeking work in the week before the 1971 census was investigated by means of the OPCS Longitudinal Study, which follows up a 1% sample of the population of England and Wales. In contrast to the current position, only 4% of men of working age in 1971 fell into this category. The mortality of these unemployed men in the period 1971-81 was higher (standardised mortality ratio 136) than would be expected from death rates in all men in the Longitudinal Study. The socioeconomic distribution of the unemployed accounts for some of the raised mortality, but, after allowance for this, a 20-30% excess remains; this excess was apparent both in 1971-75 and in 1976-81. The data offer only limited support for the suggestion that some of this excess resulted from men becoming unemployed because of their ill-health; the trend in overall mortality over time and the pattern by cause of death were not those usually associated with illhealth selection. Previous studies have suggested that stress accompanying unemployment could be associated with raised suicide rates, as were again found here. Moreover, the mortality of women whose husbands were unemployed was higher than that of all married women (standardised mortality ratio 120), and this excess also persisted after allowance for their socioeconomic distribution. The results support findings by others that unemployment is associated with adverse effects on health.

Summary

1325 INTRODUCTION

PERHAPS because of the steep rises in the late 1970s in the proportion of the working populations of western countries who were unemployed, several groups have been trying to assess the impact of unemployment on health. The published work, ranging from aggregated, national data to detailed caseI reports, has lately been reviewed by Brenner and Mooney,’ Warr,2 and Cook and Shaper.3 Although Brenner’s econometric studies,4,5 which seek to explain variations in annual UK mortality rates in terms of the unemployment rate and various other measures of economic growth, have received wide attention, investigation of the impact of unemployment on the health of an individual (and his/her family) requires a disaggregated study design. Few results from epidemiological studies of adequate power, specifically designed to measure the mortality consequences of unemployment, have been reported. Small investigations of groups of men being made redundant have been inconclusive. In larger studies, such as the Regional Heart Study,6 the Department of Health and Social Security cohort study,7,8 and the Office of Population Censuses and Surveys (OPCS) Longitudinal Study reported on here,9the relation between unemployment and mortality has not been the primary interest. As a result, weaknesses of study size, outcome and explanatory variables, control-group selection, and response and follow-up rates have reduced the weight to be attached to the results. Although there is some evidence from these studies of raised morbidity and mortality among the unemployed, potential confounding factors remain unmeasured, and inevitably the causal mechanism remains unestablished. Some light has been shed on the morbidity consequences (in particular the psychological consequences) of unemployment in case studies and other research entailing detailed interviewing of small, but high-risk, groups (see Warr2). Whilst these investigations are likewise subject to methodological limitations, they suggest that, at least in some population subgroups, being, and in particular, becoming unemployed is associated with increased morbidity, including depression, anxiety, and stress-related behaviour such as smoking and alcohol consumption. Age, sex, occupational group, and length of unemployment are among the effect-modifying factors suggested. In this paper we use data from the OPCS Longitudinal Study (LS) to examine further the relation between unemployment and mortality. Unemployed men in the LS sample have already been shown to have high mortality in 1971-75 and several possible explanations have been offered.9 Firstly, men’s health may suffer as a result of unemployment, perhaps owing to a fall in income and social

status, increased stress, and consequent behaviour. Secondly, men in poor health may be more likely to become unemployed, and the raised mortality of unemployed men may simply

reflect their health

status

before unemployment. Thirdly, the

the social distribution of high mortality may unemployed men before unemployment and the strong relation between mortality and measures of socioeconomic reflect

status.

In this paper we examine the importance that can be attached to these three explanations by (i) using mortality data for the ten years 1971-81; controlling for the socioeconomic distribution of the unemployed men; and (iii) looking at the mortality of women married to unemployed men. The interpretation ofour analysis is limited because our indicator of unemployment relates to one week in April, 1971, and we have no information on how long these men had been or were subsequently, unemployed.

(ii)

SUBJECTS AND METHODS

Source

of Data

The LS is based

1o sample of individuals enumerated in the 1971 census. Census records for sample members have been linked with information on subsequent events about which details are routinely collected, principally births, deaths, and cancer registrations. This analysis focuses on deaths in the period 1971-81. Some census information on other persons in any household containing an LS member is also linked to the information about the sample member. In this analysis the unemployed group comprises those men who indicated, in response to the 1971 census question on economic position, that they were seeking work or waiting to take up a job in the week before the census; we shall refer to them either as "seeking work" or as unemployed. This excludes other categories of economic position such as in employment, temporarily or permanently sick, and retired or otherwise inactive. Perception of economic position is dependent on many factors-primarily age, sex, and (especially for women) marital status and prevailing socioeconomic climate. Women reporting themselves as "seeking work" were a select group; 38% of women aged 15-59 in our sample were placed in the inactive category to which housewives were allocated. Consequently, we have limited ourselves here to an investigation of male unemployment and, principally, to mortality of men in the working age range 15-64. Of the quarter ofa million men in the sample in 1971, 161 699 were aged 15-64 and, of these, 5861 (3 - 6%) were "seeking work". The 1971 census provided some sociodemographic information on respondents. We were able to classify men "seeking work" into social classes from details of their most recent jobs. The LS also contains information on the household and other household members, and we have used this to investigate the mortality of women married to men "seeking work".

England and Wales

on a in

Methods

Throughout the analysis the standardised mortality ratio (SMR, the ratio of observed [0] to expected [E] deaths x is used as a summary index of mortality. Expected deaths are obtained by applying the death rates by 5-year age groups in the standard group to the person-years-at-risk in the study group In most of the analyses the standard group comprises all men in the LS aged 15 and over at census. Where a different standard group is used this is indicated in the text. Approximate 9507o confidence limits for the ratio of observed to expected deaths are also.presented.i

100)

RESULTS

The SMR for 1971-81 for

seeking work in 1971, Fig I-Mortality 1971-81 ofmen seeking work in 1971 by age at death.

men

was

confidence limits 122-152. It

aged 15-64 years at death, approximate 95%

136 with rose

from 129 for 1971-75

to

1326 TABLE I-MORTALITY

1971-81OF ALI. MEN AND MEN SEEKING

WORK IN

1971, BY SOCIAL CLASS

Fig 2- Percentage distribution of all men and men seeking work aged 15-64

by social

I-V=social

occupations;

class in 1971.

AF=armed Unoce unoccupied.

class;

forces;

ID=inadequately

described

Figures in parentheses are numbers of observed deaths. tenure-an

144 for 1976-81. Although the mortality of these men was raised at all ages, the excess seems greater at younger ages (fig 1). In age groups 25-34, 35-44, and 45-54 the SMR exceeded 170.

Effects of Socioeconomic Distribution How much of the excess mortality among the unemployed men can be explained by their socioeconomic distribution? . The difference between the social class distribution of

unemployed men and all men in the LS is shown in fig 2. Unemployed men were concentrated in the lower social classes. For example, social class V contained 16% of all unemployed men but only 7% of all men. For one-fifth of the unemployed it was not possible to allocate them to a social class on the basis of their most recent job; this is the case for less than 2% of all men. Those allocated to this "inadequately described" group mainly comprised men whose failure to complete the census question on most recent occupation reflected either the fact that they were out of work or that they were enumerated in institutions such as hospitals. There was a strong mortality gradient with social class among all men (with a social class I SMR of 73 and a social class V SMR of 120), which was also present, indeed wider, among men "seeking work" (table I). Within each social class the SMR for men "seeking work" was higher than that for all men, indicating that the raised mortality of the unemployed was maintained through all the classes. As would be expected, this is not so for the residual group, the "inadequately described"; many of this group were in hospitals and other institutions, so those among them who were "seeking work" would have been a comparatively healthy subset. To take account of the social class distribution of the new values for the expected deaths were calculated, with standardisation for age and class. The SMR for all unemployed men, standardised for age and class, was 121 with approximate 95% confidence limits 108-135; this suggests that some but not all of the excess mortality among unemployed men may be explained by their class distribution. Standardisation for the distribution of unemployed men by social class but with exclusion of men with "inadequately described" occupations suggests that the component of the raised mortality of men "seeking work" which could be explained by their social class composition was somewhat less than indicated above.12 Standardisation for housing

unemployed,

alternative

measure

of

socioeconomic

status9-reduced the SMR of men "seeking work" from to

127

136

(table II).

The increase in SMR in unemployed men, from 129 in 1971-75 to 144 in 1976-81, seemed to be accounted for by the social class distribution; SMRs standardised for age and social class were 122 and 123 for the two time-periods. The effect of standardising the SMRs for social class was greater in 1976-81 than in 1971-75 because there was a steeper social class gradient in mortality in 1976-81 than in 1971-75. This results largely from dissipation of the effects of selective health-related mobility out of employment which affected mortality differentials in the earlier period.13 Mortality seems to have been particularly high for malignant neoplasms and for accidents, poisonings, and violence (table III). Raised mortality from malignant neoplasms (0=94, E=66’5) appears to have been attributable in the main to deaths from lung cancer ( =48, E=27’4). A clear excess from this cause remained after allowance for social class. High mortality from accidents, poisonings, and violence (0 = 46, E 22 - 8) was only partly explained by the very high mortality from suicide (0 20, E 8 - 3). Although SMRs for both these causes were reduced by social class standardisation, substantial excesses remained. =

=

=

Effects of Health-related Selectaon

in the

Unemployed Group We now examine evidence for the suggestion that the men who were unemployed in 1971 became unemployed because of poor health. 14 If ill-health were a major influence on the selection of this group we would expect the health, and the relative mortality, of the group to improve over time. Such a. change would be expected because of the high initial

TABLE II-MORTALITY AGED

15-64 AT DEATH,

1971-81

OF MEN SEEKING WORK IN

STANDARDISED FOR AGE AND SOCIAL

AND AGE AND HOUSING TENURE

1971, CLASS,

1327 TABLE III-MORTALITY AGED

15-64

1971-81OF MEN

SEEKING WORK IN

1971,

AT DEATH, BY CALSE OF DEATH

TABLE IV-MORTALITY

WORK

IN

1971-75

AND

1976-81

OF MEN SEEKING

1971, AGED IS-64 AT DEATH, BY CAUSE OF DEATH

in parentheses are numbers of observed deaths. *ICD 8th revision 850-877, 942, 950-959, 980-989.

Figures

mortality of those selected on the basis of ill-health; the proportion of sick men in this category would decline over time. This and other mechanisms would lead to a reduction in mortality over time, as has been observed in other areas of our work.13.15 On the other hand, if this group were selected on the basis of positive health, we would expect the health of the group to worsen, and their relative mortality to rise with time. A further component of the analysis involves examination of patterns of cause-specific mortality over time, since the time scales for changes in mortality for acute causes would be shorter than those for chronic causes if the selection mechanism were valid. Although we have 10 years’ mortality data, this is a short time-span over which to observe trends in cause-specific mortality for such a small subgroup of the population. The lack of power of this analysis of short-term trends by cause of death is apparent from the wide confidence intervals around the SMRs in table IV. As we have said, the SMRs for all causes of death standardised for social class showed no trend between the two time periods. The fall in SMRs (standardised for social class) for lung’, cancer and ischaemic heart disease might be construed as evidence of ill-health selection, but they were not based on sufficient numbers of deaths for firm conclusions to be drawn about their trends with time. Interpretation of the pattern of deaths from bronchitis is even more problematical. Although these data provide no strong evidence for a selection effect, we cannot rule out the possibility that there is one operating. The complexity of the hypothesis and the large sampling variation make conclusive interpretation of the data difficult. ’

Figures

in

parentheses

are

numbers of observed deaths,

interpreted as more directly attributable to the effects of unemployment. Table v shows the mortality in 1971-81 of the 2906 married women in private households whose husbands were unemployed at the 1971 census. The standard population used in calculating the expected deaths was all married women resident in private households. The overall SMR for 1971-81 was 120 (approximate 95% confidence intervals 102-139), which suggests that mortality among this group of women was higher than would have been expected. The SMR

rose

from 108 in 1971-75

to

129 in

1976-81; the

approximate 95% confidence intervals were 83-138 and 105-154, respectively. Although some cause-specific SMRs were raised, the numbers of deaths from specific causes were in the main too small to make any clear interpretations of the results. However, for ischaemic heart disease the SMR was 157 with 95% confidence intervals 115-206. There was apparent trend over time for this cause; the SMR for 1971-75 was 155 and that for 1976-81 was 159. As with the excess mortality for unemployed men, part of the excess mortality among women whose husbands were

approximate no

TABLE V-MORTALITY

1971-81 OF WOMEN

SEEKING WORK IN

1971,

WHOSE

HUSBANDS

WERE

BY CAUSE OF DEATH

——————————————————& mdash;—————————————————&m dash;——————————————

Effects of Unemployment on

the Health

of Other Household

Members

Any adverse effects of unemployment-through, for example, a fall in income, or an increase in stress-may be expected to have repercussions on all members of the unemployed person’s family or household. By examining mortality among people other than the unemployed man himself we partly eliminate any health-selection effect, unless the ill-health of others in the household was associated with his being less available for work. Any adverse health experience among these family members could therefore be

Figures in parentheses are numbers of observed deaths. The standard population used in calculating the expected deaths was all married women resident

in

private households.

1328 may have been due to the socioeconomic composition of this group. The data on tenure suggest that a high proportion of these women lived in local-authority housing; 45% lived in council housing and 34% owned their houses as compared with 2707o and 56%, respectively, among married women whose husbands were in employment. The tenure mortality gradient was far steeper among women with unemployed husbands (owner occupiers had an SMR of 101 while council tenants had an SMR of 144) than among women with husbands in employment (corresponding SMRs 90 and 114). Within each tenure group the women with unemployed husbands had considerably higher mortality; the overall SMR of these women was 124, compared with 100 for women whose husbands were in employment. However, standardisation for tenure distribution reduced the SMR only from 124 to 121. This suggests that housing tenure explained little of the excess mortality among women with unemployed husbands.

unemployed

DISCUSSION

Limitations and StrerLgths

of the Analysis

This paper has been concerned with male unemployment and its relation with the mortality of unemployed men and that of women married to unemployed men. Unfortunately we have not been able to consider female unemployment and its effect on health, mainly because of difficulties in interpreting responses to the question on economic position in the 1971 census. Although we have followed about 6000 men unemployed in 1971 for up to 10 years, the 328 deaths which occurred amongst this group were too few in number to enable us to test specific relations satisfactorily-for example, for selected causes of death and by year of death. In particular, the assessment of evidence for a health-related selection effect

severely hampered. unemployment relates to one week in and we have no information on how long these April, 1971, men had been, or were subsequently, unemployed. The principal strength of the investigation is that it is a prospective study of individuals who were unemployed in 1971, rather than a study of histories collected retrospectively, or an analysis of aggregate data. In this extension was

Our indicator of

of the earlier analysis9 we have examined the social class composition of unemployed men to establish the extent to which the raised mortality of these men is explained by social class differentials in mortality. As well as looking at the mortality among unemployed men, we have now started to investigate the mortality experienced by other members of households containing an unemployed man. This is useful not only because it enables us to assess the wider health effects of unemployment but also for the light it sheds on the role of health-related selection, the weakest element of our analysis ofmortality among the unemployed themselves. These issues are discussed more fully in a working paper. 12

Findings Among men who were "seeking work" in 1971, our data indicate high mortality (SMR 136) over the next 10 years. The analysis suggests that some of this excess mortality may be explained by the fact that unemployed men were more concentrated in social classes IV and V; nonetheless, a 20-30°7o excess remained unexplained. However, it should be noted that, without details of the timing of unemployment

in relation to social mobility, it is difficult to establish the direction of the causal relation. The evidence in support of some health-related selection of unemployed men remains very unclear, although it seems probable that men "seeking work" were partly selected for good health since the least healthy men, who were not in employment, would have been recorded either as "out of work, sick" or "permanently sick". Two causes of death, lung cancer and suicide, stand out as having significantly raised levels of mortality among men "seeking work" after allowance for social class. Both causes have been linked elsewhere to stress and stress-related activity.I6,]7 However, to explain an excess oflung cancer mortality in 1971-81, one must probably invoke exposure to a risk factor before 1971. Women whose husbands were "seeking work" had raised mortality (SMR 120 compared with all married women resident in private households), most of which remained after controlling for tenure distribution. As any health-related selection effect is expected to be small among women whose husbands were unemployed-indeed the SMR rose with increased follow-up-it is reasonable to suggest that their high mortality was largely attributable to other factors, such as a direct effect of their spouses’ unemployment. The excess mortality from ischaemic heart disease may support this hypothesis since this disease has been linked with stress.]8 Further work now in progress on the mortality of other members of households where there was an unemployed man in 1971 should therefore be of particular interest. Between 1971 and 1981 unemployment became a more common experience among men in the working age range and durations of unemployment increased. These changes make it difficult to extrapolate from our findings to estimate the impact of unemployment on health today. Once we have access to mortality data for the 1980s "and information from the 1981 census on the employment status and socioeconomic circumstances of the LS sample members at this second point in time, we should be able to assess the changing health effects of unemployment. To summarise, although effects of other factors remain to be investigated, the results of this investigation do provide evidence suggesting that some of the excess mortality among unemployed men may be explained by their socioeconomic circumstances before unemployment. However, this alone does not account for all of the high mortality in unemployed men and in women married to unemployed men. The analyses are part of a review by the Social Statistics Research Unit at City University of mortality data available from the OPCS Longitudinal Study. (Crown copyright is reserved.) This programme is supported by a grant from the Medical Research Council. The views expressed are those of the

authors.

REFERENCES

MH, Mooney A Unemployment and health change Soc Sci Med 1983; 17: 1125-38.

1 Brenner

in the context of

economic

2. Warr P Twelve questions about unemployment and health. In Roberts B, Finnegan R, Gallie D, eds. New approaches to economic life Manchester University Press

(in press) 3. Cook D, Shaper AG Unemployment and health In: Harrington M, ed. Recent advances in occupational health, vol. 2. Edinburgh Churchill Livingstone (in

press).

Mortality and the national economy a review, and the experience of and Wales, 1936-1976. Lancet 1979; ii: 568-73. 5 Brenner MH. Unemployment and health Lancet 1979, ii 874-75. 6 Cook DG, Cummins RO, Bartley MJ, Shaper AG Health of unemployed middle-aged men in Great Britain Lancet 1982; i: 1290-94. 7. Wood D The DHSS cohort study of unemployed men (working paper no 1) London. Department of Health and Social Security, 1982. 4. Brenner MH.

England

1329

DNA in Medicine AGRICULTURAL PRODUCTION M. S. SWAMINATHAN International Rice Research Institute, PO Box 933, Manila, Philippines

AN impact on illness worldwide at least as great as that from the diagnostic and therapeutic possibilities aired by previous contributors to this series may well come from the application of DNA technology, not to man directly, but to the plants

(and animals) he eats. Agricultural advances

in recent years have been triggered of the development by high-yielding breeds of crops and livestock. The breeding of strains possessing a broad spectrum of desirable characters involves extensive hybridisation, followed by rigorous selection. The enhanced ability for gene transfer across sexual barriers opened up by recombinant DNA technology is thus an important gain. Gene transfer at the molecular level helps to broaden the variability available to breeders and assists in the introduction of specific characters controlled by one or a few genes into established cultivars, without much disturbance to the

adapted genetic background. The emphasis in this article will be on crops, especially the staples such as rice and wheat. However, the principles of specific gene transfer are much the same for animal husbandry as for plants. Other contributors to this series have discussed the applications of monoclonal antibodies and recombinant DNA technology to diagnosis and vaccine development in human medicine-and veterinary medicine stands to gain too. THE CONTRIBUTION OF PRE-MOLECULAR GENETICS

Before

discussing the potential applications of genetic engineering and other tools of molecular biology in agriculture, I would like to emphasise that much has been and continues to be accomplished by use of the conventional, nonmolecular tools of genetics. For example, improved varieties of major staples coupled with better crop management techniques have led to food production in Asia and Latin America remaining above population growth (fig 1). I would like to illustrate this for rice. In rice, which belongs to the genus Oryza, there are some twenty about 120 000 distinct genotypes. With-the help of scientists worldwide 74 000 of these strains are being preserved at the International Rice Germplasm Center here in Manila. This material is screened for over fifty characters and the data are computerised. On the basis of such studies, over 5000 crosses are made each year by plant breeders at this Institute. The segregating populations from the F2 generation are selected for resistance to a

species and

K. A. MOSER AND OTHERS: 8. Ramsden S, Smee C. Thehealthof

unemployedmen: DHSScohortstudy. Employment Gazette September, 1981, 397-401. 9. Fox AJ, Goldblatt PO. Socio-demographic mortality differentials: longitudinal study 1971-75, series LS no 1 London: HM Stationery Office, 1982. 10. Berry G. The analysis of mortality by the subject-years method. Biometrics 1983; 39: 173-84. 11 Vandenbroucke JP. A shortcut method for calculating the 95 per cent confidence interval of the standardised mortality ratio. Am J Epidemiol 1982; 115: 303-04. 12 Moser KA, Fox AJ, Jones DR, Goldblatt PO. Unemployment and mortality in the OPCS longitudinal study. London: SSRU, City University. Working paper no 18, 1984.

1-Trends in per caput food production in Asia, Latin America, and Sub-Saharan Africa, 1961-65 average and 1965-83.

Fig

(Source:

Food and

Agriculture Organisation, Rome.)

variety of pest and disease reactions and ability to cope with unfavourable soil conditions. The segregating populations are often grown in "hot spots" in countries where, for example, there is severe natural infestation by specific pests. The rice variety IR36 is popular with farmers because it is resistant to many of the important pests and diseases and is tolerant to alkaline, saline, and zincdeficient soils. The pedigree of IR36 (fig 2) illustrates the enormous amount of breeding work involved and the large number of parents that have gone into the making of this variety, including a wild species, 0 nivara from India, which is resistant to grassy stunt virus. The segregating populations which gave rise to IR36 were also grown in west Sumatra in Indonesia for selection for resistance to tungro virus and in Orissa, India, to identify plants resistant to gall midge. We have yet to exhaust the potential of classical genetics for improving crop and animal productivity. The new tools of molecular genetics are thus not substitutes for the techniques already available to breeders but potentially valuable supplements. Land is a shrinking resource for agriculture, and we are going to have to produce more and more food on less and less land. This will be possible only if we can achieve a continuous rise in crop and animal productivity without harming the long-term production potential of soil and water. We have to develop high-yield techniques for ecologically handicapped areas such as arid, semi-arid, and flood-prone regions. We have to help the small farmers of developing countries to produce more at minimum cost. This means helping him to substitute non-cash inputs for purchased inputs-eg, by growing his own nitrogen through Azolla (a tiny aquatic fern that is symbiotic with the nitrogen-fixing

REFERENCES—continued Fox AJ, Goldblatt PO, Jones DR Social class mortality differentials. artefact, selection or life circumstances?J Epidemiol Commun Health (in press). 14. Stern J The relationship between unemployment, morbidity and mortality in Britain. Population Studies 1983; 37: 61-74 15. Fox AJ, Goldblatt PO, Adelstein AM. Selection and mortality differentials. J Epidemiol Commun Health 1982; 36: 69-79 16. Cooper CL. Psychosocial stress and cancer Bull Br Psychol Soc 1982, 35: 456-59. 17. Plait S. Unemployment and suicidal behaviour: a review of the literature Soc Sci Med 1984, 19: 93-115. 18. Jenkins CD Recent evidence supporting psychologic and social risk factors for coronary disease. N Engl J Med 1976, 294: 987-94, 1033-38 13.