Q J Med 2000; 93: 669-675
© 2000 Association of Physicians
Use of large medical databases to study associations between diseases
From the Unit of Health-Care Epidemiology, Institute of Health Sciences, University of Oxford, Oxford, UK
 |
Summary |
|---|
 Top Summary IntroductionMethodsResultsDiscussionReferences |
|---|
We describe the use of a dataset of statistical medical records, the Oxford Record Linkage Study (ORLS), to identify diseases which occur together more commonly (association), or less commonly (dissociation), than their individual frequencies in the population would predict. We investigated some conditions known or suspected to enhance the subsequent risk of cancer, some conditions thought to be linked with schizophrenia, and some associations between conditions with a known autoimmune component. Diseases may occur in combination more often (or less often) than expected by chance because one predisposes to (or protects against) another or because they share environmental and/or genetic mechanisms in common. The investigation of such associations can yield important information for clinicians interested in potential disease sequelae, for epidemiologists trying to understand disease aetiology, and for geneticists attempting to determine the genetic basis of variation in disease course among individuals. We suggest that, through the use of datasets like the ORLS, it will be possible to ‘map’ comprehensively the phenomic expression of co-occurring diseases.
|
Introduction |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
The importance of collecting and analysing information about episodes of illness has long been recognized. In England, statistical data have been collected from hospital admission records since 19491 and from death certificates since 1837.2 One limitation of these data is that, in general, information about different episodes of illness for the same individuals cannot be linked together, precluding their use for studying the occurrence of successive events in individuals. In the Oxford health region of England, however, linkable hospital admission statistics and death certificates have been maintained from the 1960s,3,4 and statistical files in the Oxford Record Linkage Study (ORLS) have now been linked across a 32-year period. The advent of cheap, powerful data processing means that extensive analysis of very large medical databases is now eminently feasible. One potentially important use of linked statistical medical records is to determine whether particular diseases occur together in the same individuals more often than expected by chance (association), less often than expected by chance (dissociation), or as often as expected from their individual frequencies in the population.
Medical conditions may be associated, or dissociated, through a number of mechanisms. One condition may predispose to, or protect against, the other. Conditions may be associated or dissociated through related factors such as treatment: for example, treatment for one condition may predispose to, or protect against, another. Diseases may also be associated because they share genetic and/or environmental aetiologies.
The aims of this article are to describe the use of a large medical database to map associations between medical conditions; to illustrate this with brief description of examples; and to stimulate ideas about disease associations worthy of study through the use of such databases.
|
Methods |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Population
The ORLS includes statistical abstracts of records of births, hospital admissions, cancer registrations, and deaths in a defined population in southern England. The area covered had a population of about 350 000 in 1963, and it expanded to cover 900 000 from 1966, 1.9 million from 1975, and 2.5 million (all eight health districts in the former Oxford health region) from 1987. The statistical abstracts for each individual have been linked together across the years 1963 to 1994. Personal identifiers are removed from the linked file prior to analysis.
Analytical design and selection of clinical conditions
In the studies outlined in this paper, the method of analysis was a series of nested case-control studies. The ‘cases’ comprised all those individuals with the condition of interest recorded on a hospital admission record, cancer registration record, or death certificate. The ‘controls’ were individuals who had been admitted to hospital for any one of a wide range of minor medical or surgical conditions. Approximately 600 000 individuals were included as controls for each comparison. The prior ‘exposure’ condition was that which we wished to study in association with the case condition.
Selection procedures for dealing with multiple diagnoses per admission and multiple admissions per individual were as follows. Any given hospital admission record may include more than one diagnosis or operation. We classified individuals as cases if the case condition was recorded at any position on their admission record. In contrast, individuals were only classified as controls if the control diagnosis or operation was recorded as the principal one on the admission record. The logic of this was to avoid selecting as controls people with other, more major conditions. Over time, an individual could have more than one medical condition requiring admission to hospital. Individuals with both a case condition and a control condition were selected as cases (and removed from the control group). For individuals with more than one control condition, the episode of care which was included in the analysis as the control event was chosen randomly. An individual could have more than one hospital admission for any given case or control condition. In this instance, the date of the first admission was selected as the date of the ‘index’ event. Finally, to reduce the possibility of Berkson's bias,5 we excluded anyone with just a single episode of admission on which both the exposure and either the case or control condition were recorded. In all analyses, the data were stratified by year of occurrence of the case or control index event, the person's district of residence, their age in 5-year age bands, and sex.
Calculation of ‘expected’ values
The occurrence of the prior condition of interest in the control group was used to generate ‘expected’ numbers of people with the condition in the case group. The rate of occurrence of the exposure condition in the controls was calculated in each stratum, i.e. by age, sex, year of occurrence, and district of residence. These stratum-specific rates were applied to the number of people in each corresponding stratum in the case group, and then summed, to give an ‘expected’ total number of people with the exposure condition based on the controls’ experience. This is equivalent to the conventionally-used method of indirect standardization in comparing rates in epidemiology.6 The numbers of individuals with case, control and exposure conditions, and the International Classification of Diseases codes used for cases and controls, are available upon request from the authors.
The expected number was compared with the observed number for each pair of conditions in combination. For associations between diseases which we hypothesized might result from shared aetiology, rather than because the first condition predisposed an individual to the second condition, the combinations were sought both ways round (e.g. individuals with asthma and their controls were studied for prior occurrence of hypothyroidism; and individuals with hypothyroidism and their controls were studied for prior asthma).
A Poisson distribution was assumed for the observed values, and confidence intervals for the standardized incidence ratios were calculated accordingly.7
|
Results |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
Conditions prior to cancer of the colon
Table 1
shows the well recognized association between cancer of the colon and ulcerative colitis.8–10 The risk of developing cancer of the colon after Crohn's disease or diverticular disease is less well defined in the literature.10–12 Our data suggest that the risk for people with Crohn's disease is similar to that for ulcerative colitis; and our estimate of a two-fold increase in risk in people with diverticular disease is in agreement with an earlier report.12
|
The hypothesis that cholecystectomy might increase the risk of colon cancer, as a consequence of changes to bile flow through the intestine, has been controversial.13–15 The small and non-significant elevation of risk found in our data (Table 1
), when cases at all time intervals after cholecystectomy were included, is attributable to cases where cancer was diagnosed shortly after cholecystectomy. It is likely that in some people abdominal pain was initially diagnosed as gallstones, when in fact these people subsequently proved to have cancer. Our data on longer time intervals between operation and disease—intervals with sufficient latent period for a causal association to be plausible—show no increase in long-term cancer risk following cholecystectomy (Table 1).
Vasectomy prior to testicular and prostatic cancer
It has been suggested that testicular and prostatic cancer could be long-term adverse effects of vasectomy.16 Our data (Table 1) are in agreement with more recent reports, which find no evidence of elevated risk of either type of cancer after vasectomy.17–19 Indeed, it is even possible that vasectomy, which in many men invokes a strong and persistent autoimmune response, may protect against testicular cancer through enhanced immune surveillance.20
Parkinson's disease and lung cancer
There is evidence suggesting that smoking is protective against Parkinson's disease.21–24 We do not have data on smoking, but we confirm that lung cancer, as a proxy for smoking, is less common in people with Parkinson's disease than in others (Table 1).
Peptic ulcer and gastric cancer
Gastric cancer is sometimes initially diagnosed (or may initially manifest itself) as a gastric ulcer. Hansson et al.25 studied the risk of gastric cancer in a large cohort of Swedish patients with gastric or duodenal ulcers, taking account of time intervals between ulcer and cancer diagnoses. They found that, with at least a 3-year interval between the two, gastric ulcer was still more common than expected in people with gastric cancer. By contrast, at this time interval duodenal ulcer was actually less common than expected in people with gastric cancer. Our results are similar to those of Hansson et al.,25 with nearly identical odds ratios (Table 1). These findings may reflect different sequelae of Helicobacter pylori infection, depending on the anatomical distribution of the infection.
Schizophrenia, lung cancer, and rheumatoid arthritis
An inconclusive literature exists which suggests that cancer and rheumatoid arthritis may be less common in schizophrenics than in the general population. Various biological mechanisms have been suggested to account for these relationships. For example, Catts and Catts26 have proposed a genetic explanation for earlier reports that schizophrenics seemed to be protected against cancer in general, and protected against lung cancer in particular, despite the fact that they tend to be heavy smokers.27–31 They suggested that a genetic predisposition to elevated levels of apoptosis (programmed cell death) due to alterations in the p53 gene could cause schizophrenia by interfering with the normal pattern of brain development and could also explain reduced rates of cancer in those with schizophrenia.26 It has been postulated that schizophrenia and rheumatoid arthritis occur together less commonly than expected by chance, perhaps through shared genetic or immunological mechanisms tending to work in opposite directions.32,33 Our data show an elevated, rather than decreased, risk of lung cancer in patients with schizophrenia (Table 1), compatible with a high prevalence of smoking among schizophrenics. This casts doubt on the earlier reports of a phenotypic dissociation between lung cancer and schizophrenia.
However, our data (Table 2) do indeed confirm a significant dissociation between schizophrenia and rheumatoid arthritis. The dissociation is specific to schizophrenia and was not found between rheumatoid arthritis and other (non-schizophrenic) psychiatric illness. We wondered if dissociation might be found between schizophrenia and other autoimmune conditions and therefore studied schizophrenia and polyarteritis nodosa. We found no dissociation.
|
Asthma and autoimmune conditions
We analysed our data to determine whether a number of conditions with an autoimmune component, including rheumatoid arthritis, hypothyroidism, thyrotoxicosis, ulcerative colitis, and Crohn's disease, occur more commonly in people with asthma than expected by chance, perhaps through shared aetiological mechanisms. Peumery34 hypothesized a positive association between asthma and rheumatoid arthritis (based on two cases) but little epidemiological work on the co-occurrence of these two diseases has been done since. Our data show a significant dissociation between the two conditions (Table 2), a finding which does not provide epidemiological support for a recent hypothesis suggesting that asthma and rheumatoid arthritis (as well as other allergic and autoimmune conditions) could be linked aetiologically through the action of anti-IgE, anti-FcepsilonRI or anti-IgG antibodies activating basophils and mast cells.35
In contrast, our data show that asthma is positively associated with both hypothyroidism and thyrotoxicosis (Table 2). Shared aetiology in some people is a possible explanation. Another possibility is that thyroid hormones may influence the inflammatory component of asthma, perhaps by enhancing IgE production.36,37 Case studies of patients whose asthma has worsened dramatically with the onset of thyrotoxicosis are consistent with this hypothesis.38,39 Our data on the association between asthma and the two inflammatory bowel diseases (Table 2) show a positive association between asthma and ulcerative colitis but provide no strong evidence of an association between asthma and Crohn's disease.
Inflammatory bowel disease and thyroid conditions
Little work has been done on the co-occurrence of inflammatory bowel disease and thyroid disease. Jarnerot et al.40 found an excess of thyrotoxicosis among ulcerative colitis patients. Snook et al.41 found an increased prevalence of thyrotoxicosis and myxoedema among people with ulcerative colitis, and a slightly reduced prevalence of both thyroid disorders among people with Crohn's disease. Our results show positive associations between ulcerative colitis and both thyroid diseases, but they are not statistically significant, and there was no hint of a positive association between Crohn's disease and the two thyroid conditions.
|
Discussion |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
We have described, with brief examples, the ‘navigation’ of a dataset of linked statistical records for studying disease associations. The ORLS is an example of how statistics collected primarily for administrative purposes can, when enhanced by linkage, be used to address issues of biological interest. The dataset does, however, have some limitations. It includes hospital day cases and in-patients only. It misses people without any such care; it misses people whose hospital care occurs outside the area of, or before the time-period covered by, the ORLS; and it does not follow those who migrate outside the area. Accordingly, the absolute values of disease occurrence (and operation rates) are underestimated, to the extent that cases are missed for these reasons. However, provided that cases and controls migrate at similar rates, the relative risks we quote are valid measures of association, or lack of it, between diseases. Another limitation is that the dataset lacks information on some potentially important confounders, such as smoking; and confounding may be one explanation for associations found within the dataset. For these reasons, studies using datasets like this, as with all studies using vital statistical data, are best used as simple tools for testing hypotheses where seeking more definitive data in clinically-based studies would be costly and time-consuming.
The study of disease associations should lend insight to clinicians interested in disease sequelae, clinical scientists interested in disease mechanisms, and epidemiologists and geneticists interested in aetiology. At the intersection between epidemiology and genetics, we suggest that disease association studies have the potential to offer important aetiological clues to complex disease in two broad areas. First, the co-occurrence of two diseases may suggest shared aetiology, pointing, in some instances, to shared genetic susceptibility. Second, it is likely that many complex diseases are made up of distinct aetiological subtypes, which result in similar (but not identical) clinical symptoms and which have different prognoses. With appropriate ethical approval, large medical datasets could be used by researchers to identify such aetiological and prognostic subtypes, for example groups of patients whose diagnosis (e.g. ulcerative colitis) is followed by different clinical courses (e.g. those who do and those who do not develop malignancy), and search for genetic differences in these respects among those groups.
Advances in data processing and in genetics are greatly increasing the feasibility of constructing epidemiological and genetic databases covering large populations.42 One topical example is the Icelandic Healthcare Database being constructed by deCODE genetics.43 In England, Fears and Poste44 highlighted the potential of the National Health Service (NHS) as a research resource for developing health-care information systems and genomics-based medicine. While promising, a prospective national NHS database will require a significant amount of development before being a useful research resource. Meanwhile, several datasets already exist which could be used to study inter-relationships between diseases. In addition to the ORLS, they include datasets established in Western Australia;45 in Scotland;46 linkable disease registers which have been established over many years in Scandinavian countries; and the General Practice Research Database (GPRD) in the UK.47 With advances in and falling costs of data processing, there is the potential, using large linked databases such as these, to map the phenomic expression of diseases which occur together more commonly, and those which occur together less commonly, than their individual frequencies would predict.
|
Acknowledgments |
|---|
This work was supported by funding from the Oxford Regional Health Authority and the South East Regional Office of the National Health Service Executive. We thank Glenys Bettley and Myfanwy Griffith for their help in building the linked files.
|
Notes |
|---|
Address correspondence to Dr Michael Goldacre, Unit of Health-Care Epidemiology, Institute of Health Sciences, University of Oxford, Oxford OX3 7LF. e-mail:michael.goldacre{at}dphpc.ox.ac.uk
|
References |
|---|
|
TopSummaryIntroductionMethodsResultsDiscussionReferences |
|---|
1. MacKay D. Hospital morbidity statistics: preliminary study of inpatient discharges. London, General Register Office, 1951.
2. Alderson MR. Statistics as a basis for health management and planning. In: Hobson W, ed. The Theory and Practice of Public Health, 4th edn. London, Oxford University Press, 1975:29–43.
3. Acheson ED. Medical Record Linkage. Oxford, Oxford University Press, 1967.
4. Goldacre MJ, Simmons H, Henderson J, Gill LE. Trends in episode based and person based rates of admission to hospital in the Oxford record linkage study area. Br Med J1988; 296:583–5.
5. Berkson J. Limitations of the application of fourfold table analysis to hospital data. Biometrics1946; 2:47–53.[Web of Science]
6. Armitage P, Berry G. Statistical Methods in Medical Research, 2nd edn. Oxford, Blackwell Scientific Publications, 1987:399–405.
7. Esteve J, Benhamou E, Raymond L. Statistical Methods in Cancer Research 4:62–66. Lyon, International Agency for Research on Cancer, 1994.
8. Ekbom A, Helmick C, Zack M, Adami HO. Ulcerative colitis and colorectal cancer. N Engl J Med1990; 323:1228–33.[Abstract]
9. Karlen P, Lofberg R, Brostrom O, Leijonmarck CE, Hellers G, Persson PG. Increased risk of cancer in ulcerative colitis: a population-based cohort study. Am J Gastroenterol1999; 94:1047–52.[Web of Science][Medline]
10. Lewis JD, Deren JJ, Lichtenstein GR. Cancer risk in patients with inflammatory bowel disease. Gastroenterol Clin North Am1999; 28:459–77.[Web of Science][Medline]
11.
Gillen CD, Walmsley RS, Prior P, Andrews HA, Allan RN. Ulcerative colitis and Crohn's disease: a comparison of the colorectal cancer risk in extensive colitis. Gut1994; 35:1590–2.
12. Stefansson T, Ekbom A, Sparen P, Pahlman L. Cancers among patients diagnosed as having diverticular disease of the colon. Eur J Surg1995; 161:755–60.[Web of Science][Medline]
13. Ekbom A, Yuen J, Adami HO, McLaughlin JK, Chow WH, Persson I, Fraumeni JF Jr. Cholecystectomy and colorectal cancer. Gastroenterology1993; 105:142–7.[Web of Science][Medline]
14. Giovannucci E, Colditz GA, Stampfer MJ. A meta-analysis of cholecystectomy and risk of colorectal cancer. Gastroenterology1993; 105:130–41.[Web of Science][Medline]
15. Reid FD, Mercer PM, Harrison M, Bates T. Cholecystectomy as a risk factor for colorectal cancer: a meta-analysis. Scand J Gastroenterol1996; 31:160–9.[Web of Science][Medline]
16.
Rosenberg L, Palmer JR, Zauber AG. Vasectomy and the risk of prostate cancer. Am J Epidemiol1990; 132:1051–5.
17.
John EM, Whittemore AS, Wu AH, Kolonel LN, Hislop TG, Howe GR, West DW, Hankin J, Dreon DM, Teh CZ, Burch JD, Paffenbarger R Jr. Vasectomy and prostate cancer: results from a multiethnic case-control study. J Natl Cancer Inst1995; 87:662–9.
18.
Møller H, Knudsen LB, Lynge E. Risk of testicular cancer after vasectomy: cohort study of over 73 000 men. Br Med J1994; 309:295–9.
19.
Rosenberg L, Palmer JR, Zauber AG, Warshauer ME, Strom BL, Harlap S, Shapiro S. The relation of vasectomy to the risk of cancer. Am J Epidemiol1994; 140:431–8.
20. Nienhuis H, Goldacre M, Seagroatt V, Gill L, Vessey M. Incidence of disease after vasectomy: a record linkage retrospective cohort study. Br Med J1992; 304:743–6.
21.
Doll R, Peto R, Wheatley K, Gray R, Sutherland I. Mortality in relation to smoking: 40 years’ observations on male British doctors. Br Med J1994; 309:901–11.
22.
Gorell JM, Rybicki BA, Johnson CC, Peterson EL. Smoking and Parkinson's disease: a dose-response relationship. Neurology1999; 52:115–19.
23.
Grandinetti A, Morens DM, Reed D, MacEachern D. Prospective study of cigarette smoking and the risk of developing idiopathic Parkinson's disease. Am J Epidemiol1994; 139:1129–38.
24.
Vanacore N, Spila-Alegiani S, Raschetti R, Meco G. Mortality cancer risk in parkinsonian patients: a population-based study. Neurology1999; 52:395–8.
25.
Hansson LE, Nyren O, Hsing AW, Bergstrom R, Josefsson S, Chow WH, Fraumeni JF Jr, Adami HO. The risk of stomach cancer in patients with gastric or duodenal ulcer disease. N Engl J Med1996; 335:242–9.
26. Catts VS, Catts SV. Apoptosis and schizophrenia: is the tumour suppressor gene, p53, a candidate susceptibility gene? Schizophr Res2000; 41:405–15.[Web of Science][Medline]
27. Rassidakis NC, Kelepouris M, Goulis K, Karaiossefidis K. On the incidence of malignancy among schizophrenic patients. Agressologie1973; 14:269–73.[Web of Science][Medline]
28.
Fox BH, Howell MA. Cancer risk among psychiatric patients: a hypothesis. Int J Epidemiol1974; 3:207–8.
29. Dupont A, Møller Jensen O, Strömgren E, Jablensky A. Incidence of cancer in patients diagnosed as schizophrenic in Denmark. In: ten Horn GHMM, Giel R, Gulbinat WH, Henderson JH, eds. Psychiatric Case Registers in Public Health. Amsterdam, Elsevier, 1986:229–39.
30. Harris AE. Physical disease and schizophrenia. Schizophr Bull1988; 14:85–96.
31. Mortensen PB. The occurrence of cancer in first admitted schizophrenic patients. Schizophr Res1994; 12:185–94.[Web of Science][Medline]
32. Allebeck P, Rodvall Y, Wistedt B. Incidence of rheumatoid arthritis among patients with schizophrenia, affective psychosis and neurosis. Acta Psychiatr Scand1985; 71:615–19.[Web of Science][Medline]
33. Eaton WW, Hayward C, Ram R. Schizophrenia and rheumatoid arthritis: a review. Schizophr Res1992; 6:181–92.[Web of Science][Medline]
34. Peumery JJ. [Asthma and rheumatoid polyarthritis. Apropos of 2 cases of association]. Rev Fr Allergol1967; 7:33–9.[Medline]
35. Marone G, Spadaro G, Palumbo C, Condorelli G. The anti-IgE/anti-FcepsilonRIalpha autoantibody network in allergic and autoimmune diseases. Clin Exp Allergy1999; 29:17–27.[Web of Science][Medline]
36. Manzolli S, Macedo-Soares MF, Vianna EO, Sannomiya P. Allergic airway inflammation in hypothyroid rats. J Allergy Clin Immunol1999; 104:595–600.[Web of Science][Medline]
37.
Sato A, Takemura Y, Yamada T, Ohtsuka H, Sakai H, Miyahara Y, Aizawa T, Terao A, Onuma S, Junen K, Kanamori A, Nakamura Y, Tejima E, Ito Y, Kamijo K. A possible role of immunoglobulin E in patients with hyperthyroid Graves’ disease. J Clin Endocrinol Metab1999; 84:3602–5.
38. Hoffman DA, McConahey WM. Thyrotoxicosis and asthma. Lancet1982; i:808.
39. Ursu HI, Peretianu D, Bulandra T, Cucu C, Grigorie D. Graves’-Basedow's disease—asthma association. Presentation of five cases. Rom J Endocrinol1993; 31:89–94.[Medline]
40. Jarnerot G, Azad Khan AK, Truelove SC. The thyroid in ulcerative colitis and Crohn's disease. II. Thyroid enlargement and hyperthyroidism in ulcerative colitis. Acta Med Scand1975; 197:83–7.[Web of Science][Medline]
41.
Snook JA, de Silva HJ, Jewell DP. The association of autoimmune disorders with inflammatory bowel disease. Q J Med1989; 72:835–40.
42. Khoury MJ. Human genome epidemiology: translating advances in human genetics into population-based data for medicine and public health. Genet Med1999; 1:71–73.[Web of Science][Medline]
43. Gulcher JR, Stefánsson K. The Icelandic Healthcare Database and informed consent. N Engl J Med 2000; 342:1827–30.
44.
Fears R, Poste G. Building population genetics resources using the U.K. NHS. Science1999; 284:267–8.
45. Brameld KJ, Holman CDJ, Bass AJ, Codde JP, Rouse IL. Hospitalisation of the elderly during the last year of life: an application of record linkage in Western Australia 1985–1994. J Epidemiol Community Health1998; 52:740–4.[Abstract]
46. Kendrick S, Clarke J. The Scottish Record Linkage System. Health Bulletin1993; 51:72–9.[Medline]
47. Walley T, Mantgani A. The UK General Practice Research Database. Lancet1997; 350:1097–9.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  N Rajoriya, C J Wotton, D G R Yeates, S P L Travis, and M J Goldacre Immune-mediated and chronic inflammatory disease in people with sarcoidosis: disease associations in a large UK database Postgrad. Med. J., May 1, 2009; 85(1003): 233 - 237. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Keene and J. Rodriguez Are mental health problems associated with use of Accident and Emergency and health-related harm? Eur J Public Health, August 1, 2007; 17(4): 387 - 393. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M.J. Goldacre, C.J. Wotton, V. Seagroatt, and D. Yeates Immune-related disease before and after vasectomy: an epidemiological database study Hum. Reprod., May 1, 2007; 22(5): 1273 - 1278. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. J. GOLDACRE, L. M. KURINA, C. J. WOTTON, D. YEATES, and V. SEAGROAT Schizophrenia and cancer: an epidemiological study The British Journal of Psychiatry, October 1, 2005; 187(4): 334 - 338. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M J Goldacre, C J Wotton, V Seagroatt, and D Yeates Cancers and immune related diseases associated with Down's syndrome: a record linkage study Arch. Dis. Child., November 1, 2004; 89(11): 1014 - 1017. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Petrou, Z. Mehta, C. Hockley, P. Cook-Mozaffari, J. Henderson, and M. Goldacre The Impact of Preterm Birth on Hospital Inpatient Admissions and Costs During the First 5 Years of Life Pediatrics, December 1, 2003; 112(6): 1290 - 1297. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L M Kurina, M J Goldacre, D Yeates, and V Seagroatt Appendicectomy, tonsillectomy, and inflammatory bowel disease: a case-control record linkage study J Epidemiol Community Health, July 1, 2002; 56(7): 551 - 554. [Abstract] [Full Text]
|
|
|
|
|
|
J Brind, V M Chinchilli, M Goldacre, L Kurina, and V Seagroatt Abortion and breast cancer J Epidemiol Community Health, March 1, 2002; 56(3): 237 - 238. [Full Text]
|
|
|
|
|
|
L M Kurina, M J Goldacre, D Yeates, and L E Gill Depression and anxiety in people with inflammatory bowel disease J Epidemiol Community Health, October 1, 2001; 55(10): 716 - 720. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M J Goldacre, L M Kurina, V Seagroatt, and D Yeates Abortion and breast cancer: a case-control record linkage study J Epidemiol Community Health, May 1, 2001; 55(5): 336 - 337. [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||