SEARCH

SEARCH BY CITATION

Keywords:

  • children;
  • epidemiology;
  • leukaemia;
  • lymphoma;
  • small-area analysis

Abstract

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

Summary. Childhood leukaemias and lymphomas have been associated with exposure to environmental factors, including infections, which show geographical variation. This study examined the geographical distribution of the incidence of acute leukaemia and lymphoma using Manchester Children's Tumour Registry (MCTR) data 1976–2000. A total of 910 children were included, all of whom had histologically and/or cytologically verified leukaemia or lymphoma. At the time of their diagnoses, all the children were aged 0–14 years and were resident in the counties of Greater Manchester or Lancashire. Standardized morbidity ratios were calculated. Poisson regression was used to examine the relationship between incidence rates and small-area (census ward) population density, ethnic composition and deprivation index. There was a monotonic relationship between acute lymphoblastic leukaemia (ALL) incidence and population density (P = 0·05). Higher rates were seen in more densely populated areas. There was evidence for a monotonic relationship between the incidence of the mixed cellularity subtype of Hodgkin's disease (HD) and the Townsend deprivation score (P = 0·001). Markedly higher incidence was associated with greater levels of unemployment and household overcrowding. The results for ALL and mixed cellularity HD support the involvement of environmental factors, such as infections, in disease aetiology.

The Manchester Children's Tumour Registry (MCTR) collects incidence data on all cancer in children, aged 0–14 years, from a defined geographical region of north-west England. Ascertainment has been estimated to be close to 100% (Birch, 1988). The registry retains diagnostic specimens and re-review is undertaken periodically in line with changing disease understanding and new technology. The MCTR therefore provides a unique data set for the investigation of incidence patterns over a wide geographical area and time frame.

We have reported previously space–time clustering among cases of acute lymphoblastic leukaemia (ALL) (Birch et al, 2000) and precursor B-cell ALL (PB_ALL) (McNally et al, 2002) included in the MCTR. We have also reported seasonal variation in onset among cases of Hodgkin's disease (HD) and PB_ALL (Westerbeek et al, 1998). These findings are consistent with a role for infections in aetiology. Several hypotheses have been proposed (Greaves, 1988; Kinlen, 1995; Smith, 1997). If infections, or other geographically varying agents, are involved in the aetiology of certain diagnostic subgroups of leukaemias or lymphomas, then case exposure to the agents may be expected to vary according to the location of residence. Thus, the distribution of the cases may exhibit some systematic geographical variation. We have explored further the role that geographically varying agents may play in the aetiology of leukaemia and lymphoma. First, we have calculated standardized morbidity ratios (SMRs). Secondly, we have analysed the data for the possible existence of spatial clustering. That is, we have performed a test for the general presence of localized excesses, which is also known as extra-Poisson variation. Thirdly, we have examined the relationship between incidence rates and population density, measures of ethnic population composition and the level of deprivation, all at small-area (census ward) level.

Patients and methods

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

Cases, aged 0–14 years, diagnosed between 1 January 1976 and 31 December 2000, registered with the MCTR and resident in the counties of Lancashire and Greater Manchester were included in this study. The diagnostic groups analysed comprise all acute leukaemias (700 cases), ALL (571 cases), PB_ALL (361 cases), T-cell ALL (55 cases), null-ALL (32 cases), acute non-lymphoblastic leukaemia (ANNL, 129 cases), HD (99 cases), nodular sclerosis HD (NS-HD, 50 cases), mixed cellularity HD (MC_HD, 26 cases), lymphocyte predominant HD (LP_HD, 18 cases), non-Hodgkin's lymphoma (NHL, 111 cases), T-cell NHL (18 cases) and B-cell NHL (20 cases). The other cases of NHL comprised one null subtype and 72 with the subtype not known, but these were not analysed separately. Immunophenotypes of ALL were available from 1980 onwards, and immunophenotypes of NHL were consistently available from 1991 onwards.

Analyses were performed at census ward level. The childhood population in the census wards ranged between 117 and 4194. During the study period, there were two national censuses. Two time periods (1976–85 and 1986–2000) were considered for some analyses because: (i) small-area census population data were only available from the 1981 and 1991 censuses [Office of Population Censuses and Surveys Census Division, General Register Office (Scotland) Census Branch, 1983; Office for National Statistics, 1991a]; and (ii) they provided sufficiently long time periods to allow adequate case numbers for analyses, but not so long for population shifts to dilute any area-specific effects. This was confirmed by sensitivity analysis. Also, there were some boundary changes at ward level between the 1981 and 1991 censuses.

The reference address for all cases was the address at diagnosis. There were 519 wards in the 1981 census and 518 wards in the 1991 census. Cases were allocated to wards [Office of Population Censuses and Surveys Census Division, General Register Office (Scotland) Census Branch, 1981; Office for National Statistics, 1991b].

MCTR data were used to generate standard rates for each diagnostic group. The iterative procedure was used, originally developed by Mantel & Stark (1968), but simplified by Breslow & Day (1975). Observed and expected numbers of cases were calculated by age–sex–ward strata for each period. These were then summed to obtain standardized morbidity ratios (SMRs) for each ward.

Ward characteristics were derived from the small-area statistics of the censuses [Office of Population Censuses and Surveys Census Division, General Register Office (Scotland) Census Branch, 1983; Office for National Statistics, 1991a]. These included population density, ethnic composition and level of deprivation. The Townsend score for deprivation at ward level (and not individual level) was calculated (Townsend et al, 1988). This is a combination of four census measures: unemployment, households without access to a car, home tenancy and household overcrowding.

The Potthoff–Whittinghill test (Potthoff & Whittinghill, 1966a,b; Muirhead & Ball, 1989) was used to test for departures from the Poisson assumption at ward level, which would be indicative of the general presence of localized excesses (spatial clustering).

Area-based analyses were carried out. Statistical analyses (ecological regressions) were performed using Poisson regression in glim (generalized linear interactive modelling) software (Francis et al, 1993). The number of cases observed in each ward was the dependent variable, and the log of the expected number of cases was used as the offset. The ecological (independent) variables were the stratified census-derived ward characteristics.

Continuous variables were cut into strata so that each stratum contained an (approximately) equal proportion of the childhood population. The following ward variables were cut into quintiles: population density, child population density, percentage white population, percentage Pakistani population, percentage Indian population (these were the three largest ethnic groups), the Townsend deprivation score and its components.

A series of univariate analyses was carried out. For each quintile, the observed (O) and expected (E) numbers of cases were obtained, and the ratio O/E was calculated. A relative risk (RR) was calculated, for each quintile, by comparing the ratio O/E with the value of O/E for the first quintile (i.e. the 20% of the childhood population who resided in a ward that had the lowest value of the explanatory variable). The first quintile was assigned a RR of 1. Confidence intervals (CIs) were obtained using glim (Francis et al, 1993). A test for linear trend was performed. Statistical significance was taken as P < 0·05 throughout the analyses.

Results

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

The observed numbers of cases by age group, time period, diagnostic group and gender are given in Table I. The person–years at risk for the analyses are also presented.

Table I.  Observed cases and person–years at risk, by diagnostic group (all acute leukaemias (leuks), ALL, PB_ALL, T-ALL, ANLL, HD, NS_HD, MC_HD, NHL, T-NHL).
 All acute leuksALLPB_ ALLT-ALLANLL Person–years at risk
1976–85
 0–4133117504162 402 260
 5–97961244182 667 820
 10–147051185193 235 580
 Male1581334910254 263 330
 Female12496433284 042 330
 Total2822299213538 305 660
1986–2000
 0–423819416610444 087 665
 5–996806412163 874 350
 10–1484683920163 587 340
 Male23119815227335 911 845
 Female18714411715435 637 510
 Total418342269427611 549 355
 HDNS_HDMC_HDNHLT-NHLPerson–years at risk
  1. Patient age is stratified into age groups of 0–4 years, 5–9 years and 10–14 years and also males and females.

1976–85
 0–42021902 402 260
 5–94041902 667 820
 10–14311271813 235 580
 Male267113804 263 330
 Female11521814 042 330
 Total3712135618 305 660
1986–2000
 0–45411524 087 665
 5–9231091883 874 350
 10–14342432273 587 340
 Male40201036135 911 845
 Female221831945 637 510
 Total623813551711 549 355

Acute leukaemias

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

There was no evidence for the general presence of localized excesses (spatial clustering) for 1976–85, P = 0·78, and for 1986–2000, P = 0·37. Ecological analyses showed that there was a significant monotonic relationship between ALL incidence and population density (Table II, P = 0·05). Higher rates were seen in the more densely populated wards. Further analysis showed that this was largely due to cases of PB_ALL and T-cell ALL and not null-ALL. However, as numbers were small, statistical significance was not achieved (Table II). There was also a marginally significant monotonic relationship between ANLL incidence and population density (Table II, P = 0·06). However, in contrast to ALL, higher rates were seen in the less densely populated wards.

Table II.  Acute leukaemias analyses of population density *   (data available for wards: 1976–2000) [quintile 1 = lowest; quintile 5 = highest], showing RRs and 95% CIs.
QuintileALLPB_ALLT-cell ALLNull-ALLANLL
  • *

    Quintile boundaries: 1 [5–1245]; 2 [1246–2158]; 3 [2158–3176]; 4 [3178–4356]; 5 [4367–10 420] persons per square km.

1RR = 1RR = 1RR = 1RR = 1RR = 1
2RR = 1·11 (0·84–1·46)RR = 1·10 (0·78–1·57)RR = 5·33 (1·55–18·30)RR = 0·66 (0·19–2·33)RR = 1·17 (0·70–1·93)
3RR = 1·20 (0·92–1·57)RR = 1·36 (0·98–1·90)RR = 3·91 (1·10–13·85)RR = 1·28 (0·44–3·69)RR = 0·94 (0·55–1·59)
4RR = 1·21 (0·92–1·58)RR = 1·17 (0·83–1·65)RR = 3·63 (1·01–13·01)RR = 1·29 (0·45–3·71)RR = 0·77 (0·44–1·34)
5RR = 1·30 (0·99–1·69)RR = 1·26 (0·89–1·76)RR = 4·44 (1·26–15·57)RR = 0·97 (0·31–3·02)RR = 0·67 (0·37–1·19)
Test for linear trendP  = 0·05P  = 0·20P  = 0·14P  = 0·66P  = 0·06

Hodgkin's disease (HD)

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

The MC_HD exhibited evidence for the general presence of localized excesses (spatial clustering) during 1986–2000 (P = 0·04). Although based on only 26 cases, ecological analyses showed striking significant monotonic relationships (Table III) between the MC subtype and the Townsend deprivation score (P = 0·001), unemployment (P = 0·003), households without access to a car (P = 0·02), home tenancy (P = 0·05) and household overcrowding (P = 0·005). Markedly higher incidence of the MC subtype was associated with greater levels of deprivation, unemployment and household overcrowding. There was a significant monotonic relationship between the LP subtype and population density (P = 0·05). A lower incidence was associated with greater levels of population density.

Table III.  Mixed cellularity subtype of HD analyses (data available for wards: 1976–2000) [quintile 1 = lowest, quintile 5 = highest], showing significant results, RRs and 95% CIs.
Quintile Townsend score* UnemploymentHouseholds without access to a car Home tenancy§ Household overcrowding
  • *

    Quintile boundaries: 1976–85 1 [−5·7 to −2·4]; 2 [−2·3 to −0·7]; 3 [−0·7–1·5]; 4 [1·5–4·0]; 5 [4·1–13·1]; 1986–2000 1 [−5·0 to −2·3]; 2 [−2·3 to −0·7]; 3 [−0·7–1·3]; 4 [1·3–4·3]; 5 [4·3–14·7].

  • Quintile boundaries: 1976–85 1 [1·2–7·1]; 2 [7·1–9·5]; 3 [9·5–12·1]; 4 [12·1–16·3]; 5 [16·6–33·5] per cent; 1986–2000 1 [1·8–6·0]; 2 [6·0–8·2]; 3 [8·2–10·8]; 4 [10·8–16·7]; 5 [16·7–39·2] per cent.

  • Quintile boundaries: 1976–85 1 [9·0–30·7]; 2 [30·7–40·8]; 3 [41·0–49·3]; 4 [49·4–58·8]; 5 [58·9–86·0] per cent; 1986–2000 1 [4·6–25·0]; 2 [25·1–34·1]; 3 [34·2–42·2]; 4 [42·3–51·9]; 5 [51·9–81·4] per cent.

  • §

    Quintile boundaries: 1976–85 1 [2·8–20·3]; 2 [20·3–30·3]; 3 [30·3–39·8]; 4 [39·8–55·0]; 5 [55·2–99·5] per cent; 1986–2000 1 [1·7–16·1]; 2 [ 16·1–24·6]; 3 [24·8–33·7]; 4 [33·8–48·4]; 5 [48·5–97·9] per cent.

  • Quintile boundaries: 1976–85 1 [0·2–1·9]; 2 [1·9–2·8]; 3 [2·8–3·8]; 4 [3·8–5·1]; 5 [5·1–21·8] per cent; 1986–2000 1 [0–1·0]; 2 [1·0–1·5]; 3 [1·5–2·1]; 4 [2·1–2·9]; 5 [3·0–21·0] per cent.

1RR = 1RR = 1RR = 1RR = 1RR = 1
2RR = 5·02 (0·59–43·00)RR = 4·03 (0·45–36·02)RR = 2·50 (0·48–12·86)RR = 1·32 (0·30–5·92)RR = 4·95 (0·58–42·35)
3RR = 3·02 (0·31–29·04)RR = 4·07 (0·46–36·43)RR = 2·53 (0·49–13·02)RR = 1·00 (0·20–4·97)RR = 4·03 (0·45–36·07)
4RR = 4·09 (0·46–36·58)RR = 7·13 (0·88–57·96)RR = 1·52 (0·25–9·11)RR = 2·70 (0·72–10·19)RR = 5·05 (0·59–43·19)
5RR = 13·08 (1·71–100·02)RR = 10·14 (1·30–79·24)RR = 5·57 (1·23–25·11)RR = 2·66 (0·70–10·01)RR = 11·04 (1·43–85·50)
Test for linear trendP  = 0·001P  = 0·003P  = 0·02P  = 0·05P  = 0·005

Non-Hodgkin's lymphoma (NHL)

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

There was no evidence for the general presence of localized excesses (spatial clustering) for 1976–85, P = 0·67, and for 1986–2000, P = 0·56. Ecological analyses showed no statistically significant relationships for NHL overall. Additional analyses were performed by immunophenotype of NHL and also for T-cell ALL together with T-cell NHL, as these are closely related diseases in children, and a marginally significant result was obtained for T-cell ALL. Therefore it was of interest to see whether combining T-cell ALL with T-cell NHL strengthened the result for T-cell ALL. However, the results by immunophenotype of NHL were not statistically significant, and the results for T-cell ALL together with T-cell NHL added little to those for T-cell ALL alone.

Discussion

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

This is the first analysis to examine the geographical distribution of ALL by immunophenotype. The study has only been made possible by the availability of high-quality and consistent population-based diagnostic and residential address data. As ascertainment is close to 100%, there is no reason to suspect that there is any artifactual bias by small-area of diagnosis.

Several methodological issues should be emphasized. The ethnic composition of the ward is not necessarily related to characteristics of individual cases and should only be regarded as an ecological measurement. Likewise, population density and the deprivation scores are ward based and are not individual characteristics. Area-level data have been assigned to individual cases. Care should be exercised when using such grouped data to make inferences about individuals. There may be various confounding factors that exhibit the same pattern of geographical variability (Richardson & Montfort, 2000).

We have found previously space–time clustering among cases of ALL aged 0–4 years and specifically among cases of PB_ALL, aged 18–54 months (Birch et al, 2000; McNally et al, 2002). We have also identified seasonal variation in the onset of PB_ALL (Westerbeek et al, 1998). Furthermore, we have found that recent temporal increases in the incidence of ALL are specifically caused by cases of PB_ALL in the childhood peak (McNally et al, 2000, 2001). All these findings, together with current hypotheses concerning an infectious origin for childhood leukaemia (Greaves, 1988; Kinlen, 1995; Smith, 1997), would lead us to predict that there should be an increased incidence of PB_ALL in areas of lower population density and greater affluence, as most epidemiological evidence points to a lack of immune stimulation particularly in the early years of life. Unexpectedly, we identified a non-significant increased incidence of PB_ALL in areas of greater population density.

However, as hypothesized by Greaves (1988) and Kinlen (1995), delayed immune stimulation is probably somewhat more related to the types of population mixing and exposure to infections and not the amount of opportunity for exposure to specific infectious agents. There have been a number of other studies that have reported systematic geographical differences in the incidence of ALL but with apparently contradictory findings (Greenberg & Shuster, 1985). Although some studies have reported an increased incidence in less densely populated areas and areas of greater affluence (Alexander et al, 1990, 1996), other studies have reported a more complex pattern, with increased incidence in both intermediately populated areas and less densely populated areas (Alexander et al, 1998, 1999). We are not aware of any other study that has examined the geographical distribution of ALL by immunophenotype. However, the current analysis has not aided the interpretation of the previous contradictory studies. Thus, for PB_ALL, we suggest that there may be two aetiological mechanisms operating, giving rise to two subsets of cases. One subset of cases of PB_ALL may result from in utero exposure to infections, which would be more likely to be associated with areas of greater deprivation and higher population density. A second subset of cases of PB_ALL may result from delayed exposure to common infections, which would be more likely to be associated with areas of greater affluence and lower population density. The findings of the previous studies would be dependent on which of the two subsets predominated during the particular time period and geographical area.

An infectious aetiology involving the Epstein–Barr virus (EBV) has been suggested for HD. EBV has been detected in numerous studies of HD. The EBV has been particularly associated with childhood cases and the MC subtype (Glaser et al, 1997; Preciado et al, 1997), and MC_HD may arise by direct viral transformation. One prediction from this mechanism is the greater incidence of MC_HD in small areas that have a greater opportunity for such a mechanism to take effect, that is where there is greater person-to-person contact. A second prediction is that there will be localized excesses of cases of MC_HD. The present findings of a strong association with increased levels of deprivation (particularly unemployment and household overcrowding) and of localized excesses of cases, which were only found for MC_HD, are supportive of the role of a directly transforming virus. Although the results are based on small numbers (only 26 cases), they are most striking.

In summary, we have found geographical heterogeneity in certain leukaemias and lymphomas. There is evidence that MC_HD had greater incidence in more deprived wards, particularly those that had greater unemployment and household overcrowding, and also exhibited localized excesses, which is very supportive of a direct role for infections. There was weaker evidence for the higher incidence of PB_ALL in areas of greater population density, but these latter results should be treated with caution. Finally, the results are based on ecological analyses, and care must be used when interpreting the results at individual level.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References

The Manchester Children's Tumour Registry (MCTR) is supported by Cancer Research UK. Jillian M. Birch is Cancer Research UK Professorial Research Fellow in Paediatric Oncology, and Osborn B. Eden is Cancer Research UK Professor of Paediatric Oncology at the University of Manchester. We thank Mrs E. A. Dale, Mrs D. A. Elliott, Mrs J. F. Hogg and Mr C. Nikolaisen for all their hard work on data processing and verification. The work is based on census data, which are copyright of The Crown. The work is based on data provided with the support of the ESRC and JISC and uses Census boundary material, which is copyright of The Crown and the ED-LINE Consortium.

References

  1. Top of page
  2. Abstract
  3. Patients and methods
  4. Results
  5. Acute leukaemias
  6. Hodgkin's disease (HD)
  7. Non-Hodgkin's lymphoma (NHL)
  8. Discussion
  9. Acknowledgments
  10. References
  • Alexander, F.E., Ricketts, T.J., McKinney, P.A. & Cartwright, R.A. (1990) Community lifestyle characteristics and risk of acute lymphoblastic leukaemia in children. Lancet, 336, 14611465.
  • Alexander, F.E., Leon, D.A. & Cartwright, R.A. (1996) Isolation, car ownership, and small area variation in incidence of acute lymphoblastic leukaemia in children. Paediatric and Perinatal Epidemiology, 10, 411417.
  • Alexander, F.E., Boyle, P., Carli, P.M., Coebergh, J.W., Draper, G.J., Ekbom, A., Levi, F., McKinney, P.A., McWhirter, W., Michaelis, J., Peris-Bonet, R., Petridou, E., Pompe-Kirn, V., Plisko, I., Pukkala, E., Rahu, M., Storm, H., Terracini, B., Vatten, L. & Wray, N. on behalf of the EUROCLUS project (1998) Spatial clustering of childhood leukaemia: summary results from the EUROCLUS project. British Journal of Cancer, 77, 818824.
  • Alexander, F.E., Boyle, P., Carli, P.M., Coebergh, J.W., Ekbom, A., Levi, F., McKinney, P.A., McWhirter, W., Michaelis, J., Peris-Bonet, R., Petridou, E., Pompe-Kirn, V., Plisko, I., Pukkala, E., Rahu, M., Stiller, C.A., Storm, H., Terracini, B., Vatten, L. & Wray, N. (1999) Population density and childhood leukaemia: results of the EUROCLUS Study. European Journal of Cancer, 35, 439444.
  • Birch, J.M. (1988) Manchester Children's Tumour Registry 1954–70 and 1971–83. In: International Incidence of Childhood Cancer (ed. by D.M.Parkin, C.A.Stiller, G.J.Draper, C.A.Bieber, B.Terracini & J.L. Young), pp. 299304. IARC Scientific Publications no. 87. IARC, Lyon.
  • Birch, J.M., Alexander, F.E., Blair, V., Eden, O.B., Taylor, G.M. & McNally, R.J.Q. (2000) Space-time clustering patterns in childhood leukaemia support a role for infection. British Journal of Cancer, 82, 15711576.
  • Breslow, N.E. & Day, N.E. (1975) Indirect standardisation and the multiplicative model for rates with reference to the age adjustment of cancer incidence and relative frequency data. Journal of Chronic Diseases, 28, 289303.
  • Francis, B., Green, M. & Payne, C. (eds) (1993) The GLIM System. Release 4 Manual. Clarendon Press, Oxford.
  • Glaser, S.L., Lin, R.J., Stewart, S.L., Ambinder, R.F., Jarrett, R.F., Brousset, P., Pallesen, G., Gulley, M.L., Khan, G., O'Grady, J., Hummel, M., Preciado, M.V., Knecht, H., Chan, J.K. & Claviez, A. (1997) Epstein–Barr virus-associated Hodgkin's disease: epidemiologic characteristics in international data. International Journal of Cancer, 70, 375382.
  • Greaves, M.F. (1988) Speculations on the cause of childhood acute lymphoblastic leukaemia. Leukemia, 2, 120125.
  • Greenberg, R.S. & Shuster, Jr, J.L. (1985) Epidemiology of cancer in children. Epidemiologic Reviews, 7, 2248.
  • Kinlen, L.J. (1995) Epidemiological evidence for an infective basis in childhood leukaemia. British Journal of Cancer, 71, 15.
  • McNally, R.J.Q., Birch, J.M., Taylor, G.M. & Eden, O.B. (2000) Temporal increase in the incidence of childhood peak precursor B-cell acute lymphoblastic leukaemia seen in north west England. Lancet, 356, 485486.
  • McNally, R.J.Q., Cairns, D.P., Eden, O.B., Kelsey, A.M., Taylor, G.M. & Birch, J.M. (2001) Examination of temporal trends in the incidence of childhood leukaemias and lymphomas provides aetiological clues. Leukemia, 15, 16121618.
  • McNally, R.J.Q., Alexander, F.E. & Birch, J.M. (2002) Space–time clustering analyses of childhood acute lymphoblastic leukaemia by immunophenotype. British Journal of Cancer, 87, 513515.
  • Mantel, N. & Stark, C.R. (1968) Computation of indirect-adjusted rates in the presence of confounding. Biometrics, 24, 9971005.
  • Muirhead, C.R. & Ball, A.M. (1989) Contribution to the discussion at the Royal Statistical Society meeting on cancer near nuclear establishments. Journal of the Royal Statistical Society, Series A, 152, 376.
  • Office for National Statistics (1991a) 1991 Census: Small Area Statistics and Local Base Statistics [computer file]. ESRC/JISC Census Programme. Census Dissemination Unit, University of Manchester.
  • Office for National Statistics (1991b) 1991 Census: Digitised Boundary Data (Great Britain). ESRC/JISC Census Programme. Census Geography Data Unit (UKBORDERS), University of Edinburgh.
  • Office of Population Censuses and Surveys Census Division, General Register Office (Scotland) Census Branch (1981) 1981 Census: Digitised Boundary Data (Great Britain). ESRC/JISC Census Programme. Census Geography Data Unit (UKBORDERS), University of Edinburgh.
  • Office of Population Censuses and Surveys Census Division, General Register Office (Scotland) Census Branch (1983) 1981 Census Small Area Statistics: 100% Population and Households Aggregated to Ward Level (Great Britain) [computer file]. UK Data Archive [distributor], SN: 1893, Colchester, Essex.
  • Potthoff, R.F. & Whittinghill, M. (1966a) Testing for homogeneity. I. The binomial and multinomial distributions. Biometrika, 53, 167182.
  • Potthoff, R.F. & Whittinghill, M. (1966b) Testing for homogeneity. II. The Poisson distribution. Biometrika, 53, 183190.
  • Preciado, M.V., Diez, B. & Grinstein, S. (1997) Eptein Barr virus in Argentine pediatric Hodgkin's disease. Leukemia and Lymphoma, 24, 283290.
  • Richardson, C. & Montfort, C. (2000) Ecological correlation studies. In: Spatial Epidemiology: Methods and Applications (ed. by P.Elliott, J.Wakefield, N.Best & D. Briggs). Oxford University Press, Oxford.
  • Smith, M. (1997) Considerations on a possible viral etiology for B-precursor acute lymphoblastic leukemia of childhood. Journal of Immunotherapy, 20, 89100.
  • Townsend, P., Phillimore, P. & Beattie, A. (1988) Health and Deprivation. Inequality and the North. Croom-Helm, London.
  • Westerbeek, R.M.C., Blair, V., Eden, O.B., Kelsey, A.M., Stevens, R.F., Will, A.M., Taylor, G.M. & Birch, J.M. (1998) Seasonal variations in the onset of childhood leukaemia and lymphoma. British Journal of Cancer (Suppl.), 78, 119124.