Abstract

Background Creutzfeldt-Jakob disease (CJD) is a rare fatal dementia caused by a transmissible agent. However, the mechanism leading to the disease is unknown in the majority of cases. The presence of geographically clustered cases might indicate a common environmental exposure to the transmissible agent, or case-to-case transmission of the agent. This study sought evidence of clustering of cases of sporadic CJD in France.

Methods A total of 402 individuals who died from definite or probable sporadic CJD in France between 1992 and 1998 were analysed. The geographical distribution of cases was analysed using three different clustering methods. An analysis of the distribution of the distances between pairs was performed to look for evidence of clustering. Then, two methods of cluster detection were used to identify the locations of clusters.

Results Each of our analyses found some evidence of clustering, though the extent of that clustering differed between approaches. The strongest evidence, statistically, related to three cases living in a small rural area in South-West France (P = 0.001). Two of the three cases lived in the same area throughout life. They had also both undergone surgery on several occasions. Little information is available on the third case.

Conclusion Some sporadic CJD cases in France may be aetiologically linked. There was strong evidence that three cases in South-West France formed a cluster but the precise mechanism underlying this cluster of cases remains unclear. The potentially long incubation period of the disease makes the identification of links between such cases difficult.

Creutzfeldt-Jakob disease (CJD) is the most commonly occurring human transmissible spongiform encephalopathy. The annual incidence of the disease in Europe is approximately 1 per million population.1 Some 10–15% of CJD cases are associated with mutations in the prion protein (PRNP) gene (inherited cases), while another few per cent are iatrogenic, the result of human-to-human transmission during medical procedures.2 In France, most iatrogenic Creutzfeldt-Jakob disease cases are attributable to human growth hormone treatment.3,4 In Japan, dura mater grafts constitute the most frequent reported source of iatrogenic Creutzfeldt-Jakob disease.5 In 1996, a new variant of CJD (vCJD) was identified in the UK.6 These cases have been attributed to infection with the agent responsible for bovine spongiform encephalopathy (BSE). Until now, only 4 such cases have been identified outside the UK (3 in France and 1 in Ireland). Thus the great majority of CJD cases, 80–85% of the total, are of unknown origin and are commonly referred to as ‘sporadic’ cases.

One possible origin of sporadic CJD is a spontaneous conformational change in an individual's prion protein (PrP), from PrPsen into the disease-related isoform PrPres.7 Epidemiological data indicate that an individual's risk of the disease is influenced by their PRNP genotype at codon 129.8 Individuals who are homozygous at codon 129 (methionine-methionine or valine-valine) are at higher risk of developing sporadic CJD than heterozygous individuals (methionine-valine). Alternatively, the conformational change might be induced by exposure to PrPres from an unknown external source.9

Case-control studies of environmental risk factors for sporadic CJD have investigated place of residence, occupation, exposure to animals or animal products, dietary habits and medical history as potential risk factors.10–12 No consistent findings have emerged from these studies, and age and codon 129 genotype remain the only well-established risk factors.

Small, unexplained ‘clusters’ of sporadic CJD have been reported in Italy, Japan, and Canada,13–15 but rigorous statistical analyses of these clusters have not generally been performed. Geographical clustering of sporadic CJD cases, if it occurs, might reflect case-to-case transmission of the agent, common exposure to an environmental source or, perhaps, a shared genetic susceptibility lying outside the PrP coding region.

Following the epidemic of BSE in the UK, national surveillance of CJD was implemented in France in 1991. Since 1992 it has provided data on the incidence of CJD throughout the country.16,17 In the present study, a variety of statistical methods were used to look for objective evidence of clustering in 402 individuals who died from sporadic CJD between 1992 and 1998.

Data

Case ascertainment is based on direct notification of suspected cases of CJD by neurologists, neuropathologists and laboratories involved in the examination of the prion protein gene or, since 1998, the detection of 14–3–3 protein in cerebrospinal fluid. Suspected cases were examined, when possible, by the study neurologist (JPB) and approximately 60% of patients with suspected CJD underwent post-mortem. The diagnostic criteria for CJD used over the study period have been published.18 In this paper, analyses are based on definite and probable cases.

Cases were classified as iatrogenic when the patient had a known history of exposure to pituitary growth hormone or dura mater. They were classified as inherited if they carried a disease-associated mutation of the prion protein gene or if, in the absence of genetic analysis, they had a family history of definite or probable CJD. One case was classified as vCJD on the basis of their neuropathological features.19 All other cases were classified as sporadic.

A total of 503 patients with definite or probable CJD died between January 1992 and December 1998, yielding an annual mortality rate of 1.2 per million inhabitants. Among these 503 CJD cases, 60 were classified as iatrogenic (most of them being related to human growth hormone), 40 were classified as inherited and 1 case was a variant CJD. Among the remaining 402 sporadic CJD cases, the median age at death was 67.5 years and ranged from 38 to 88 years (sex ratio: 170 males/ 232 females). In all, 296 (74%) underwent analysis of the PRNP gene, of whom 121 (40.9%) were methionine homozygotes at codon 129, 50 (16.9%) were valine homozygotes and 34 (11.5%) were methionine-valine heterozygotes.

Place of residence at the time of disease onset was available for all sporadic cases and place of birth was known for all but two cases. Eight cases were excluded from our analyses because they lived outside France at the time of disease onset. A total of 56 patients were born outside France, mostly in North Africa. Thus, the geographical distribution of places of birth could be analysed for 344 cases and distribution of place of residence at onset could be analysed for 394 of the 402 cases. For 126 (31.3%) of the 402 cases, places of residence throughout life were known because these individuals were included in a separate case-control study.

Central grid co-ordinates of 36 609 districts across France were obtained from the French Institut National de Géographie. Data on the populations of these districts were obtained from the 1995 census of the French population. The smallest occupied district had less than 100 inhabitants, the largest (Paris) 2 175 200. Large districts such as Paris, Lyon (415 487 inhabitants) or Marseilles (800 550 inhabitants) were subdivided into Arrondissements with populations ranging from several thousands to about 200 000. Thus, in total, 36 654 areas with data on location and population size were identified. The median population size of these areas was about 300 and 95% of the areas had a population of less than 5000.

Statistical Methods

Analysis 1

Initially statistical evidence of clustering of cases of sporadic CJD in France was sought through an analysis of the distribution of the distances between all possible pairs of cases.20 We first calculated the observed number of pairs of cases within a particular distance of each other. This observed number of pairs was then compared to distribution of this number under the null hypothesis of no clustering. This null distribution was obtained by Monte-Carlo simulation, by repeated random sampling of the population with the probability of including a person from a specific area being proportional to the population of the area (1000 simulations performed). This method provides an overall test for presence of clustering but does not aim to give precise locations of possible clusters. Analyses were performed separately for place of residence at onset and place of birth. Analyses stratified by codon 129 genotype were also performed.

Analysis 2

Further analyses were performed to identify possible clusters, initially using the method of Besag-Newell. This method was originally developed for the study of the geographical distribution of childhood leukaemia. According to the authors, the method is suitable for rare diseases of unknown aetiology.21 The basis of this method is relatively simple: circles are drawn centred on each case that encompasses its k nearest neighbouring cases. Then, for each circle, the Poisson probability of observing k or more cases within this circle is calculated given the population at risk within the circle. Ideally, the age structure of the population at the district level should be used to adjust for age in calculating the expected number of cases, but these data were not available to us. The analysis was repeated for different values of k. A cluster was classified as ‘consistent’ if its statistical significance at the 5% level persisted over three values of k. A case was then defined as clustered if it was included in the largest circle of one of the consistent clusters (which could result in clusters of size possibly larger than k). The value of k specified, together with the total number of cases, largely determines the radii of the circles that are explored by the method. Large values of k focus on large ‘clusters’ and thus tend to investigate circles (and clusters) of large radius. For the purposes of this analysis, k was set successively to 4, 6 and 9 in order to consider cluster radii up to about 70 km.

Analysis 3

A criticism of the Besag-Newell method is that the control of multiple testing through the definition of consistent clusters is arbitrary and informal. The method is therefore said to provide many false positive clusters. To overcome this problem of multiple testing, we applied another method of cluster location according to Kulldorff et al.22,23 This method uses a spatial scan statistic which can detect clusters of any size located anywhere in the study region, whether or not they conform with administrative borders. The maximum cluster size to be investigated was set to 50% of the total population at risk. Essentially the method examines circles of continuously varying size centred on a large number of different locations. For each circle, a likelihood ratio is computed for the alternative hypothesis that there is an increased risk of disease inside the circle against the null hypothesis that the risk inside the circle is the same as the risk outside it. The ‘most likely cluster’ is that with the largest likelihood ratio. The statistical significance of this largest likelihood ratio is assessed by determining its distribution under the null hypothesis through Monte Carlo simulation (1000 simulations performed). This takes account of the multiple testing inherent in the procedure. Each analysis by this method identifies a single ‘most likely cluster’. To look for further clusters, we re-applied the same procedure to datasets from which the cases and the population data for the most likely cluster had been removed.

For Analyses 1 and 2, computer programs were written in C. For Analysis 3, we used the StsWin (SaTScan) software developed and provided by the method developer (URL: http://dcp.nci.nih.gov/bb/satscan.html).

Results

Figure 1 shows the standardized mortality ratios of sporadic CJD by Department. Table 1 summarizes the results of the first analysis of clustering and shows the observed and expected numbers of pairs of cases born, or living at onset, within distances of 1, 5, 10, 20 km of each other. For example, among the 394 cases of sporadic CJD, 117 pairs of cases were found to have places of residence at onset less than 1 km apart compared with a median of 218 pairs observed under the null hypothesis. Thus, there is no evidence of an excess of pairs of cases living within 1 km of each other. The same was true for pairs of cases living within 5 km of each other. However, the table does indicate that for sporadic CJD there is some evidence of clustering of places of onset at distances of less than 10 and 20 km.

Analyses using place of birth instead of place of residence at onset did not detect any evidence of clustering. An analysis of the geographical distribution of sporadic cases by codon 129 genotype did not reveal any evidence of clustering (data not shown).

The Besag-Newell method identified five possible clusters of sporadic CJD by place of residence at onset, persisting over the three values of k used in the analyses. These clusters were widely distributed across France (North, West, Centre and South-West). To protect patient confidentiality, the precise location of the clusters is not given. One cluster included 7 cases, three included 10 cases each, and one included 25 cases. In three of the five identified clusters, more than 95% of individuals forming these clusters lived in urban areas. This proportion was about 70% in the two other clusters.

The method of Kulldorff et al. identified only one cluster of three cases located in the South-West (P = 0.001, radius 4.5 km). These three cases formed part of one of the larger clusters identified by the Besag-Newell method. The next most likely cluster (three cases, North of France) had a P-value of 0.18 (radius 2.5 km). None of the other clusters identified by the Besag-Newell method approached statistical significance using the spatial scan statistic (P > 0.20 for all three).

We examined the records of the cases involved in the two most likely clusters detected by Kulldorff's method. Some characteristics of the cases are summarized in Table 2. In the South-West cluster, three cases aged 60–84 years at onset lived at the time of disease onset in two small rural parishes less than 5 km apart. Two had lived in the same parish throughout life. The third case was born about 100 km from his place of residence at the time of disease onset, but his residential history throughout life was not recorded. Two cases were methionine homozygotes, the third case was not genotyped. Detailed information on medical history and other possible exposures were available for only two cases. One case with onset in 1996 had a history of five surgical procedures between 1959 and 1996 including vertebral column surgery in 1991. The other case, with onset in 1997, had ocular surgery in 1994 and had had other surgical treatments in 1977 and 1993. The two patients had been operated on in the same hospitals, but not within a one-year period of each other. They had both undergone dental treatment by the same dentist who practised in the area. One case was reported to eat brain of veal or lamb at least once per month, while the other reportedly ate brain occasionally. Both ate veal thymus several times per year and venison occasionally and both had lived on farms.

In region 2 (North), three cases, aged 57–72 years at onset, were living less than 3 km apart in an urban area at disease onset. One case, whose lifetime residential history was known, had lived in the same parish throughout life, while the other two cases were born less than 10 km from the place where they lived at disease onset. One had a past history of vertebral column surgery. None were genotyped. In the patients with known past surgical history, the possibility of iatrogenic transmission of the disease through dura mater implants was investigated by examining records and by interviewing a referral expert in neurosurgery. According to this expert, none of the cases reported here were likely to have had dura mater implants.

Discussion

A major limitation of the present study is that data on lifetime places of residence were not available for all cases. Studies of iatrogenic CJD and kuru indicate that the incubation period of these diseases can be very long.3,24 Therefore, evidence of residential proximity 10, 20 or more years before disease onset is likely to be more meaningful than geographical clustering at disease onset. We observed no evidence of clustering of places of birth of sporadic CJD cases. A weakness of this analysis is that the reference population for this analysis is based on the 1995 census data, when we should be using the population distribution at around the time of birth of the cases (60+ years before on average).

Another limitation of the present study is that the age-structure of the reference population used for calculating expected number of cases was not available. In France, the greatest variations in the age distribution are likely to be found between rural and urban areas, the mean age in rural areas being greater than that in urban areas. Because the incidence of sporadic CJD is much higher in older individuals, it is possible that ‘false-positive’ clustering resulting from confounding by age might be observed in rural areas. Such an effect could account for some of the clustering observed in Analysis 1. With respect to Analysis 3, the ‘North’ cluster is centred in an urban area and is unlikely to be explained by such confounding. The ‘South-West’ cluster is located in a rural area. Even if we were to assume that all of the population in that area was aged greater than 55 years, the expected number of cases would increase only to 0.08, compared with three observed cases.

Apparent clustering of sporadic CJD cases might also be observed if diagnostic practices vary across France. In 1998, the 14–3–3 test was added to the diagnostic criteria for probable CJD. However, this was very close to the end of the period considered in this paper and the availability of this test became very rapidly well-known by neurologists following the emergence of vCJD in Great Britain. In the two clusters considered, three cases had a positive 14–3–3 result, all of whom underwent a post-mortem. We think it unlikely that regional differences in the utilization of the 14–3–3 test explain the observed clusters.

On the other hand, if clustering of sporadic CJD does occur through common environmental exposure to the disease agent, analyses based solely on place of residence at onset are likely to underestimate the true strength of clustering.25 Interestingly, at least two of the individuals in the ‘most likely cluster’ had lived in the same small area throughout life and thus may have been exposed to common environmental factors many years before disease onset. Alternatively, one of the cases may have transmitted the infectious agent to the others, although the short time delay between the onsets of the cases in the absence of any neurosurgery linking the cases argues against this explanation. Given that a full PrP gene analysis was not performed on these cases, we cannot exclude the possibility that they are familial cases perhaps with unidentified co-paternity.

Given the limitations outlined above, it is perhaps not surprising that, overall, the evidence for geographical clustering of sporadic CJD in France is rather weak, with some inconsistencies apparent between the results obtained by the different methods. The cluster identified by the method of Kulldorff et al. involves only a small proportion of all cases (<1%) and this cluster cannot explain the excess of pairs of cases living close together identified in our initial analysis of clustering (Table 1). While the method of Besag-Newell identifies a higher proportion of cases as occurring in clusters (17%), which could account for the excesses of pairs observed in Table 1, this method is known to be susceptible to false positive results.

A case-control study performed in Australia reported an increased risk of sporadic CJD associated with surgery, though the methodological limitations of the study prevent any firm conclusions being drawn with regard to this association.11 In the present study, we found that some of the cases belonging to two possible clusters had undergone several episodes of surgery. One possible explanation for the presence of these clusters is, therefore, that some of these cases were iatrogenically infected, although we could not identify any clear surgical links between any of the cases. We should note that cluster analyses based on place of residence at onset might miss clusters of unrecognized iatrogenic CJD if the surgery took place relatively far away from the place of residence of individuals at onset.

In summary, we used three methods to investigate the geographical distribution of sporadic CJD cases in France. Each of these methods found some evidence of clustering, though the extent of that clustering differed between approaches. The strongest evidence, statistically, related to three cases living in a small rural area in South-West France. Two of the three cases lived in the same area throughout life and could have shared an environmental exposure to the infectious agent. They had also both undergone surgery on several occasions. Little information is available on the third case. The precise mechanism underlying this cluster of cases remains, therefore, unclear.

Table 1

Observed numbers of pairs of Creutzfeldt-Jakob disease cases born or living at onset within 1, 5, 10 or 20 km of each other

  No. of observed pairs at a distance of (median/95th/99th percentile of number of pairs under the null hypothesis) 
Sporadic Creutzfeldt-Jakob disease N <1 km <5 km <10 km <20 km 
aIndicates that the value exceeds the 95th percentile of the expected distribution (thresholds values are given below in parenthesis). 
At onset 394 117 370 931a 1878a 
  (218/332/381) (358/488/550) (651/904/1025) (1249/1678/1931) 
Place of birth 344 114 303 628 1027 
  (124/219/255) (226/332/391) (447/637/741) (916/1277/1448) 
  No. of observed pairs at a distance of (median/95th/99th percentile of number of pairs under the null hypothesis) 
Sporadic Creutzfeldt-Jakob disease N <1 km <5 km <10 km <20 km 
aIndicates that the value exceeds the 95th percentile of the expected distribution (thresholds values are given below in parenthesis). 
At onset 394 117 370 931a 1878a 
  (218/332/381) (358/488/550) (651/904/1025) (1249/1678/1931) 
Place of birth 344 114 303 628 1027 
  (124/219/255) (226/332/391) (447/637/741) (916/1277/1448) 
Table 2

Characteristics of six cases of sporadic Creutzfeldt-Jakob in two clusters identified in France

Cluster Case # Year of birth, sex, codon 129 status Year of death Diagnostic status Diagnostic criteria Past surgical history Notes on dura mater implants 
South-West 1 1914,F, Met-Met 1998 Definite 14–3–3+ Post-mortem Eye surgery, 1994 Abdominal surgery, 1978 Dental surgery with evulsion, 1978 No dura mater graft documented 
 2 1938,M, Met-Met 1998 Definite 14–3–3+ Post-mortem Endoscopy (stomach), 1997 Discale hernia, 1991 Inguinal hernia, 1976 Appendectomy, 1975 Dental surgery with evulsion, date unkown No dura mater graft documented 
 3 1933,F, Met-Met 1995 Probable EEG+ Not documented Unknown 
North 1 1936,M, unknown 1993 Probable EEG+ Coronary surgery, date unkown No dura mater graft documented 
 2 1931,M, Met-Met 1997 Definite 14–3–3+ EEG+ Post-mortem Vertebral surgery, date unknown No dura mater graft documented 
 3 1923,F, unknown 1995 Probable EEG+ Not documented Unknown 
Cluster Case # Year of birth, sex, codon 129 status Year of death Diagnostic status Diagnostic criteria Past surgical history Notes on dura mater implants 
South-West 1 1914,F, Met-Met 1998 Definite 14–3–3+ Post-mortem Eye surgery, 1994 Abdominal surgery, 1978 Dental surgery with evulsion, 1978 No dura mater graft documented 
 2 1938,M, Met-Met 1998 Definite 14–3–3+ Post-mortem Endoscopy (stomach), 1997 Discale hernia, 1991 Inguinal hernia, 1976 Appendectomy, 1975 Dental surgery with evulsion, date unkown No dura mater graft documented 
 3 1933,F, Met-Met 1995 Probable EEG+ Not documented Unknown 
North 1 1936,M, unknown 1993 Probable EEG+ Coronary surgery, date unkown No dura mater graft documented 
 2 1931,M, Met-Met 1997 Definite 14–3–3+ EEG+ Post-mortem Vertebral surgery, date unknown No dura mater graft documented 
 3 1923,F, unknown 1995 Probable EEG+ Not documented Unknown 
Figure 1

Standardised mortality ratio of sporadic Creutzfeldt-Jakob diseases by French Department

Figure 1

Standardised mortality ratio of sporadic Creutzfeldt-Jakob diseases by French Department

J Huillard d'Aignaux was supported by a fellowship of the Institut National de la Santé et de la Recherche Médicale (INSERM) and by the UK Department of Health. The authors thank the neurologists, the neuropathologists and all the French physicians who participated in the prospective surveillance, and the families of the cases for providing information going back many years. The authors thank the Professor Peter Smith for helpful discussion and for a critical review of the manuscript, and the anonymous reviewers for their helpful comments.

References

1
Will RG, Alperovitch A, Poser S et al. Descriptive epidemiology of Creutzfeldt-Jakob disease in six European countries, 1993–1995. EU Collaborative Study Group for CJD.
Ann Neurol
 
1998
;
43
:
763
–67.
2
de Silva R, Esmonde T. Iatrogenic transmission of Creutzfeldt-Jakob disease.
CNS Drug
 
1992
;
2
:
96
–101.
3
Huillard d'Aignaux J, Costagliola D, Maccario J et al. Incubation period of Creutzfeldt-Jakob disease in human growth hormone recipients in France.
Neurology
 
1999
;
53
:
1197
–201.
4
Brandel JP, Agid Y. Creutzfeldt-Jakob disease after treatment with growth hormone: the point in July, 1998.
Rev Neurol
  (Paris)
1998
;
154
:
654
–55.
5
Nakamura Y, Aso E, Yanagawa H. Relative risk of Creutzfeldt-Jakob disease with cadaveric dura transplantation in Japan.
Neurology
 
1999
;
53
:
218
–20.
6
Will R, Ironside J, Zeilder M et al. A new variant of Creutzfeldt-Jakob disease in the UK.
Lancet
 
1996
;
347
:
921
–25.
7
Prusiner S, DeArmond S. Prion disease and neurodegeneration.
Annu Rev Neurosci
 
1994
;
17
:
311
–39.
8
Alperovitch A, Zerr I, Pocchiari M et al. Codon 129 prion protein genotype and sporadic Creutzfeldt-Jakob disease.
Lancet
 
1999
;
353
:
1673
–74.
9
Manuelidis L, Sklaviadis T, Akowitz A, Fritch W. Viral particles are required for infection in neurodegenerative Creutzfeldt-Jakob disease.
Proc Natl Acad Sci USA
 
1995
;
92
:
5124
–28.
10
Zerr I, Brandel JP, Masullo C et al. European surveillance on Creutzfeldt-Jakob disease: a case-control study for medical risk factors.
J Clin Epidemiol
 
2000
;
53
:
747
–54.
11
Collins S, Law MG, Fletcher A, Boyd A, Kaldor J, Masters CL. Surgical treatment and risk of sporadic Creutzfeldt-Jakob disease: a case-control study.
Lancet
 
1999
;
353
:
693
–97.
12
van Duijn CM, Delasnerie-Laupretre N, Masullo C et al. Case-control study of risk factors of Creutzfeldt-Jakob disease in Europe during 1993–95. European Union (EU) Collaborative Study Group of Creutzfeldt-Jakob disease (CJD).
Lancet
 
1998
;
351
:
1081
–85.
13
D'Alessandro M, Petraroli R, Ladogana A, Pocchiari M. High incidence of Creutzfeldt-Jakob disease in rural Calabria, Italy.
Lancet
 
1998
;
352
:
1989
–90.
14
Arakawa K, Nagara H, Itoyama Y et al. Clustering of three cases of Creutzfeldt-Jakob disease near Fukuoka City, Japan.
Acta Neurol Scand
 
1991
;
84
:
445
–47.
15
Nosal R, Kapoor A, Shanin R. Cluster of cases of Creutzfeldt-Jakob disease—Ontario.
Can Dis Wkly Rep
 
1991
;
17
:
12
.
16
d'Aignaux JH, Laplanche JL, Delasnerie-Laupretre N et al. Trends in mortality from sporadic Creutzfeldt-Jakob disease in France 1992–7.
J Neurol Neurosurg Psychiatry
 
2000
;
68
:
787
–89.
17
Ruffie A, Delasnerie-Laupretre N, Brandel JP et al. [Incidence of Creutzfeldt-Jakob disease in France, 1992–1995].
Rev Epidemiol Sante Publique
 
1997
;
45
:
448
–53.
18
Brandel JP, Delasnerie-Laupretre N, Laplanche JL, Hauw JJ, Alperovitch A. Diagnosis of Creutzfeldt-Jakob disease: effect of clinical criteria on incidence estimates.
Neurology
 
2000
;
54
:
1095
–99.
19
Chazot G, Broussolle E, Lapras L, Bläter T, Aguzzi A, Kopp N. New variant of Creutzfeldt-Jakob disaese in a 26-year old French man.
Lancet
 
1996
;
347
:
1181
.
20
Mantel N, Bailar JC, 3rd. A class of permutational and multinomial tests arising in epidemiological research.
Biometrics
 
1970
;
26
:
687
–700.
21
Besag JE, Newell JN. The detection of clusters in rare diseases.
J R Statist Soc A
 
1991
;
265
:
143
–55.
22
Kulldorff M. A spatial scan statistic.
Communication in Statistics: Theory and Methods.
 
1997
;
26
:
1481
–96.
23
Kulldorff M, Feuer EJ, Miller BA, Freedman LS. Breast cancer clusters in the northeast United States: a geographic analysis.
Am J Epidemiol
 
1997
;
146
:
161
–70.
24
Prusiner S, Gajdusek D, Alpers M. Kuru with incubation periods exceeding two decades.
Ann Neurol
 
1982
;
12
:
1
–9.
25
Cousens SN, Linsell L, Smith PG et al. Geographical distribution of variant CJD in the UK (excluding Northern Ireland).
Lancet
 
1999
;
353
:
18
–21.