[Emerging Infectious Diseases]
[Volume 4 No. 2 / April - June 1998]

Perspectives

Could Myocarditis, Insulin-Dependent Diabetes Mellitus, and Guillain-Barré
Syndrome Be Caused by One or More Infectious Agents Carried by Rodents?

Bo Niklasson,*,† Birger Hörnfeldt,‡ Berit Lundman§,¶
*Swedish Institute for Infectious Disease Control, Stockholm,
Sweden;†National Defense Research Establishment, Umeå, Sweden; ‡Department
of Animal Ecology, Umeå University, Umeå, Sweden; §Department of Advanced
Nursing, University of Umeå, Umeå, Sweden; ¶Research and Development Unit,
Sundsvalls Hospital, Sundsvall, Sweden

  --------------------------------------------------
      The numbers of small rodents in northern Sweden fluctuate
      heavily, peaking every 3 or 4 years. We found that the
      incidence of Guillain-Barré syndrome and insulin-dependent
      diabetes mellitus, as well as the number of deaths caused by
      myocarditis, followed the fluctuations in numbers of bank
      voles, although with different time lags. An environmental
      factor, such as an infectious agent, has been suggested for all
      three diseases. We hypothesize that Guillain-Barré syndrome,
      myocarditis, and insulin-dependent diabetes mellitus in humans
      in Sweden are caused by one or more infectious agents carried
      by small rodents. Also, a group of novel picornaviruses
      recently isolated from these small rodents is being
      investigated as the possible etiologic agent(s).

Nephropathia epidemica (NE) is a disease caused by Puumala virus (genus
Hantavirus, family Bunyaviridae). The vector and natural reservoir of
Puumala virus is a small rodent, the bank vole (Clethrionomys glareolus)
(1). In most parts of southern Sweden, bank vole populations are noncyclic,
whereas in the north, populations fluctuate on a 3- or 4-year cycle of
abundance (2-4). NE is endemic only in the northern part of the country,
and the number of human cases cycles with the bank vole population (5-6).

We found the incidence of myocarditis, Guillain-Barré syndrome (GBS), and
insulin-dependent-diabetes mellitus (IDDM) to lag behind the population
density fluctuations of bank voles, although with different time delays. We
hypothesize that in Sweden the bank vole is a vector for one or more
infectious agents, which are pathogenic and cause these diseases in humans.

The Study

Density of Bank Voles

Long-term records (>10 years) on small rodent abundance are scarce.
However, in Sweden such data are available from the area with cyclic rodent
populations (2, 4-8). We used data from Grimsö (59°40' N, 15°25' E), where
small rodents have been collected every autumn since 1973 (6-8).
Snap-trapping was performed where possible in six adjacent 5x5-km areas,
each with four 1-ha plots. Of 24 permanent 1-ha plots, 20 were trapped. At
each trapping, approximately 940 traps were set during three consecutive
days, amounting to approximately 2,800 "trap-nights."

Species, date, and location of trapped animals were recorded (4, 6-8). As
an index of bank vole abundance, we calculated the number of animals
trapped per 100 trap nights (i.e., density). Bank vole abundance fluctuated
in cycles with 3- or 4-year intervals between density peaks (Figure 1).

[fig 1]
Figure 1. Bank vole abundance in Grimsö, 1973-150;1994. Untransformed data.


Incidence of Disease

We used the national census to calculate (according to counties and
age groups) annual disease incidences per 100,000 population. Because of
the lack of long-term records on noncyclic small rodent populations,
we analyzed data from the counties with cyclic rodent populations: 
Norrbotten, Västerbotten, Västernorrland, Jämtland, Gävleborg, Kopparberg, 
Värmland, Örebro, Västmanland, Uppsala, Stockholm, Södermanland, Göteborg, 
-lvsborg, Jönköping, and Kronoberg (2-3). We excluded Stockholm and Uppsala
counties as they are predominantly urban and on the border between counties 
with cyclic and noncyclic rodents.

Myocarditis

Myocarditis is a clinical condition in which the heart is infiltrated by
inflammatory cells. Diagnosis is very difficult during the acute stage and
is often made by postmortem microscopic investigation of the myocardium.
All causes of death in Sweden are registered by the Swedish Cause-of-Death
Statistics, Statistics Sweden, S-115 81 Stockholm, Sweden. Our study
included all deaths between 1970 and 1986, in the age group 11 to 46 years,
caused by myocarditis (ICD 8 code 422), if the diagnosis was given as
either the direct (primary) cause of death or as the first out of six
recorded contributing causes of death. The age group 11 to 46 years was
chosen because congenital or arteriosclerotic cardiovascular disease is
less common in this age group. In 201 (92%) of 218 cases, the diagnoses
were based on clinical or forensic autopsies. The remaining diagnoses were
based on clinical examination before death. We did not use data collected
after 1986, as the disease classification changed from ICD 8 to ICD 9 in
1987, making comparison with the period after 1986 difficult.

Guillain-Barré Syndrome

In different parts of Europe (9), reported incidence of GBS varies
considerably depending on study method. Our data, from the hospital
discharge registry, Swedish National Board of Health and Welfare, S-106 30
Stockholm, Sweden, are considered adequate for epidemiologic surveillance
of GBS (9). We selected patients hospitalized with the diagnosis GBS (ICD 8
code 354) and gathered information on age, sex, county of residence, county
of hospitalization, diagnosis, date of admission and date of discharge. We
only analyzed data of GBS patients 46 years of age and younger, as the
diagnoses in older patients frequently are obscured by nonspecific
illnesses and other problems (10).

Insulin-Dependent Diabetes Mellitus

IDDM was studied in Medelpad (part of Väternorrland County), an area with
cyclic rodent populations. The patients were identified through registries
of patients with diabetes at the only hospital and at 16 out of the 17
health-care centers, as well as through common registries of diagnosis at
one health-care center. The World Health Organization (WHO) diagnostic
criteria were applied. We analyzed case data in all age groups because
precision in the diagnosis of IDDM is not age related.

Statistical Analysis

A cross-correlation function (CCF) (11) can indicate any direct or delayed
dependence between two different time series. A CCF graph is a plot of
positive and negative correlation coefficients between pair-wise values
from the two series. The correlations have been calculated between the two
series for different time shifts (lags), with lag 0 meaning no time shift,
lag -1 meaning the first series has been shifted backwards one time unit,
and lag 1 meaning the first series has been shifted forward by one time
unit.

We used CCFs, calculated in Statistical Package for the Social Sciences
(SPSS) (12), to indicate any temporal association between the incidence in
cyclic versus noncyclic areas of GBS and of deaths from myocarditis,
respectively. We also used CCFs to indicate any direct (at lag 0) or
delayed dependence (at negative lags) of the different disease incidences
on vole density. At negative lags, the vole time series is the leading
indicator that is shifted "backwards" along the time scale, relative to the
disease series. Thus, the correlation coefficients at negative lags are in
focus, as they indicate any disease incidence dependence on past vole
densities, i.e., on densities 1, 2, or 3 years ago. At positive lags, the
disease time series is shifted "backwards" relative to the vole time
series. The correlation coefficients at positive lags may be large because
of the similarities between the two series, but we see no causal
relationship between vole density and past disease incidence.

Log transformation of the time series was used because of the variable
amplitudes of the fluctuations in the disease and vole series. In the case
of IDDM, we also ran one CCF with the disease and vole time series
difference-transformed by one year.

Findings

Myocarditis

A total of 218 patients, ages 11 to 46 years, died of acute myocarditis in
1970 to 1986; 148 in counties with cyclic rodent density and 70 in counties
with noncyclic rodent density. Sex ratio was 2:1 (146 male, 72 female
patients). Myocarditis death incidences in cyclic and noncyclic areas
(Figure 2) were not correlated as revealed by cross-correlation of the
disease series (CCF not shown; n = 17 computable 0-order correlations).
Cross-correlation of the annual incidence of myocarditis deaths in 1970 to
1986 in the cyclic area with bank vole abundance (Figure 3) showed that the
incidence of myocarditis was most highly correlated with vole abundance in
the previous year, according to the high positive and significant
correlation at lag -1 in the CCF (Figures 4, 5) (r = 0.635, p < 0.05, n =
13).

 [fig 2]                              [fig3]
 Figure 2. Incidence of death from    Figure 3. Myocarditis deaths,
 myocarditis, 1970–1986.              1974–1986 relative to bank vole
 Untransformed data.                  abundance 1 year previously (vole
                                      data from 1973-1985). Untransformed
                                      data.

 [fig 4]                              [fig 5]
 Figure 4. Cross-correlation          Figure 5. Myocarditis deaths,
 function of incidence of death from  1974–1986, relative to bank vole
 myocarditis with bank vole           abundance 1 year previously (vole
 abundance, 1973–1986. Time series    data from 1973-1985). Log
 are log transformed; n = 14          transformed data. r = 0.635, n = 13.
 computable 0-order correlations.
 Lines represent + 2 SE. The
 standard error is based on the
 assumption that the series are not
 cross-correlated and one of the
 series is white noise.

Guillain-Barré Syndrome

In five cyclic counties and two noncyclic counties, where complete
data were available, 258 GBS patients (</= 46 years of age) were hospitalized
in 1973 to 1982; the sex ratio was 1.1:1 (135 male, 123 female
patients), 104 from the cyclic and 154 from the noncyclic area.

No correlation was found between GBS in the cyclic and noncyclic areas by
cross-correlation of the disease series (CCF not shown; n = 10 computable
0-order correlations). Cross-correlation of the annual incidence of GBS in
cyclic areas in 1973 to 1982 with bank vole abundance (Figure 6) suggested
that the incidence of GBS co-varied with current vole abundance, as shown
by the high positive and significant correlation at lag 0 in the CCF
(Figures 7, 8) (r = 0.757, p < 0.05, n = 10).

 [fig 6]
Figure 6. Time series of Guillain-Barré‚ syndrome incidence, 1973&#150;1982, 
relative to bank vole abundance in the same years. Untransformed data.


 [fig 7]                              [fig 8]
 Figure 7. Cross-correlation          Figure 8. Guillain-Barré syndrome
 function of Guillain-Barré syndrome  incidence, 1973–1982, relative to
 incidence with bank vole abundance,  bank vole abundance in the same
 1973–1982. Time series are log       years. Log transformed data. r =
 transformed; n = 10 computable       0.757, n = 10.
 0-order correlations. Lines
 represent + 2 SE. The stan-dard
 error is based on the assumption
 that the series are not
 cross-correlated and one of the
 series is white noise.

Insulin-Dependent Diabetes Mellitus

A total of 318 cases of IDDM were recorded in 1971 to 1991,
representing a sex ratio of 1.3:1 179 male, 139 female patients). No
cross-correlation was found between the annual incidence of IDDM in the
cyclic area and bank vole abundance(Figure 9)(CCF not shown; n = 19
computable 0-order correlations), as was found for myocarditis and GBS
with vole abundance. On the other hand, after the alternative
transformation (to log transformation) by differencing the IDDM and vole
time series by 1 year, changes of the IDDM incidence from one year to the
next covaried with the changes of vole abundance 2 years previously. This
was suggested by the high positive and significant correlation at lag -2 in
the CCF (Figures 10,11) (r = 0.595, p < 0.05, n = 16).

 [fig 9]
 Figure 9. Time series of insulin-dependent diabetes mellitus incidence in 
 1975-1991 relative to bank vole abundance 2 years previously (vole data 
 from 1973-1989). Untransformed data.

 [fig 10]                             [fig 11]
 Figure 10. Cross-correlation         Figure 11. Change of
 function of insulin-dependent        insulin-dependent diabetes mellitus
 diabetes mellitus incidence with     incidence, 1975–1991, relative to
 bank vole abundance, 1973–1991.      change of bank vole abundance 2
 Time series are differenced (1); n   years previously, after
 = 18 computable 0-order              transformation of time series by
 correlations. Lines represent +/- 2  differencing (1) (vole data from
 SE. The standard error is based on   1973-1989). r = 0.595, n = 16.
 the assumption that the series are
 not cross-correlated and one of the
 series is white noise.

Calculation of Confidence Limits

The confidence limits (+/-2 SE) for the most important cross correlations, at
lag 0 for GBS (Figure 7), lag -1 for myocarditis (Figure 4), and at lag -2
for IDDM (Figure 10), were calculated in an alternative way to the default
calculation in SPSS (12) according to formula 11.1.11, p. 377 (11). The
alternative calculations were made according to formula 11.1.7, p. 376
(11). These calculations were made assuming that 1) the three disease time
series represented white noise (i.e. contain no autocorrelation); 2) the
vole time series contained autocorrelation, but it was zero after lag 5; 3)
for the CCF of GBS with the vole time series, correlation existed at lag 0
and +1 but was zero at other lags; 4) for the CCF of myocarditis with the
vole series, correlation existed at lag -1 but was zero at other lags; and
5) for the CCF of the differenced IDDM with the differenced vole time
series, correlation existed at lag 0 to lag -4 but was zero at other lags
and the differenced vole time series contained autocorrelation up to, but
not after, lag 7. The calculations of the alternative confidence limits
showed that the most important cross-correlations were significant at p <
0.05. The confidence limits for GBS at lag 0 were 0.757 +/- 0.542; for
myocarditis at lag -1, 0.635 +/- 0.450; and for the differenced IDDM at lag
-2, 0.595 +/- 0.478.

Conclusions

Zoonotic disease reservoir or vector population density data can indicate a
link between the etiologic agent and its reservoir/vector and between the
agent and the disease. This type of data linked rodents to NE in Sweden in
the 1930s and to Korean hemorrhagic fever in the 1950s (13-14). In a
previous study, we found that the Puumala virus infection rate in bank
voles in the spring tracked the vole density in the previous autumn and
that infection rate peaked as vole population declined (5). We believe that
this explains the approximately 9-month time lag in the incidence of NE in
the spring relative to vole abundance in the previous autumn, although the
incubation period in humans is only 2-6 weeks.

Our present data show significant temporal correlations of bank vole
population density with incidence of death from myocarditis (Figures 3-5)
and incidence of GBS (Figures 6-8). Our data also show a positive and
significant correlation between changes in IDDM incidence and changes in
bank vole density (Figures 9-11). We hypothesize that these three diseases
are caused or triggered by one or more infectious agents carried by bank
voles in Sweden. In this context, several sudden deaths among Swedish
orienteers were recognized during 1989 to 1992. Six of approximately 200
elite orienteers died of acute myocarditis (15). Orienteering involves
finding the fastest/shortest way between several checkpoints, often in
forested areas. Because of extensive exposure to nature, it has been
speculated that the sudden deaths of the orienteers were caused by a
vector-borne (rodent or arthropod) infectious agent.

If our hypothesis is correct, the different time lags of different diseases
relative to vole abundance, as seen in the present study, may depend in
part on the specific infection rate dynamics of any agent in the bank vole
population as seen in the case of NE (5). The time elapsed from infection
to disease onset or death may also be important in IDDM and myocarditis.

The disease time series presented here were not very long, and the results
should therefore be treated with caution. The present findings do not prove
any direct link between the diseases investigated and the bank vole. Many
infectious diseases and other biological phenomena have a cyclic variation,
and the relationships presented here may be spurious or may involve another
reservoir or vector population (e.g., small rodents and small game species
that fluctuate synchronously with the bank vole at these latitudes)
(4-5,7-8,16-17).

The identification of the etiologic agent or agents is critical to "prove"
our hypothesis. However, as seen in several zoonotic diseases, humans are
often dead-end hosts, and it may be easier to isolate the etiologic agent
from the reservoir or vector than from the patient. Also, because of the
time elapsed between primary infection and onset of disease, the infectious
agent may be present only in very small amounts or absent. It is also
possible that several human diseases with no etiologic agent identified are
caused by agents that are difficult to cultivate, at least by standard
techniques.

Thus, because we need to identify the etiologic agent(s) and it is
difficult to isolate any virus from humans, we have initially attempted to
isolate new viruses from small rodents, the primary suspects. The bank vole
is the most common wild-living Swedish mammal. Its abundance varying by
>300 times (4), this rodent is likely to have a high and highly variable
potential over time to transmit any etiologic agent to humans. In addition,
bank voles are known to enter buildings, thus transporting disease risk
closer to humans.

Three novel virus isolates from bank voles, resembling picornaviruses in
size and morphology, have been found. The first isolate was named Ljungan
after the Ljungan river in Medelpad County, Sweden, where the animals were
trapped. The second and third isolate originated from animals trapped in
Västerbotten County. The amino acid sequences of predicted Ljungan virus
capsid proteins were closely related (approximately 70% similarity) to the
human pathogen echovirus 22. Partial 5' noncoding region sequence of
Ljungan virus was most closely related to cardioviruses. Two additional
isolates were serologically and molecularly related to the prototype
(Niklasson et al., unpub. observation). Echovirus 22 is a known human
pathogen, and other known cardioviruses can induce not only myocarditis but
also neurologic diseases and IDDM in several species of animals.

We hope to elucidate the role of Ljungan virus as a human pathogen by
serologic assays for Ljungan virus, diagnostic polymerase chain reaction
based on generated sequence data, and specific antisera for viral antigen
detection. Studies are also under way to identify and determine the role of
other viruses of these small rodents in inducing myocarditis, GBS, and IDDM
in humans.

Acknowledgments

We thank Sara Sjöstedt for statistical help and Stiftelsen Olle Engkvist,
Byggmästare and by the Swedish Environment Protection Agency (via the
National Environmental Monitoring Programme) for financial support for the
vole trapping.

Address for correspondence: Bo Niklasson, Swedish Institute for Infectious
Disease Control, S-105 21 Stockholm, Sweden; fax: 46-8-735-66-15; e-mail
BO.NIKLASSON@SMI.KI.SE

References

  1. Niklasson B, Le Duc J. Epidemiology of nephropathia epidemica in
     Sweden. J Infect Dis 1987;155:269-76.
  2. Hansson L, Henttonen H. Gradients in density variations of small
     rodents: the importance of latitude and snow cover. Oecologia 1985;
     67:394-402.
  3. Hansson L, Henttonen H. Rodent dynamics as community processes. Trends
     in Ecology & Evolution 1988;3:195-200.
  4. Hörnfeldt B. Delayed density dependence as a determinant of vole
     cycles. Ecology 1994;75:791-806.
  5. Niklasson B, Hörnfeldt B, Lundkvist Å, Björsten S, LeDuc J. Temporal
     dynamics of Puumala virus antibody prevalence in voles and of
     nephropathia epidemica incidence in humans. Am J Trop Med Hyg
     1995;53:134-40.
  6. Hörnfeldt B. Smsdäggdjursinventering i PMK:s referensområden - rapport
     från verksamheten 1992. Solna, Sweden: Swedish Environmental
     Protection Agency; 1994. Report 4294.
  7. Lindström ER, Andrén H, Angelstam P, Cederlund G, Hörnfeldt B,
     Jäderberg L, Lemnell P-A, Martinsson B, Sköld K, Swenson JE. Disease
     reveals the predator: sarcoptic mange, red fox predation and prey
     populations. Ecology 1994;75:1042-9.
  8. Lindström ER, Hörnfeldt B. Vole cycles, snow depth and fox predation.
     Oikos 1994;70:156-60.
  9. Jiang GX, de Pedro-Cuesta J, Fredriksson G. Guillain-Barré syndrome in
     South-West Stockholm, 1973-1991. Quality of registered hospital
     diagnosis and incidence. Acta Neurol Scand 1995;91:109-17.
 10. Winner SJ, Evans JG. Age-specific incidence of Guillain-BarrT syndrome
     in Oxfordshire. Quarterly Journal of Medicine, New Series
     1990;77:1297-1304.
 11. Box GEP, Jenkins GM. Time series analysis, forecasting and control.
     San Francisco: Holden Day; 1970.
 12. SPSS Base System Syntax Reference Guide, Release 6.0. Carey (NC): SPSS
     Inc.; 1993.
 13. Myhrman G. En njursjukdom med egenartad symptombild. (In Swedish).
     Nord Med Tidskr 1934;7:793-94.
 14. Gajdusek DC. Virus haemorrhagic fevers-special reference to
     haemorrhagic fevers with renal syndrome (epidemic haemorrhagic fever).
     J Pediatr 1962;60:841-57.
 15. WesslTn L, Pshlson C, Friman G, Fohlman J, Lindquist O, Johansson C.
     Myocarditis caused by Chlamydia pneumonie (TWAR) and sudden unexpected
     death in a Swedish elite orienteer. Lancet 1992;340:427-8.
 16. Hörnfeldt B. Synchronous population fluctuations in voles, small game,
     owls and tularemia in northern Sweden. Oecologia 1978; 32:141-152.
 17. Hörnfeldt B, Löfgren O, Carlsson B-G. Cycles in voles and small game
     in relation to variations in plant production indices in northern
     Sweden. Oecologia 1986;68:496-502.


Emerging Infectious Diseases
National Center for Infectious Diseases
Centers for Disease Control and Prevention
Atlanta, GA

URL: http://www.cdc.gov/ncidod/EID/vol4no2/niklason.htm

Please note that figures and equations are not available in ASCII format; 
their placement within the text is noted by [fig] and [eq], respectively. 
Greek symbols are spelled out. The following codes are used: 
(ft) for footnote; (sup) for superscript; (sub) for subscript; 
/= for greater than or equal to.