Clin Microbiol Infect 2011; 17: 888–893
Since 2000, the epidemiology of C. difficile infections (CDI) has changed in the US and Europe. Few population-based assessments of both incidence and case fatality of CDI have been performed. In this study, the Finnish nationwide laboratory-based surveillance data from the year 2008 were analysed to assess the incidence and case fatality of CDI, and to detect regional differences in relation to molecular epidemiology. A total of 6201 episodes of CDI were identified (118.3/100 000 population; range by regions, 57.2–189.1). The incidence increased by age and was highest in persons aged >84 years (1286.0). Of the CDI episodes, 711 (11.5%; range by regions, 2.2–15.0%) led to death within 30 days. The 30-day case fatality was highest (22.0%) in persons aged >84 years. In total, 334 (5% of all episodes) isolates from 13/21 regions were sent for genotyping: 120 (36%) were of PCR ribotype 027, and it was found in 6/13 regions. Among the rest of the isolates, 53 (16%) were of type 001, and 19 (6%) of 002 and 014. The incidence and case fatality were highest in elderly persons and varied regionally. This may be explained by uneven spread of hypervirulent PCR ribotypes, such as 027, but also differences in diagnostic activity or the patient populations among which the outbreaks are occurring.
Clostridium difficile is the most frequent cause of antibiotic-associated diarrhoea. Recently, the epidemiology of C. difficile infections (CDI) has changed. Since 2000, CDI-related morbidity and mortality have increased in North America as well as in many parts of Europe [1–3]. Simultaneously, the emergence of the hypervirulent PCR ribotype 027 was observed and the virulence properties of the pathogen were characterized. However, other PCR ribotypes may also have similar genotypic virulence properties .
In Finland, hospitalizations with CDI doubled from 1996 to 2004 . The increase was most notable for patients over 64 years of age. During 1998–2004, there was also a nearly two-fold increase in CDI-associated mortality rate in the same age group. The increase in these numbers might to some extent be explained by the changes detected in diagnostic methodology . Furthermore, in October 2007, the first case of CDI caused by the hypervirulent PCR ribotype 027 was reported in Finland . In 2008 none of the Finnish clinical microbiology laboratories performed routine cell culture cytotoxicity assays . All of them used methods detecting both toxins A (TcdA) and B (TcdB), and 87% used both culture and toxin detection. In addition, three laboratories were setting up a method to detect pathogenicity locus genes by PCR for primary diagnostics.
In January 2008, toxin-positive CDI became a laboratory-notifiable disease, and all clinical microbiology laboratories were requested to send C. difficile isolates from severe cases and/or outbreaks to the National Institute for Health and Welfare (THL) for genotyping. In Finland, the use of unique personal identity codes enables the linkage of individual information within National Infectious Disease Register (NIDR) and other national databases, and to study the outcome of illness systematically within a certain time interval. The objective of this study was to assess the incidence and case fatality of CDI in Finland, and to detect regional differences in CDI epidemiology based on the surveillance and genotyping data retrieved in 2008.
Materials and Methods
In Finland (population 5.3 million), the national healthcare system is organized into 21 geographically and administratively defined healthcare districts (HDs), with populations ranging from 67 800 to 1.7 million. Fifteen HDs have only secondary and primary care hospitals, and five provide also tertiary care services.
Laboratory-based epidemiological surveillance
Since 1 January 2008 all Finnish microbiology laboratories have reported all C. difficile findings (positive culture and/or toxin production) from stools to the National Infectious Disease Register. Each notification included the following information: specimen date, each individual’s unique national identity code, date of birth, sex and place of residence. Using this information and a 3-month time interval, multiple notifications of the same person were merged as a single episode. An episode of CDI was defined as detection of toxin-positive C. difficile from stools of persons aged ≥1 years. The dates of deaths were obtained from the National Population Information System by using the national identity code.
The data in this study were analysed within the epidemiological research purposes authorized by the Finnish Contagious Diseases Act (clause 4 in section 40 a, Finlex, Valtion säädöstietopankki. Available at: http://www.finlex.fi/fi/laki/alkup/2003/20030935. Accessed 16 April 2010).
Reference laboratory activities/molecular typing
In 2008, clinical microbiology laboratories were also asked to send C. difficile isolates from severe cases (CDI-related intensive care, colectomy or death) [1,8,9] and outbreaks (≥3 C. difficile cases within 4 weeks in the same unit) to the reference laboratory for genotyping. All the isolates obtained were PCR ribotyped and classified according to the reason for sending (severe case/outbreak/other reasons or unknown).
PCR ribotyping was performed according to the protocol of the Anaerobe Reference Laboratory (ARL), Cardiff, UK , using the Cardiff-ECDC culture collection as a set of reference strains. After gel electrophoresis, the band patterns were analysed using the BioNumerics 5.0 software (Applied Maths NV, Sint-Martens-Latem, Belgium).
When a local outbreak was suspected, the isolates were typed further using pulsed-field gel electrophoresis (PFGE) typing. The PFGE patterns were analysed by BioNumerics using the Dice coefficient to analyse the similarity of the banding patterns, and the unweighted pair group method using arithmetic averages (UPGMA) for cluster analysis. Relatedness of ≥ 80% was used to define the lineages .
In addition, PaLoc-CDT- multiplex PCR was used for detection of toxin genes tcdA, tcdB, cdtB and tcdC on selected isolates. The primers for the PCR were designed based on the published sequences of the pathogenicity locus (PaLoc) and binary toxin CDT (cdtA and cdtB) genes of the C. difficile strains. The primers for the binary toxin were designed so that the cdtB gene with the known deletion was not amplified.
Incidence, case fatality and statistical analysis
Data from the National Population Information System for the year 2007 were used as denominators to calculate age- and sex-specific incidence rates, including rate ratios and 95% confidence intervals. The case fatalities were calculated by dividing all deaths due to any cause within 7 and 30 days after positive diagnostic specimen for CDI, by the total number of CDI episodes. The genotyped isolates from severe cases, outbreaks, and those of unknown or other reason for sending were divided into two groups, type 027 and non-027. To compare outcomes (30-day case fatality) between the two groups, statistical comparison was only performed among the outbreak-associated patients with CDI episode and their isolates. The analyses between groups were carried out by logistic regression using the R statistical software version (R Development Core Team, Vienna, Austria) and continuous variables using the Mann–Whitney U-test.
In 2008, 6201 episodes of CDI among 5717 individual patients were identified; 4217 (68%) occurred in persons aged >64 years and 3679 (59%) in female patients. The annual incidence rate was 118.3 per 100 000 population. The rate increased by age and was highest in persons aged >84 years (1286.0 per 100 000 population) (Table 1). The rate was higher in female patients than in male patients, but the difference was statistically significant only in persons aged 15–44 years. Of the CDI episodes, 259 (4.2%) led to death within 7 days and 711 (11.5%) led to death within 30 days. The 30-day case fatality was highest in persons aged >84 years (22.0%) (Table 2). The 30-day case fatality was significantly higher in male patients than in female patients in two age groups, persons aged 65–74 years and persons aged 75–84 years (p-values <0.01 and <0.05, respectively). The overall annual incidence varied regionally (range by 21 HDs, 57.2–189.1 per 100 000 population; Fig. 1), as well as the 30-day case fatality (range by HDs, 2.2–15.0%).
|Age group (years)||Incidencea (number of episodes)|
|Female||Male||Rate ratio||95% CI||Total|
|1–14||18.6 (76)||14.3 (61)||1.3||0.9–1.8||16.4 (137)|
|15–44||39.9 (393)||25.3 (261)||1.6||1.3–1.8||32.4 (654)|
|45–64||75.0 (570)||82.5 (623)||0.9||0.8–1.0||78.7 (1193)|
|65–74||217.2 (543)||265.3 (564)||0.8||0.7–0.9||239.3 (1107)|
|75–84||577.4 (1131)||603.8 (715)||1.0||0.9–1.1||587.3 (1846)|
|>84||1308.6 (966)||1217.8 (298)||1.1||0.9–1.2||1286.0 (1264)|
|All||137.5 (3679)||98.3 (2522)||1.4||1.3–1.5||118.3 (6201)|
|Age group (years)||Number of deaths at 30 days (case fatality, %)|
|1–14||0 (0)||2 (3.3)||2 (1.5)|
|15–44||0 (0)||5 (1.9)||5 (0.8)|
|45–64||26 (4.6)||36 (5.8)||62 (5.2)|
|65–74||30 (5.5)||64 (11.4)||94 (8.5)|
|75–84||148 (13.1)||122 (17.1)||270 (14.6)|
|>84||202 (20.9)||76 (25.6)||278 (22.0)|
|All||406 (11.1)||305 (12.1)||711 (11.5)|
In total, 334 (5% of all episodes) isolates (one per patient) from 13/21 (62%) HDs were sent for genotyping (range in number of isolates sent by HDs, 1–110; Fig. 1). Of these isolates, 54 (16%) originated from severe cases, and 191 (57%) from outbreaks. For the rest of the isolates (n = 89; 27%), the reason for sending was either unknown or other (e.g. screening for the hypervirulent PCR ribotype 027) (Table 3).
|PCR ribotype||Case fatality, % (number of deaths/isolates sent)|
|Type 027||30.4 (7/23)||23.2 (19/82)||13.3 (2/15)|
|Non-027||19.4 (6/31)||10.1 (11/109)||6.8 (5/74)|
Of the isolates sent for genotyping, 120 (36%) were of PCR ribotype 027, 53 (16%) were of PCR ribotype 001, 19 (6%) were of types 002 and 014, and 29 (9%) were of PCR ribotypes other than type 027, which have deletions of different sizes in their tcdC genes (i.e. 023, 045, 075, 078 and 126). Overall, 57 distinct PCR ribotype profiles were identified. Among the 171 isolates analysed with both PFGE and PCR ribotyping, 47 PFGE types were detected among 30 PCR ribotypes. On several occasions PCR ribotypes could be divided further using PFGE typing (e.g. eight PFGE (sub)types were detected within PCR ribotype 014). PCR ribotype 027 was divided into five PFGE (sub)types (Fig. 2); however, among one of them (NAP1 subtype1) three different PCR ribotype profiles were identified (data not shown).
PCR ribotype 027 was found in 6/13 HDs (range, 1–42 isolates per HD), as well as the other ribotypes with tcdC deletions. Type 027 was found in four HDs where the incidences were among the five highest (range, 135.9–188.7 per 100 000 population), but also in two HDs with low incidence (77.8 and 93.6 per 100 000 population). The HDs where type 027 was found had the highest 30-day case fatalities among the HDs that sent isolates for genotyping, with only one exception (Fig. 1). There was no significant correlation between the ranks of HDs by incidence and case fatality (correlation coefficient, 0.23).
Among the 334 patients with CDI episodes from whom an isolate was sent for typing, the 30-day case fatality was 15.0% (Table 3). Type 027 was the most common ribotype among severe cases (n = 23; 43%) and outbreaks (n = 82; 43%). The patients with CDI episodes caused by type 027 isolates were older than those with episodes caused by non-027 isolates (median age 81.8 vs 77.0, respectively; p <0.01). Among patients with CDI episodes related to outbreaks, the 30-day case fatality was higher when type 027 was isolated (23.2% vs 10.1% for non-027 isolates; p <0.05). However, when the age difference was taken into account by standardization, no statistically significant difference was detected.
Our study showed the first nationwide estimates of incidence and case fatality of CDI in Finland. Although many regional and outbreak-related investigations have been carried out, to our knowledge, this is the first population-based study on CDI incidence and case fatality.
In the current literature, most estimates of CDI incidence available, especially from Europe, are proportioned per patient-days or hospital discharges [3,12], and hence not comparable with the results of this study. The observed overall incidence of CDI in Finland (118.3/100 000 population in 2008) was lower than in Quebec in 2003 . However, in a couple of regions the rates were at a similar level. In Quebec the CDI incidence rose from 35.6/100 000 population to 156.3/100 000 in 1991–2003. The emergence of PCR ribotype 027 might have contributed to the rise. In Finland, the hospital discharges for which CDI was listed as any diagnosis doubled from 1996 to 2004 . According to a recent Canadian report by Miller et al. there was a strong correlation between the presence of type 027 and incidence of CDI when their data were analysed by province. However, in our study no clear correlation between high incidence and occurrence of PCR ribotype 027 was found regionally in 2008. The incidence rate was significantly higher in females than in males aged 15–44 years. This might be explained by qualitatively or quantitatively differing antibiotic consumption patterns between the genders (e.g. treatment of urinary tract infections in women).
The overall 30-day case fatality proportion (11.5%) (i.e. crude mortality rate) was lower than in Quebec in 2003 (13.8%) , although in four regions the proportion was even higher. The regions where PCR ribotype 027 was found seemed to suffer from higher 30-day case fatality. In concordance with this, a relation between increased mortality and/or more severe disease and type 027 has also been found in the Netherlands and Quebec [15–17], but not in England .
The isolates from severe cases and suspected outbreaks were genotyped to study the differences in virulence and epidemic potential of different ribotypes. Among the isolates sent for genotyping, type 027 was the most prevalent PCR ribotype, followed by types 001, 002 and 014. Type 027 was the dominant PCR ribotype among the isolates from severe cases and outbreaks. The number of cases was quite small, but among these isolates the 30-day case fatality was more than double for type 027 when compared with non-027 types. However, this difference was no longer statistically significant after age-standardization. This may imply that type 027 has a substantial effect on the elderly population because they are likely to have more underlying illnesses. This is basically in concordance with the findings of Miller et al. , who found a clear correlation between the presence of type 027 and more severe disease and outcomes among patients aged 60–90 years.
We detected major regional variations in CDI incidence and case fatality. According to a laboratory survey conducted in 2006 and updated in 2008, the C. difficile diagnostics methodology is rather uniform in Finland . Although our material is nationally representative, there are some regions with low incidence but high case fatality, which may be explained by a lower diagnostic activity. Differences in diagnostic methodology and sensitivity of the methods may also have an effect. Indeed, the regions with laboratories using diagnostic PCR assays had relatively high CDI incidences (i.e. second, fifth and sixth highest of the 21 regions). There are no accurate data available for diagnostic activity in 2008, though in 2006 considerable regional variation was detected . The small number of cases (and the relatively higher proportion of deaths) in some small regions might also explain the differences.
Our aim was to support outbreak control measures locally and detect strains with increased virulence properties by genotyping isolates from severe cases and outbreaks. In relation to methods used at the reference laboratory, having more than one genotyping method proved useful in sorting out local epidemics in hospital wards; generally, PFGE typing was more discriminatory compared with PCR ribotyping, as previously reported , with few exceptions. One of the challenges is the question of how to select a representative sample of isolates from all CDI cases for genotyping. The simplest way would be to request the clinical microbiology laboratories to send a predesignated proportion of C. difficile isolates (e.g. every tenth isolate) to the national reference laboratory; however, in this study the aim was to understand especially the genotypes and virulence factors of the isolates associated with severe cases and support the hospitals struggling with persistent C. difficile problems. In Finland, it can be argued that the data on molecular epidemiology of CDI in 2008 are not nationally representative because the isolates were received from only 5% of the reported CDI cases. Moreover, there was notable regional variation in numbers of isolates. No isolates were sent from eight of the 21 regions, and unfortunately, the two regions with highest case fatalities did not send any isolates at all. Either CDI is not regarded as causing notable problems in these regions, or genotyping is seen as useless in solving these issues, or the criteria for sending isolates are considered too strict. The problem may lie in the flow of information between the laboratory and the clinic: an indication about severe disease/suspected outbreak is usually not included in the background information on the samples received by the clinical laboratories. In addition, the definition of an outbreak and the threshold for reaction might vary between hospitals. The 30-day case fatality was 15% among the cases from which isolates were received, compared with the overall case fatality rate of all the cases reported to the NIDR (11.5%).
In conclusion, CDI affected mostly elderly people. The incidence and 30-day case fatality were highest in persons aged >84 years. The overall incidence and case fatality varied regionally. PCR ribotype 027 seemed to be associated with higher case fatality than the non-027 ribotypes. However, much of this can be explained by the settings and patient populations where the outbreaks are occurring.
The study was funded using institutional resources. The authors have no conflicts of interest.