Background
Chikungunya infection (CHIKV) is a self-limiting, two-stage disease caused by an
Aedes mosquito-transmitted virus. The CHIKV acute stage typically consisting of fever, rash and multiple joint disorders is frequently followed by a chronic stage, mainly characterized by peripheral joint disorders [
1‐
9], which it is estimated can persist up to five years [
10]. Different studies have investigated prognostic factors of CHIKV non-recovery, generally assimilating non-recovery with the persistence of rheumatic disorders, but with non-standardized definitions of either “recovery” or “post-CHIKV rheumatic disorders” [
3,
11,
12]. Moreover, if a severe acute stage of CHIKV was reported to be associated with persistent rheumatic disorders, no standard definition of severity was used [
6,
8,
13]. No prognostic study distinguished between arthralgia and arthritis, although this distinction makes sense in terms of management. While arthralgia is the most common post-CHIKV rheumatic disorder (RD), arthritis is much more incapacitating and can be the first step towards destructive inflammatory rheumatism [
14,
15]. Thus, the identification of specific prognostic factors for post-CHIKV arthritis should be of interest as an aid to therapeutic decision-making to control the inflammation process early enough before any joint destruction occurs, as has recently been proposed by some rheumatologists [
15,
16].
In a previous study, we observed that, in a cohort of French gendarmes (members of a French military force responsible for carrying out police duties among civilian populations), CHIKV-infected patients (CHIKV+) presented a higher frequency and intensity of joint disorders and an impaired quality of life 30 months after infection, whether they self-considered themselves recovered or not, than non-infected (CHIKV-) subjects [
2]. The objective of the present study was to identify specific prognostic factors of persistent arthritis in the same cohort, focusing on the symptoms of infection during the initial stages of the disease. In order to take into account the patients’ perception of the disease, the association of these factors with self-perceived recovery was also studied.
Methods
Ethical aspect
This study is based on data collected retrospectively at two different times among the same population of gendarmes. A major outbreak of CHIKV in Reunion Island ended in June 2006. In that context, it was decided that a study on the prevalence and consequences of infection among gendarmes would be conducted, because CHIKV was considered to be a possible work-related disease. All the gendarmes on duty in Reunion Island were offered a serological test for CHIKV and self-report questionnaires by their military physician. Testing was considered to be part of routine care, but patients were informed about the study by a written notice attached to the self-questionnaire and the military physician received their oral consent before inclusion. Only those who accepted the blood test and questionnaire were included. In 2008, follow-up questionnaires were offered to each participant of the 2006 study. Questionnaires were sent out by mail with written information on the study’s aim, which explained that it was an anonymous epidemiological study and instructed patients who agreed to participate, to send back the completed questionnaire using the pre-identified attached envelope. The two thirds of the patients who returned the questionnaires with self-reported, hand-written data were thus considered to have consented to participate in the study. Both the studies and their consent procedures were approved by the Clinical Research Committee of the Central Directorate of the Army Health Service (DCSSA). Data were analyzed anonymously.
Study population and definitions
The study population was made up of the French gendarmes who completed both the 2006 survey (n = 662) [
1] and the 2008 survey (n = 404) [
2], minus one subject excluded for redundant data. The 259 non-respondents in 2008 did not differ from the 403 included in terms of proportion of CHIKV-infected patients, sex and age.
In accordance with previous studies performed on the same cohort [
1,
2], and due to the fact that serology results were highly concordant with self declaration; subjects were considered CHIKV + (n = 101) or CHIKV- (n = 302) according to their self declaration.
In both the 2006 and 2008 studies, participants were asked to mention which joints presented pain, stiffness or swelling at different times. To distinguish between arthralgia and arthritis, rheumatic disorders were classified in three categories: 1) no rheumatic disorders, if no pain, stiffness or swelling was reported, 2) arthralgia, if only joint pain and/or stiffness was reported, 3) arthritis, if joint swelling was reported in addition to pain and/or stiffness, regardless of the number of joints involved.
Data on initial CHIKV stages were those collected in the 2006 survey, which concerned acute stage symptoms and those persisting at the time of the survey (i.e. a median lapse of six months after CHIKV infection), the latter being called symptoms of the “early chronic stage” or “early chronic symptoms”. Clinical symptoms of the acute stage (fever, rash, joint pain, joint swelling, muscle pain, headache) were taken into account as dichotomous variables coded “yes” for present and “no” for absent. As fever was common, the level of fever (>39° or ≤39° according to the median), and its duration (>or ≤ 2 days [median duration]) were added. Asthenia was coded “yes” if it was self-evaluated as “quite significant”, “significant”, “very significant” or “totally disabling” on the 6-item Likert scale, and “no” if evaluated as “absent” or “minor”, and mood was coded “depressed” if it was self-evaluated as “affected” or “totally depressed” on the 4-item Likert scale, and “not depressed” if it was evaluated as “normal or weak but confident”. The notion of sick leave during the acute stage was added using two variables: sick leave (yes/no) and duration of sick leave (≤4 days or >4 days; [4 days = median duration]). Three items for the early chronic stage were kept: joint disorders (categorized as no rheumatic disorders/arthralgia/arthritis); asthenia (yes, no) and depressed mood (yes, no), using the same cut-off as that used on the Likert scales.
As age and rheumatic comorbidity have been associated with recovery in different studies [
3,
11,
12], we also integrated in the models age (coded: ≤ 40 and >40 years, according to the median age of the participants in 2006), and a “comorbidity” variable coded “yes” if the patient had reported a history of “traumas in tendons, joints or bones” or “osteo-articular diseases”.
Finally, self-perceived recovery was defined by the answer “yes” or “no” to the 2008 survey question: “Do you consider yourself recovered?”.
Statistical analysis
To identify the prognostic factors associated with persistent arthralgia or arthritis 30 months after infection (median duration of follow-up in 2008), we first analyzed the entire 2008 study sample using a multinomial logistic regression with “no rheumatic disorders” as the reference category. To implement the regression model, all the predictive variables were first assessed individually using univariate regression models. Then, variables with a significant p-value < 0.20 were included in the multivariate regression model, and a backward stepwise procedure was used to keep as predictors the variables with p-value <0.05. These regressions were performed using STATA software, version 9.
The CHIKV acute stage symptoms were largely shared by CHIKV + patients and absent in CHIKV- patients. This led to interactions and colinearity in the multivariate regression model. Identifying the independent predictive role of CHIKV infection in RDs was of epidemiological interest and keeping CHIKV- subjects in the study increased the power of the study, but with regard to the identification of CHIKV initial symptoms of severity that could help manage CHIKV + patients, keeping CHIKV- patients in analyses was useless. Thus, our second step was to focus on CHIKV + patients, and in order to avoid interaction and colinearity problems, a multiple correspondence analysis (MCA) was used. With the help of MCA, we sought to estimate predictors of both rheumatic disorders (separately for arthralgia and arthritis) and self-estimated recovery, but also to find a standardized definition of severity of early CHIKV infection which would be easy to identify and reproduce by clinicians. For these analyses, XL-Stat 9 and R 2.15.1 softwares were used.
The MCA displays the individuals and variable categories as labeled points in a multiple dimensional space. It decomposes all the information represented by variables and individuals into multiple factorial axes, such that each successive factorial axis captures a part of the total information in decreasing order [
17]. Thus, the first two factorial axes capture the maximum amount of information available in the data. With the help of these two factorial axes, the graphic display of variable categories can reveal the structural relation between variables. In the graph, the more distant a variable category is from the origin, the higher its contribution is (i.e. the contribution of categories near the origin is small), but outliers situated too far from the origin are no longer informative. In the study, sex categories were thus excluded as there were only 3 females, which led to a distortion of the MCA graph (the category “women” representing an outlier that was too far from the origin). Most interestingly, the interpretation of data associations can easily be read on the MCA graph: the closer the variable categories are on the graph, the stronger they are associated. Here, the categories of self-perceived recovery at 30 months and rheumatic disorders were superimposed on the MCA graph to observe which symptoms were closer to (i.e. were associated with) self-perceived recovery, arthralgia and arthritis. The strength of these associations was tested in binary analyses using the Chi
2 test.
One property of MCA is the possibility of providing a severity scale from 0 (corresponding to the theoretically least severe case) to 100 (corresponding to the theoretically most severe case) for each factorial axis [
18]. The score is calculated using the following formula: [1-(X-Xmin)/(X-Xmax)]*100, where X = coordinate of an individual on a factorial axis F, Xmin = minimum coordinate on F, Xmax = maximum coordinate on F.
This formula was used to calculate 3 severity scores based on the MCA coordinates of variables on the first factorial axis, which was the most informative. The first score was calculated using both acute and early chronic stage symptoms, the second using only the acute symptoms and the third using only the early chronic symptoms. Associations between these scores and self-perceived recovery, arthritis and arthralgia at 30 months were also tested using Kruskal-Wallis tests.
Discussion
Both analyses underlined the determinative role of acute depressed mood and early chronification of arthritis in the risk of persistent arthritis at 30 months. However, when CHIKV- patients were taken into account (multiple regression model), the main determinants of long-term rheumatic disorders whether inflammatory or not were CHIKV infection and underlying rheumatic comorbidities. The MCA restricted to CHIKV + patients enabled us to demonstrate that CHIKV initial symptoms were highly interrelated, and to calculate severity scores that were strongly associated with persistence of arthritis after 30 months. Moreover, it made it possible to isolate simple initial severity indicators easily identifiable by clinicians for patient management, such as long sick leave and joint swelling during the acute stage, in addition to chronification of depressed mood and arthritis at six months. Although long term arthralgia was linked to underlying diseases according to the first analysis, the MCA did not find it to be associated with CHIKV severity.
The association between severity of the acute CHIKV stage and persistent rheumatic disorders has already been described by some studies [
6,
8,
13]. But the present study is the first to quantify the overall severity of the acute and early chronic stages of CHIKV infection in the form of severity scores, and thus to objectify the association between initial severity and long-term post-CHIKV rheumatic disorders and self-perceived recovery. These scores showed that severity of both acute and early chronic stages of CHIKV infection was associated with long-term rheumatic disorders. However, the association was much stronger and the severity scores higher for long-term arthritis than for arthralgia, suggesting that the more severe the initial CHIKV infection, the more likely it was that patients would develop long-term arthritis. At a mild level of severity, they would develop arthralgia (with a far lower risk of joint destruction), and at a lower level no rheumatic consequence. The location of 30-month arthralgia on the MCA graph confirmed this hypothesis.
In addition, our study is the first to distinguish common arthralgia from arthritis among post-CHIKV rheumatic disorders, which made it possible to identify specific prognostic factors of arthritis. In a recent study, Gérardin
et al.[
19] distinguished relapsing post-CHIKV rheumatic musculoskeletal pain (RMSP) from lingering RMSP based on the frequency of episodes of pain. They found that age and initial severity (defined by severe rheumatic involvement and level of CHIKV IgG titers during the acute phase) were predictive of both lingering and relapsing RMSP [
19]. The distinction based on intensity of rheumatic disorders, arthritis being more severe than arthralgia, seems more relevant, since it made it possible to isolate specific predictors and has consequences for patient management, in particular for therapeutic decisions [
15,
16]. The study was based on self-reported symptoms, so the diagnosis of arthritis cannot be certain. However, after discussion with clinicians, swelling was considered specific enough to synovitis to warrant proposing this distinction. Moreover, the consistency of data between 2008 and 2006, as well as with incapacitation reported on the quality of life SF-36 questionnaire completed in 2008 (data not shown), reinforced the reliability of patients’ declarations.
The MCA graph enabled an easy visual identification of associations between certain symptoms and long-term arthritis or self-perceived non-recovery. Some of the symptoms identified, such as joint swelling during the acute stage, have already been reported as being associated with CHIKV recovery [
11], but with a different definition of recovery. In our results, there was a clear continuum in the inflammatory symptoms, as swelling in the acute stage was also found in the early and late chronic stages of CHIKV. Depression was reported to be frequent after CHIKV infection [
2,
4]. In our study, depressed mood during both acute and early chronic stages were prognostic of both persistence of arthritis at 30 months and self-reported non-recovery. This link suggests that attention should be paid to this symptom, which may be indicative of either the severity of the disease or the patient’s anxiety concerning delay in recovery. However, a direct action of the CHIK virus on the brain should be explored.
The MCA found other indirect markers of severity associated with long-term arthritis, such as absence from work and its duration. This is consistent with the notion that the more severe the acute stage is, the greater the risk will be to develop arthritis. Last but not least, reporting arthritis during the early chronic stage was found to be associated with the long-term persistence of arthritis, while early arthralgia was not. This was also found with the multinomial logistic regression, which identified early chronic arthralgia as being associated with late persistent arthralgia and early arthritis with late arthritis. This fact is of major importance, because it supports the recommendation that early detection of persistent arthritis within the first 6 months should be carried out in order to initiate adequate treatment. Rheumatologists in Reunion Island and India have recommended that the treatment should be based on disease-modifying antirheumatic drugs, with methotrexate as the first-line therapy to control the inflammation process before joint destruction occurs [
15,
16].
In no analysis was age contributive, although it has been widely reported to be associated with CHIKV recovery [
3,
8,
11]. This might be due to the specificity of our population, made up of young, healthy workers, none of whom were over 55 years of age.
Moreover, according to the MCA, the contribution of comorbidity was very small, although it was also frequently reported [
3,
8,
11]. The frequency of previous musculoskeletal pain or traumatic history among patients suffering from post-CHIKV rheumatic disorders often led to the consideration that CHIKV may reveal or stimulate underlying rheumatic diseases. Our results on the entire cohort sample are in favor of this conclusion, as the association between rheumatic comorbidity and late rheumatic disorders existed for both infected and non-infected patients. It seemed more likely that the association could be explained by arthralgia, as it was the most frequent rheumatic disorder when considering CHIKV- subjects and comorbidity did not contribute to long-term arthritis according to the MCA.
The two different assessments of recovery (self-perceived or evaluated by persistent rheumatic disorders at 30 months) led to differences in the prognostic factors found to be associated with recovery. Self-perceived recovery was only associated with the severity score based on “early chronic stage” symptoms, suggesting that the severity of entry into the chronic stage is the main predictor of the perception of recovery. Since the question about recovery was asked 30 months after the acute infection, it seems logical that patients who did not enter the chronic stage (no or few symptoms reported after six months of infection) considered themselves recovered regardless of the severity of their acute stage. On the other hand, most of the subjects who declared themselves recovered did report rheumatic disorders (35% reported arthralgia and 37% arthritis, data not shown). They may either have considered that their symptoms were not linked to CHIKV or considered that they were a bearable after-effect. It is therefore up to physicians to estimate the severity of symptoms and their possible consequences on joint integrity in order to decide on future management. The existence of severity characteristics during the previous stages of CHIKV disease should be an additional aid to therapeutic decision-making.
Acknowledgments
The authors are grateful to the Central Directorate of the Army Health Service (DCSSA), the Direction Générale de la Gendarmerie Nationale, which authorized the study, and all the French gendarmes who participated in the cohort. They would like to add special thanks to Professor Boutin and Dr. Queyriaux, who initiated the cohort by performing the first enquiry in 2006.
Funding
This work was supported by a grant from the Central Directorate of the Army Health Service. [2008-RC −15].
Competing interests
The authors declared no competing interests.
Authors’ contributions
HMY performed the statistical analysis and wrote the first draft of the manuscript. FS designed and coordinated the cohort study and helped to draft the manuscript. XD participated in the present study design and supervision of the statistical analysis and helped draft the manuscript. CM participated in the design and coordination of the cohort study, in the design of the present study and supervision of the analyses and was in charge of the final draft of the manuscript. All authors read and approved the final manuscript.