# Pityriasis rosea: elucidation of environmental factors in modulated autoagressive etiology and dengue virus infection.

IntroductionPityriasis rosea (PR) is suspected to be associated with an infection. However, an exact cause has not been found. Drago et al. reported human herpesvirus 7 to be the causative agent (1). Other investigators reported findings supporting and refuting such an association. However, the distinct clinical course, a lack of recurrences in most of the patients, and the presence of temporal case clustering support an infectious etiology. Furthermore, seasonal variation, association with respiratory tract infections, and a history of contact with PR patients in some patients do support an infectious etiology (2).

Cluster analysis is a useful approach for elucidating possible infectious etiologies. Several studies have evaluated the presence of clustering in PR (3-12). In 1982, Messenger et al. reported significant spatial-temporal clustering only in female patients with PR and a temporal cluster of 16 patients within a 28-day period (3). However, there was no control and the impact of seasonal variation was not studied. Later on, some studies reported seasonal variation and/or case clustering for patients with PR (4, 8-11), whereas others did not find any significant association with seasonal variation and incidence of PR (6, 12). To the best of our knowledge, no study has reported an association of dengue fever with PR. We thus report here a retrospective study investigating the epidemiology of PR and the incidence of dengue fever and its association with PR at a tertiary referral center in Assam.

Methods

The study was conducted at a tertiary referral center in Guwahati, Assam, India. We searched for and retrieved all medical records of patients diagnosed with PR by dermatologists from December 1st, 2014 to July 31st, 2017. The diagnosis was made only if the patient had fulfilled at least three out of the following four clinical features: 1) herald patch, 2) peripheral collaret scales, 3) predominant truncal and proximal limb distribution of the lesions, and 4) orientation of lesions along the lines of cleavage. These diagnostic criteria were laid down and validated by us (13, 14). For each visit by every patient, we retrieved data for the monthly mean air temperature, mean total rainfall, and mean relative humidity. PR patients that had dengue fever were studied along with healthy controls. The detection of [NS.sub.1] antigen was done using the Panbio Dengue Early enzyme-linked immunosorbent assay (ELISA) (Inverness Medical Innovations, Australia). The detection of IgM antibodies was done using the Dengue-IgM capture ELISA kit (National Institute of Virology, Pune). IgG anti-dengue antibodies were detected using the dengue IgG capture ELISA (PanBio Pty Ltd, Queensland, Australia).

The following steps were used for the statistical analysis: a) 2 x 2 contingency tables were drawn to calculate the odds ratio (OR) and risk ratio (RR), as well as a chi-square test, and finally a two-tailed Fisher's exact test (p < 0.05 was considered statistically significant);

Temporal clustering was investigated using a regression model. The monthly incidence of PR was taken as a dependent variable, with meteorological variables such as monthly average temperature, monthly average precipitation, and the occurrence of dengue fever considered as independent variables. The statistical relationship was evaluated using Pearson's correlation analysis in the statistical software package SPSS (version 20.0, SPSS Inc., Chicago, IL) and online statistics programs.

Time-series analysis

Monthly PR and dengue incidences were cross-correlated using the cross-correlation function (CCF). In a cross-correlation in which the direction of influence between two time series is hypothesized, the influential time series is called the "input" time series and the affected time series is called the "output" time series. The application of cross-correlations in this text infers that the input time series refers to the incidence of dengue in a patient and the output time series refers to an occurrence of an auto-immune response to the dengue virus manifesting as a PR rash.

Results

A total of 136 PR patients were found to fulfll the diagnostic criteria. The male:female ratio was 1:1.13. They were between 13 months and 59 years old with maximum incidences in the age clusters 2029 and 30-39 (Tables 1 and 2).

For the epidemiological data analysis, the seasonality plot indicates a trend characterized by a peak in post-monsoon and winters (September-January, peak month November) and a trough in summers (peak, April), and the magnitude of the seasonal variation increases at the same rate as the yearly mean levels. Therefore, we tested this distribution pattern to determine whether it was statistically significant. The expected incidence for 12 months was calculated for a year from the total number of new PR patients and the number of hospital working days in each month during the same year. Then the mean [+ or -] standard deviation of 3 years was obtained for each month for the expected number of PR patients. Statistical tests were performed to compare actual and expected numbers of first visits during each month. Statistical signifcance was detected in September, November, and December (p < 0.01 for 2014 and 2017, but not in the 2016 seasonal cohort) and also in January and February (p < 0.05 for 2016 and 2017).

Regarding precipitation and temperature as independent predictive parameters for the incidence of PR, it was found that heavy rainfall is associated with decreased incidence of PR (this correlates with our hypothesis of dengue virus being one of the etiological factors in the development of PR because high rainfall is associated with decreased breeding of the dengue vector (i.e., Aedes mosquitoes) and, as discussed above, increased PR was observed with a drop in temperature.

Temperature and pityriasis rosea (PR) incidence

The regression equation for Y (where Y = PR incidence and X = temperature in Celsius) was y = -0.61521X + 58.50307 (Fig. 1a). Our interpretation is that the negative value showed an inverse relationship; that is, the incidence of PR increased with decreased temperatures.

Rainfall and PR incidence

The regression equation for Y (where Y = PR incidence and X = rainfall in mm) was y = -0.05421X + 19.25344 (Fig. 1b). Our interpretation is that increased rainfall was associated with decreased PR incidence.

The results from linear regression plots were further analyzed for Pearson's coefficient, and we found that the monthly incidence of PR is significantly associated with months with less rainfall (R = -0.55, p = 0.0001). Such an incidence is also associated with the colder months, although the association is insignificant (R = -0.38; p = 0.23).

PR and dengue incidence

The regression equation for Y (where X = incidence of Ns1Ag-positive dengue cases per month and Y = incidence of PR per month) is y = 2.68596X + 3.0778.

The correlation coefficient (PMMC) r was found to be 0.8714 (p = 0.0002; highly significant), which shows a positive correlation between the incidence of Ns1Ag or antibody positivity and PR (Fig. 1c).

I. Cross correlation Function-SARIMA model results - PR:

a) Autocorrelation (ACF) and partial autocorrelation function (PACF) for PR incidence (Figs. 2a-2c):

ACF and PACF plots were deployed to identify patterns in the above data, which are stationary on both mean and variance, to identify the presence of AR (autoregressive) and MA (moving average) components in the residuals. The ACF function shows a perfect sinusoidal pattern with a spike at lag 1; on extrapolating the data to the PAC function, the same correlation is seen at lag 1 (p = 0.037).

b) SARIMA forecast for PR incidence:

For prewhitening, the model SARIMA (p = 1, d = 0, q = 0, P = 2, D = 1, Q = 2 p = 1, d = 0, q = 0, P = 2, D = 1, Q = 2) was selected. Strong negative correlation coefficients were found at lags of the 7th and 8th months. Weak negative associations were found at lags of 7 to 9 months (Fig. 2d).

II. Cross correlation Function-SARIMA model results - PR with preceding history of Ns1Ag or antibody positivity (Figs. 3a-3c): Both the ACF and PACF functions showed a significant positive correlation at 0 and 1 lag (p = 0.028).

c) SARIMA forecast for PR cases with Ns1Ag positivity or antibody incidence (Fig. 3d):

A significant positive correlation was found at lag 12 months (p = 0.04). Of the 136 PR patients, 38 were seropositive for either/both IgG and IgM or Ns1Ag (27.94%) in contrast to 19 (13.97%) Ns1Ag or IgM and IgG antibody seropositive cases in 136 matched controls. Seropositivity for Ns1Ag or antibody in PR patients was significantly higher than those found in controls (OR = 2.3878, 95% confdence interval (CI) = 1.294 to 4.4061; RR = 2, 95% CI = 1.217 to 3.2868; Yates [x.sup.2] = 7.19 p = 0.0073; two-tailed Fisher's exact probability test p = 0.00698), indicating a higher risk of developing PR with a preceding history of dengue viral infection. Furthermore, the bivariate Granger causality for PR incidence and NsAg1 and/or antibody positivity revealed that the incidence of seropositivity to dengue virus infection can be used to forecast the development of PR rash as a significantly positive correlation at lag 2 months (F = 10.3, p = 0.0237).

Discussion

This retrospective study found temporal clusters of PR in the dry winter months of September to January, with the correlation being statistically significant for the months of September, November, and December (p < 0.01 for 2014 and 2017 but not in the 2016 seasonal cohort) and also in January and February (p < 0.05 for 2016 and 2017); however, the overall correlation was weak. The association between the infectious etiology, especially human herpesvirus 6 and 7, with PR is controversial; reasonable evidence suggests that PR is not associated with cytomegalovirus, Epstein-Barr virus, parvovirus B19, picornavirus, influenza and parainfluenza viruses, Legionella spp., Mycoplasma spp., and Chlamydia spp. (15, 16). Interestingly, in this study, the retrospective histories of dengue fever emerged as a significant correlate against a matched cohort of 136 patients visiting the dermatology outpatient department for other ailments. The average duration between the onset of PR and dengue was 78.34 days. The most interesting example of PR with dengue was that of a pair of twins, both of whom presented with typical PR lesions with a history of dengue fever 5 weeks earlier. The outbreaks of dengue occurred from August to October, indicating increased vector transmission in the monsoon and postmonsoon periods. However, we admit that the monthly rate of dengue fever may only be a confounding variable. The signifcance of this should be investigated further in future studies.

The age and sex distribution in our study is in line with other epidemiological studies on PR (3--7). Some of these studies reported a higher incidence of PR during winter (10, 11), whereas one reported a higher incidence in the early rainy season (8) and some reported no seasonal variation (6, 12) (Table 1).

Our study has certain limitations. The most important limitation of this study is that the data were collected at one clinic in one geographical location only. Having adequate resources, we previously performed epidemiological studies in multiple geographical locations (9). However, we lack a similar scale of material support in this study. Confounding variables could thus negate the generalization of our results to other geographical locations and other clinical settings. We also failed to elucidate the underlying mechanisms for our results being similar to or different from those of other investigators (3-11) to an acceptable level of evidence. Although our study followed the morphological features delineated by Chuh et al., there is another proposed classification of PR by Drago et al., in which PR variants, including atypical forms, are classified on the basis of differences in pathogenesis, clinical fea tures, and the course of the disease (19). This classification included pregnant patients, who were not part of our study population.

The temporal clustering documented in this study might suggest a role of dengue virus as an autoimmune trigger, modulated by environmental factors that cause the syndrome in previously unexposed, genetically susceptible individuals, with asymptomatic infection leading to protective immunity in the majority of the population. The fact that PR is self-limited strongly suggests a defnitive immune response that terminates the inflammatory process.

Conclusions

We found temporal clusters of PR in the dry winter months of September to January, with the correlation being statistically significant for the amount of rainfall. Interestingly, retrospective histories of dengue fever emerged as a significant correlate. Thus, temporal clustering and dengue infection as significant correlates may imply the infectious etiology of PR. However, the signifcance of this warrants further multicentric investigations, preferably at different geographic locations.

References

(1.) Drago F, Ranieri E, Malaguti F, Losi E, Rebora A. Human herpesvirus 7 in pityriasis rosea. Lancet. 1997;349:1367-8.

(2.) Chuh A, Zawar V, Sciallis GF, Lee A. The diagnostic criteria of pityriasis rosea and Gianotti-Crosti syndrome--a protocol to establish diagnostic criteria of skin diseases. J R Coll Physicians Edinb. 2015;45:218-25.

(3.) Messenger AG, Knox EG, Summerly R, Muston HL, Ilderton E. Case clustering in pityriasis rosea: support for role of an infective agent. Br Med J (Clin Res Ed). 1982;284:371-3.

(4.) Harman M, Aytekin S, Akdeniz S, Inaloz HS. An epidemiological study of pityriasis rosea in the eastern Anatolia. Eur J Epidemiol. 1998;14:495-7.

(5.) Nanda A, Al-Hasawi F, Alsaleh QA. A prospective survey of pediatric dermatology clinic patients in Kuwait: an analysis of 10,000 cases. Pediatr Dermatol. 1999; 16:6-11.

(6.) Tay YK, Goh CL. One-year review of pityriasis rosea at the National Skin Centre, Singapore. Ann Acad Med Singapore. 1999;28:829-31.

(7.) Traore A, Korsaga-Some N, Niamba P, Barro F, Sanou I, Drabo YJ. Pityriasis rosea in secondary schools in Ouagadougou, Burkina Faso. Ann Dermatol Venereol. 2001;128:605-9. [French]

(8.) Chuh AA, Lee A, Molinari N. Case clustering in pityriasis rosea: a multicenter epidemiologic study in primary care settings in Hong Kong. Arch Dermatol. 2003; 139:489-93.

(9.) Chuh AA, Molinari N, Sciallis G, Harman M, Akdeniz S, Nanda A. Temporal case clustering in pityriasis rosea: a regression analysis on 1379 patients in Minnesota, Kuwait, and Diyarbakir, Turkey. Arch Dermatol. 2005;141:767-71.

(10.) Sharma L, Srivastava K. Clinicoepidemiological study of pityriasis rosea. Indian J Dermatol Venereol Leprol. 2008;74:647-9.

(11.) Ayanlowo O, Akinkugbe A, Olumide Y. The pityriasis rosea calendar: a 7 year review of seasonal variation, age and sex distribution. Nig Q J Hosp Med. 2010;20: 29-31.

(12.) Ganguly S. A clinicoepidemiological study of pityriasis rosea in south India. Skinmed. 2013;11:141-6.

(13.) Chuh AAT. Diagnostic criteria for pityriasis rosea - a prospective case control study for assessment of validity. J Eur Acad Dermatol Venereol. 2003;17:101-3.

(14.) Zawar V, Chuh A. Applicability of proposed diagnostic criteria of pityriasis rosea: results of a prospective case-control study in India. Indian J Dermatol. 2013;58: 439-42.

(15.) Chuh AA, Chan HH. Prospective case-control study of chlamydia, legionella and mycoplasma infections in patients with pityriasis rosea. Eur J Dermatol. 2002; 12:170-3.

(16.) Chuh A, Chan H, Zawar V. Pityriasis rosea--evidence for and against an infectious aetiology. Epidemiol Infect. 2004;132:381-90.

(17.) Mubki TF, Bin Dayel SA, Kadry R. A case of pityriasis rosea concurrent with the novel influenza A (H1N1) infection. Pediatr Dermatol. 2011;28:341-2.

(18.) Kwon NH, Kim JE, Cho BK, Park HJ. A novel influenza a (H1N1) virus as a possible cause of pityriasis rosea? J Eur Acad Dermatol Venereol. 2011;25:368-9.

(19.) Drago F, Ciccarese G, Rebora A, Broccolo F, Parodi A. Pityriasis rosea: a comprehensive classification. Dermatology. 2016;232:431-7.

Supplementary data and figures

Statistical methods used

For the cross-correlation function, the correlation coefficient, or Pearson product-moment correlation coefficient (PMCC), was calculated using the formula:

[mathematical expression not reproducible]

where n is the total number of samples, [x.sub.i] ([x.sub.1], [x.sub.2], . . . , [x.sub.n]) are the x values, [y.sub.t] is the y values, and r (PMCC) is a numerical value between -1 and 1 that expresses the strength of the linear relationship between two variables. When r is closer to 1 it indicates a stronger positive relationship.

The cross-correlation calculation for univariate time series was calculated as follows:

The cross-correlation of time series requires the time series to be stationary and prewhitened. Stationarity is defined by a constant mean and equal variance at all times, and it can be achieved by detrending or differencing. Prewhitening removes spurious correlations based on temporal dependencies between adjacent values of the input time series and it removes these influences from the output time series. The parameters lambda, d, D, and seasonality were used to apply a Box-Cox transformation and (non-)seasonal differencing in order to induce stationarity of the time series. The confdence interval was computed assuming a white noise time series (CI type = white noise).

SARIMA modeling

Multiplicative seasonal auto-regressive integrated moving average (SARIMA) models with all possible combinations of parameters p, q, P, Q [member of]{0, 1, 2} and with d, D [member of]{0, 1} were evaluated using Akaike's information criterion (AIC) on untransformed and logarithmically transformed monthly meteorological data from 2014 to 2017. The selected SARIMA model was then used to prewhiten meteorological data series, PR, and Ns1Ag positivity and PR incidence time series.

For the formulas used, the seasonal ARIMA model incorporates both non-seasonal and seasonal factors in a multiplicative model. One shorthand notation for the model is

ARIMA(p, d, q) x (P, D, Q)S, with p = non-seasonal AR order, d = non-seasonal differencing, q = non-seasonal MA order, P = seasonal AR order, D = seasonal differencing, Q = seasonal MA order, and S = time span of repeating seasonal pattern.

The model could be written more formally as:

(1) [PHI]([B.sup.s])[phi](B)([[x.sub.t] - [mu]) = [THETA]([B.srp.s])[theta](B)[w.sub.t]

The non-seasonal components are:

AR: [phi](B) = 1 - [[phi].sub.1]B - . . . - [[phi].sub.p][B.sup.p]

MA: [theta](B) = 1 + [[theta].sub.]B + . . . + [theta].sub.q][B.sup.q]

The seasonal components are:

Seasonal AR: [PHI]([B.sup.S]) = 1 - [[PHI].sub.1][B.sup.S] - . . . - [[PHI].sub.p][B.sup.PS]

Seasonal MA: [THETA]([B.sup.S]) = 1 + [[THETA].sub.t][B.sup.S] + . . . + [[THETA].sub.Q][B.supQS].

Analysis was carried out using Wessa online: Wessa P., (2017), (Partial) Autocorrelation Function (v1.0.15) in Free Statistics Software (v1.2.1), Office for Research Development and Education, URL http://www.wessa.net/rwasp_autocorrelation.wasp.

Mehak Singh (1), Manoj Pawar (2) [??], Antonio Chuh (3), Vijay Zawar (2)

(1) Department of Dermatology and Venerology, H. B. T. Medical College and Dr. R. N. Cooper Municipal General Hospital, Mumbai, Maharashtra, India. (2) Department of Dermatology, Dr. Vasantrao Pawar Medical College, Nashik, India. (3) Department of Family Medicine and Primary Care, University of Hong Kong and Queen Mary Hospital, Pokfulam, Hong Kong. [??]Corresponding author: manojpawar624@yahoo.com b) linear regression model.

Received: 18 August 2018 | Returned for modification: 6 December 2018 | Accepted: 25 December 2018

doi: 10.15570/actaapa.2019.4

Table 1 | Epidemiological data and its comparison with other studies. Study Location PR patients Harman et al. (1998) Eastern Anatolia, Turkey 399 Nanda et al. (1999) Kuwait 117 Tay et al. (1999) Singapore 368 Traore et al. (2001) Burkina Faso 36 Chuh et al. (2003) Hong Kong 41 Chuh et al. (2005) Minnesota, United States, Kuwait, 1,379 and Diyarbakir, Turkey Sharma et al. (2010) Uttar Pradesh, India 200 Ayanlowo et al. (2010) Lagos, Nigeria 427 Ganguly et al. (2013) Southern India 73 This study (2018) Northeast India 136 Study Male:female Seasonal variation Harman et al. (1998) 1:1.21 Peak during spring, autumn, and winter Nanda et al. (1999) 1:1.38 Not reported Tay et al. (1999) 1.19:1 No variation Traore et al. (2001) Not reported Not reported Chuh et al. (2003) 1:1.05 February, July, April Chuh et al. (2005) Not reported Clusters found but did not mention the seasons Sharma et al. (2010) 2:1 September to December Ayanlowo et al. (2010) 1:1.55 October, August, March Ganguly et al. (2013) Male preponderance No variation This study (2018) Female preponderance September to January PR = pityriasis rosea. Table 2 | Distribution of patients with pityriasis rosea by month, meteorological data, and Ns1Ag and/or IgM/IgG antibody positivity. Cases with Ns 1Ag and Average Average Month Patients temperature precipitation or IgG/IgM antibody ([degrees]C) (mm) positivity (only documented cases) January 14 17.5 12 3 February 7 19.5 16 0 March 1 23.3 60 0 April 0 26.0 141 0 May 5 26.8 278 1 June 3 28.1 315 0 July 2 28.9 313 2 August 1 29.0 261 0 September 25 28.6 181 8 October 17 26.2 100 9 November 36 22.5 15 11 Total 136 - - 38

Printer friendly Cite/link Email Feedback | |

Author: | Singh, Mehak; Pawar, Manoj; Chuh, Antonio; Zawar, Vijay |
---|---|

Publication: | Acta Dermatovenerologica Alpina, Pannonica et Adriatica |

Date: | Jan 1, 2019 |

Words: | 3444 |

Previous Article: | Combination of [CO.sub.2] laser therapy and curettage for sebaceous gland hyperplasia. |

Next Article: | Pulsed-dye laser versus intralesional Candida albicans antigen injection in treatment of genital warts. |

Topics: |