If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
We report on an epidemic of a necrotic skin infection with a high attack rate
•
The cause of the epidemic remained unclear because of lack of diagnostic capacity
•
Relative humidity and heavy rain were associated with increased infection cases
Abstract
Introduction
In 2016–18, the Democratic Republic of São Tomé and Príncipe suffered a necrotic skin infection epidemic.
Methods
A surveillance system was established after increased hospitalisations for this infection. Microbiology results were available for samples analysed in December 2016 and March 2017 using whole genome sequencing and metagenomics. Negative binomial regression was used to study the association of weather conditions with monthly case counts in a time-series analysis.
Results
From October 2016 to October 2018, the epidemic cumulative attack rate was 1.5%. The first peak lasted 5 months, accounting for one-third of total cases. We could not conclusively identify the aetiological agent(s) due to the country's lack of microbiology capacity. Increased relative humidity was associated with increased monthly cases (incidence rate ratio (IRR) 1.05, 95% CI 1.02–1.09), and higher precipitation in the previous month with a higher number of cases in the following month (months with 0–49 mm rainfall compared with months with 50–149 mm and ≥150 mm: IRR 1.44, 95 % CI 1.13–1.78 and 1.50, 95% CI 1.12–1.99, respectively).
Discussion
This epidemic was favoured by increased relative humidity and precipitation, potentially contributing to community-based transmission of ubiquitous bacterial strains superinfecting skin wounds.
Funding
World Health Organization Regional Office for Africa, Ministry of Health
Association of athlete's foot with cellulitis of the lower extremities: diagnostic value of bacterial cultures of ipsilateral interdigital space samples.
). Cellulitis can progress to a necrotising skin infection that can affect soft tissues. Necrotising fasciitis is a severe and rapidly progressing soft tissue polymicrobial infection caused by the same bacteria as cellulitis, such as beta-haemolytic streptococci, and Gram-negative bacteria found in water, such as Vibrio spp. It causes substantial morbidity, including substantial rates of skin graft, and up to 30% mortality (
The Democratic Republic of São Tomé and Príncipe is a Portuguese-speaking archipelago of volcanic origin located in the Gulf of Guinea, close to the equator. The country consists of 2 islands (São Tomé and Príncipe) and numerous islets, has approximately 200 000 inhabitants and a surface area of 1001 km2. Cellulitis cases have been described in the country in the last few years. However, from September 2016, hospital beds were being quickly filled with patients with a severe necrotic skin infection of unknown aetiology requiring lengthy hospital stays. Unlike necrotising fasciitis, this infection did not affect the fascia, and it was therefore referred to as ‘necrotising cellulitis’, although this is not a recognised clinical term. A surveillance system and an epidemic investigation team were quickly established. Here, we describe the epidemiology of this event, quantify its extent and investigate the aetiology and risk factors for the disease, including climatic factors.
Methods
Study design and methodology
After notification to the Ministry of Health of an increase in the number of cellulitis hospitalisations, disease surveillance was quickly established in the 6 districts of São Tomé and the autonomous region of Príncipe. Suspected cases were defined as any person who presented with an inflammatory process circumscribed to a region of the body, painful or not, with or without injury, oedema or lumps and/or serous or purulent secretion; probable cases were defined as any suspected case with extension of the inflammatory process or any person with epidermal detachment or cutaneous necrosis.
Probable necrotising cellulitis cases were recruited from the regional hospital in Príncipe or the Ayres de Menezes hospital, or other health unit with hospitalisation capacity, in São Tomé. In São Tomé, controls were matched by age (intervals of 2 years for aged <18 years, intervals of 5 years for aged ≥18 years), sex and week of hospital admission. Controls were patients hospitalised for any disease other than cellulitis, but no microbiological investigation was available for cases or controls. Due to the limited number of cases available in Príncipe, no formal sample size calculation was made, and a case-control ratio of 1:4 was used to improve power. For São Tomé, using an expected odds ratio of 3 for the estimated least frequent independent variable in the population, a power of 80%, a CI of 5% and a case-control ratio of 1:2, the sample size calculation predicted the need of 44 cases and 88 controls. Exclusion criteria for cases and controls were being aged <5 years, having chronic conditions (e.g., diabetes, alcoholism) or having other skin lesions diagnosed as diabetic foot, ulcer, or other superinfected chronic lesions.
Laboratory procedures
Microbiological investigation was carried out on a limited number of samples in December 2016 and March 2017. In March 2017, a field laboratory was set up with quantitative-polymerase chain reaction (qPCR) tests for the following organisms based on the potential causes of cellulitis and necrotising fasciitis: Staphylococcus aureus, Streptococcus pyogenes, Mycobacterium ulcerans and marinum, Pseudomonas aeruginosa, Aspergillus spp., and Vibrio spp. The following additional qPCR tests were used as positive and negative controls: Brucella sp., Bacteroides sp., Rickettsia sp., and Legionella sp. In addition to the above tests, the laboratory was able to undertake the following untargeted testing approaches on the MinIon sequencing platform: i) full length (V1-6) and partial (V1-3) 16S—to identify bacterial species without culture; ii) fungal internal transcribed spacer for zygomycetes and pan-fungal primers; iii) full bacterial genome sequencing of cultured isolates (Supplementary Figure 1). In addition to the mobile capabilities, testing on key samples was selectively performed in the UK using a next-generation sequencing platform (Illumina HiSeq 2500) and metagenomics.
Statistical analysis
Logistic regression was used to investigate risk factors associated with the disease in the case-control studies and to hospitalisation using the database of notified cases. The following variables were studied in the case-control studies: age, sex, drinking alcohol every day, having recurring health problems, having had a severe skin lesion in the past. In addition, the following were noted if they occurred in the 2 weeks prior to hospital admission: having had skin lesions/injuries, having visited a water source, having been at the beach or in the sea, having lived close to a river, having had contact with stagnating water, preparing fish for selling or eating, having fished, having had contact with a person with a severe skin lesion, farming pigs or having been scratched/injured/bitten by a pig or other animal, having taken drugs, having travelled or having taken drugs.
We retrieved information by month on temperature, relative humidity, rainfall and number of rainy days from September 2013 until October 2018 registered by the only weather station in São Tomé and Príncipe, located close to São Tomé international airport.
Seasonal and temporal trends in case occurrence were evaluated using an approach similar to that used by Jensen et al. (
) with construction of a negative binomial regression model incorporating month and year and accounting for seasonality trends. For all models, variables, unless forced, were retained in the final model if they had P<0.1. Data were entered in Epi Info software (version 7), and statistical analyses were carried out using R (version 3.5.4) or STATA version 14.0.
Results
Clinical presentation of suspected and probable necrotising cellulitis and treatment of choice
Reports from international infectious disease experts that visited a limited number of cases (n=35) meeting the definition of suspected and probable necrotising cellulitis between October 2016 and February 2017 underlined the main clinical features. A skin lesion, deep or superficial, was noticed in most cases. The start of the disease was most often severe, with high fever (>38°C), inflammatory oedema of variable extension and depth and at times presence of lumps, either evolving to desquamation (suspected cases, Figure 1, Panel A) or evolving to superficial and localised necrotic plaques or even large and diffused necrotic plaques in 4–10 days but without involvement of the fascia, and without heavily affecting the general state of the patient (probable cases, Figure 1, Panel B). Most new cases did not have any epidemiological link to previous cases and there was no evidence of person-to-person transmission. Localisation was to the lower limb in almost all cases (Figure 1). The delay between the initial wound and the symptoms varied from 48 hours to 2 months. The interval between onset of symptoms and admission varied from 8 days to 3 months.
Figure 1Typical clinical presentation of suspected (A) and probable cases of necrotic skin infection (B).
Recommended first-line treatment during the acute phase for mild cases (for those not allergic to penicillin) was Clavamox 875 mg or Amoxicillin 1 g or Ampicillin 500 mg; for more severe cases, Clavamox 1g or Ampicillin plus Clindamycin 600 mg. Second-line treatment during the acute phase was Floxapen 1g + Ceftriaxone 1g. In the late phase (ulcer phase), Clavamox 1g was recommended.
Descriptive epidemiology
From October 2016, the Ministry of Health began notifying of suspected and clinically probable necrotizing cellulitis cases. Until February 2017, the median number of cases per week was 61 (interquartile range (IQR), 40–79), with a notification peak at week 52 (108 cases). The second wave of the epidemic, from March 2017 to October 2018, registered a median number of cases per week of 18 (IQR 15–25). From November 2019 (until February 2019), the median number of cases decreased to 6 per week (IQR 4–8); this roughly corresponded to the background hospitalization rate for cellulitis that a retrospective analysis estimated at approximately 20 per month for the main hospital in São Tomé (Figure 2). Therefore, we analysed cases, divided into suspected and probable, from the first and the second wave of the epidemic only, excluding cases notified after week 44 of 2018.
Figure 2Epidemic curve of the suspected and probable cases of necrotic skin infection in Sao Tome, 2016-18 After October 2018 (*), the monthly incidence of hospitalisations returned to background levels.
We had demographic data and information on hospitalisation for 2441/2841 (85.9%) of notified cases. Of the cases, 52.9% were male (1501/2839), 21% were children (<15 years, 605/2836), 42.5% were young adults (15–44 years,1206/2836), 23.9% were aged 45–64 years (679/2836) and 12.2% ≥65 years (346/2836). Of the 2441 cases with information on hospitalization, 50% were hospitalized for at least 24 hours (n=1176). Cases came from all 7 districts (6 in São Tomé, 1 in Príncipe). Approximately one-third (35.3%) of cases were notified during the first wave of the epidemic, with the rest notified during the second wave (64.7%).
Age distribution did not differ between suspected and probable cases, while males were more common among suspected cases (53.8%) than among probable cases (46.9%). The hospitalisation rate was higher among probable cases than suspected cases (63.0% vs 45.9%, Table 1).
Table 1Characteristics of suspected and probable necrotizing cellulitis cases (N=2841) Missing data for age=5; sex=2; district=1; consultation period=0; hospitalization=400 (14%).
By the end of the epidemic, the country's cumulative incidence was 15.5 per 1000 inhabitants. The incidence was highest in Caué (28.5 per 1000 inhabitants) and Cantagalo (24.2 per 1000) and lowest in Agua Grande and Príncipe (10.2 and 12.0 per 1000, respectively, Figure 3).
Figure 3Cumulative incidence of suspected and probable cases of necrotic skin infection per 1000 inhabitants, by health district
Of the total notified cases (2441), only 28 had benefited from surgical interventions, of which 19 were skin graft surgeries. No single death directly caused by necrotising cellulitis was reported during the epidemic.
Laboratory investigation
Microbial investigation was largely unavailable throughout the epidemic; however, international laboratory investigation teams travelled to São Tomé and Príncipe and took a limited number of samples. Infection with Mycobacterium ulcerans was excluded, as 68 samples from October 2016 were negative by Ziehl/Nielsen stain and/or PCR. Sampling in April 2017 (46 samples from 21 patients) identified polymicrobial infections, with S. pyogenes in 12/21 cases, S. aureus in 15/21 cases and co-infections with both in half of the cases (11/21), as well as other opportunistic infections in most cases. Corynebacterium diphtheriae was not isolated (Gram-negative was not attempted for these samples); however, it was identified by metagenomics in 9/12 tested samples (Figure 4). The strains identified by shotgun metagenomics are reported in Supplementary Table 1.
Figure 4Sample results summary matrix The results below summarise the qPCR, 16S, full genome and metagenomics analysis. Results are binary, i.e., they simply display the presence or absence of an organism. Metagenomics included robust negative controls to identify possible environmental contamination, namely 3 negative control extractions, prepared alongside the patient samples. (p) Principe; (s) Sao Tome; (a) 16S positive only; (b) Corynebacterium sp. other than diphtheria detected by partial 16S only; Patient IDs underlined were subjected to metagenomics.
Risk factors for necrotising cellulitis and hospitalisation
The first case-control study was conducted from 1 October 2016 to 28 February 2018 in Sao Tomé and from 1 January 2017 to 15 April 2018 in Príncipe. Forty-five cases and 90 controls were recruited in São Tomé Island, and 27 cases and 108 controls in Príncipe Island. The lesion was in the lower leg in 100% of cases with a known location (16/16 in Príncipe).
Presence of an injury in the 2 weeks before disease onset (adjusted odds ratio (aOR) 7.7, 95% CI 3.2–18.6 in São Tomé; aOR 8.3, 95% CI 2.8–23.2 in Príncipe) and living close to a river (aOR 2.59, 95% CI 1.01–6.87 in São Tomé only) were identified as risk factors while having a recurrent health problem was identified as a protective factor (aOR 0.38, 95% CI 0.16–0.90 in São Tomé; aOR 0.31, 95% CI 0.10–0.88 in Príncipe). In addition, increased age was identified as a risk factor (in Príncipe only, P=0.01). Having had contact with a person with a severe skin lesion was not found to be a risk factor.
Multivariable analysis using the database of all suspected and probable necrotic skin infections from October 2016 until October 2018 to investigate risk factors for hospitalisation found a lower risk of hospitalisation for suspected cases, as compared with probable cases (aOR 0.50, 95% CI 0.38–0.67), no difference in risk between males and females (P=0.53), and an increased risk of hospitalisation in older age groups (P<0.001, Supplementary Table 2).
Environmental risk factors for increase in the monthly number of cases
Because the increase in the number of cases appeared to correlate with the amount of rainfall in São Tomé, we investigated if climatic factors had a potential role in this epidemic.
From São Tomé’s only weather station, we collected minimum, maximum and average temperatures, precipitation (mm), number of rainy days and relative humidity by month for the last 5 rainy seasons, including the 2 associated with the necrotising cellulitis epidemic (Supplementary Table 3). São Tomé is crossed by the equator and the mean temperature range for the period under study was 23.8–27.0°C (minimum 18.5°C, maximum 32.7°C).
From October to February, the period corresponding to the first epidemic wave, the average monthly rainfall was 273.6 mm (range 142.6–435.0 mm) in the pre-epidemic seasons 2013/14 to 2015/16. The same rainy season period in 2016/17 registered a substantial increase in the amount of rainfall (640.5 mm, 134% increase compared with the mean rainfall for the 2013/14 to 2015/16 seasons; 160% increase compared with the previous rainy season, 2015/16). From October 2017 to February 2018, 438.3 mm rainfall was registered, i.e., a 60% increase compared with the mean rainfall for the seasons 2013/14 to 2015/16, or a 50% decrease compared with the previous rainy season, 2016/17.
We studied time trends, seasonality and weather effects using aggregate monthly cases. The overall goodness of fit of the time trend model was good (Chi2 22.0; P<0.001). Observed and predicted case counts are presented in Supplementary Figure 2. We found that each month registered an average 2% decrease in the number of cases (Table 2). However, a 1% increase in the relative humidity resulted in a 5% increase in the monthly number of cases. Furthermore, a mean maximum temperature below the median (i.e., 30.9°C) in the previous month resulted in double the number of notifications in the following month (Table 2). Similarly, compared with the lowest amount of monthly rainfall (0–49 mm), the mid amount (50–149 mm) and highest amount (≥150 mm) were associated with a 1-month lagged higher number of cases (IRR 1.41, 95% CI 1.13–1.77 and IRR 1.50, 95% CI 1.12–1.99, respectively) (Table 2).
Table 2Incidence rate ratio (IRR) estimates from the time-series model taking environmental factors into account. LRT: Likelihood ratio test.
We described a large epidemic of a necrotic skin infection with the first peak of infections overlapping the first peak of the rainy season in 2016/17, characterized by unusually abundant rains.
Hygienic conditions in São Tomé and Príncipe are limited and skin infections are common in this type of setting. Poverty is widespread, with over 50% of the population living under the poverty threshold and 15% in extreme poverty (as of 2001), and a fast-growing population in the last few years (
). The pathogens identified in this study were a range of virulent, multi-drug resistant bacterial strains. Characterized strains and antimicrobial resistance patterns differed in whole-genome sequencing analysis, indicating enhanced transmission of virulent strains rather than the epidemic being caused by a single strain. Among other tetracycline-resistant S. pyogenes, strains included methicillin-sensitive Panton-Valentine leucocidin (PVL) toxin positive S. aureus, methicillin-resistant S. aureus, and Corynebacterium diphtheriae, identified by metagenomics only in three-quarters of the samples that were tested. The epidemic captured here could potentially have been 2 overlapping outbreaks of skin and soft tissue infection over a ∼24-month period, with a "first wave" potentially being driven by cutaneous diphtheria (the usual manifestations of which fit the clinical case description of the reported condition quite neatly) and the "second wave" representing a potentially climate-associated increased incidence of ubiquitous bacterial strains superinfecting skin wounds. However, vaccination against diphtheria started in 1977 in Sao Tomé, and vaccine coverage for the third dose of the pentavalent vaccine (DTP-HepB-Hib) in 2016 was 93% (95% CI 88%–96%) in 12–23 month-old children, with a dropout rate of <10% in all districts (
Despite this microbiological evidence, because clinical and epidemiological data of these patients were not well characterised, we cannot exclude that 1 unidentified strain was the cause of the epidemic with increased pathogenesis through human skin commensals as described for S. aureus (
Results from the case-control studies identified the presence of a wound and/or intertrigo in the 2 weeks prior to the disease used by ubiquitous bacteria as an entry site, a known risk factor for diagnosis of cellulitis with common superinfection with S. pyogenes. (
). The protective effect of having a recurrent health problem was most likely because control subjects were more likely to have a recurrent health problem and/or health-seeking behaviour than the general population. Rather than patients admitted to the hospital for other medical conditions, it would have been more appropriate to select geographically, age and sex-matched community-based individuals as controls
Environmental investigation confirmed the hypothesis that favourable climatic conditions for bacterial survival and growth played a role in the increased transmission of bacterial strains, including virulent ones, in the community. An increase in relative humidity was associated with a larger number of cases, suggesting that it may have helped the infection to occur, often in the presence of a skin lesion, most often located in the legs. Rainfall in São Tomé and Príncipe can be sudden and heavy, and these factors can facilitate the spread of bacteria in the environment. People from São Tomé and Príncipe often walk barefoot or in flip flops, and rivers are used for several activities, including washing clothes. Seasonality of cellulitis, associated with warmer months, has been described (
). Total rainfall was also recently associated with cellulitis incidence in Taiwan; weather conditions such as temperature and humidity may interfere with the skin barrier and promote bacterial growth (
In São Tomé and Príncipe, the rainy season is slightly warmer than the dry season, but the temperature differences between seasons are limited, as the country lies on the equator. Islands can be more susceptible to climate change when they host fragile ecosystems. Precipitation in São Tomé and Príncipe has decreased an average of 1.7 mm per year from 1950 until 2000 and temperature has increased 1.2°C from 1978 until 2000 (
Though hospitals may have played a role in the spread of virulent strains and possibly in the severity of the disease, the infection causing first symptoms mostly happened in the community. Unfortunately, data on the follow-up of the infection course was suboptimal for most cases as they were not systematically requested by the surveillance system. Moreover, the case definition for suspected patients was sensitive and included skin inflammation of any cause (trauma, burn, wound), while patients with lymphangitis of any infectious cause, ulcer linked to varicosity, or ulcer related to underlying disease such as diabetes, were rarely properly diagnosed and therefore excluded from the counts. Therefore, an overestimation of the number of cases of necrotising cellulitis is likely to have occurred. According to the international medical experts who reviewed admitted patients to the São Tomé hospital (n=20) during the epidemic peak, only 60% were actual necrotising infections. Interestingly, cellulitis was rarely found in soft tissue, and no cases of necrotising fasciitis were reported, which is in line with the finding of zero mortality that would be very unusual for necrotising fasciitis.
The lack of microbiological capacity was amongst the biggest challenges of the response to this epidemic. Human expertise was present, but reagents were unavailable and basic microbiology was seldom carried out. Furthermore, the laboratory investigation of the subset of samples analysed abroad was limited by the lack of medical and epidemiological information attached to the samples.
Similarly, despite strict guidelines for treating cases, the antibiotic treatment of choice was irregularly prescribed because it was often unavailable. Drug susceptibility testing was unavailable for most cases, and response to antibiotic treatment was not formally documented, with very few exceptions. It is also surprising that despite the severity and the extent of the epidemic, no deaths were reported. Although deaths may have been underestimated, clinicians in Sao Tomé and Principe agreed that the great majority of patients recovered, even if morbidity and disability-adjusted life years count would have been high if quantified since surgical skin treatment and skin graft were not available, and many patients are likely to have had sequelae.
Control measures included improved hygiene in the hospitals and messages from the Ministry of Health in different media, such as newspapers and television, to improve personal hygiene in the community.
Changes in environmental conditions are likely to have played a role in this large-scale event, with relative humidity favouring infection and abundant rainfalls likely to have had a negative effect on general hygienic conditions on the islands. Though it is possible that the same highly virulent strain caused most of the severe cases, we could not identify the strain due to a lack of microbiological capacity. This experience calls on investing in diagnostics capacity, especially now that climate change has started to increase the burden of infectious diseases in fragile environments such as Sub-Saharan Africa and island countries.
Disclaimer
The authors alone are responsible for the views expressed in this article and they do not necessarily represent the views, decisions or policies of the institutions with which they are affiliated.
Author contributions
LS, AS, IB, EA, BCS, ABD, CADC, FS, MTP, MC and VSG performed field data collection. VL, JQ and LM performed laboratory investigations. LS and JL collected the climatic data and performed the time-series analysis. LS, AP, MHD and ISF performed data analysis and interpretation. LS wrote the manuscript and all authors revised the manuscript and approved its final version.
Acknowledgements
We acknowledge the participants to the study for their collaboration. We also thank Timothée Dub for stimulating discussions.
Ethics approval
The study followed international guidelines for ethics in research, such as the Universal Declaration of Bioethics and Human Rights and the Helsinki Declaration and was in agreement with São Tomé and Príncipe legislation. Informed consent was signed by every participant of the case-control studies, which were coordinated by the Ministry of Health with technical support from the World Health Organization.
Funding source
The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Conflicts of interest
None declared
Transparency declaration
This article is published as part of a supplement titled, “Field Epidemiology: The Complex Science Behind Battling Acute Health Threats,” which was supported by Cooperative Agreement number NU2HGH000044, managed by TEPHINET (a program of The Task Force for Global Health) and funded by the Centers for Disease Control and Prevention (CDC). Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention, the Department of Health and Human Services, The Task Force for Global Health, Inc., or TEPHINET.
Association of athlete's foot with cellulitis of the lower extremities: diagnostic value of bacterial cultures of ipsilateral interdigital space samples.
row percentages
00541-5&id=gr1.jpg){kind=link}
00541-5&id=gr2.jpg){kind=link}
00541-5&id=gr3.jpg){kind=link}
00541-5&id=gr4.jpg){kind=link}