Study design and setting
This was a hospital–based case-control study carried out in the Bamenda Regional Hospital (BRH) between the months of September 2015 and August 2016. Bamenda is the capital of the Northwest region of Cameroon with an estimated population of 500,000 inhabitants. The BRH has a total bed capacity of about 600 beds and serves as the main referral hospital in the region, as well as a teaching hospital for the Faculty of Health Sciences, University of Bamenda. The study was conducted mainly at the surgical department of the hospital. The hospital does not have a microbiologic service and superficial swabs with gram staining are the main method of screening for germs from skin specimens.
Study participants and sampling
All consenting patients hospitalised for cellulitis of the lower limbs during the study period were recruited. For each case, two controls were selected and matched for age and sex. Those who presented with erysipelas (with a localised and well demarcated area of erythema, oedema, redness and pain), necrotising fasciitis of the lower limb, myositis, abscess, other variants of dermo-hypodermitis, or refused to sign an informed consent were excluded from the study.
Sample size calculation
The sample size was calculated using the formula for difference in proportions [12]:
$$ \mathrm{n}=\left(\mathrm{r}+1/\mathrm{r}\right)\ \left[\left(\mathrm{p}\right)\left(1\hbox{-} \mathrm{p}\right)\ {\left({\mathrm{Z}}_{\mathrm{B}}+{\mathrm{Z}}_{\upalpha 2}\right)}^2\right]/{\left({\mathrm{p}}_1\hbox{-} {\mathrm{p}}_2\right)}^2 $$
Where n is the sample size in the case group, r is the ratio of control to cases (r = 2), p is the average proportion exposed. Assuming that the desired power = 80%, ZB = 0.84 and Zα2 = 0.96 for a statistical significance at 5%. P1 is the proportion of cases exposed and p2 is the proportion of controls exposed to one of the risk factors of cellulitis – obesity; which is assumed to be 13% [13]. The minimum odds ratio required to detect an effect will be 3. Therefore, the number of cases required for the study will be 59 and the number of controls – 118.
Case definition
Cases were defined as patients admitted to the surgical unit who presented with a localised area of lower limb erythema, warmth, oedema and pain, associated with fever (temperature > 38 °C) and/or chills of sudden onset. The cases were hospitalised in the surgical ward after consultation by the surgeon without prior referral as the hospital neither has a dermatology unit nor a dermatologist. Patients who presented with leg ulcers were generally excluded. Only patients who presented with the above case definition and developed an inherent leg ulcer during inpatient monitoring were included in the study.
Our controls were subjects hospitalised for diseases other than necrotising fasciitis, myositis, abscess or other variants of dermo-hypodermitis. Controls were recruited mainly from the same surgical ward where the cases were found except in a few cases where they were obtained from the paediatrics, obstetrics and gynaecology wards.
Data collection
Data was collected using a pre-structured questionnaire and included: demographic characteristics (age, sex, marital status, occupational status, level of education), clinical profile (diseased systems for controls), general risk factors for cellulitis [histories of diabetes, hypertension and human immunodeficiency virus (HIV)], local risk factors of cellulitis [past history of surgery to the lower limb and history of trauma or disruption of skin barrier (recent trauma of the lower limb leading to break of skin with or without break in subcutaneous tissues)], social history (histories of alcohol consumption, tobacco smoking and use of other recreational drugs) and length of hospital stay. The participants were then examined clinically for presence of toe-web intertrigo (mild scaling to an exudative, macerated or erosive process of the toe web space) [14], their heights and weights were recorded after measurement. Samples were collected from the toe-web and/or an ulcer of the affected limb of participants with cellulitis and sent to the laboratory for gram staining.
Obesity was defined as a body mass index (BMI) of 30 kg/m2 and above [15]. We defined alcohol misuse as a consumption of more than 11.2 g/l (14 units) of alcohol per week. History of HIV was sub-categorised into; severe immunodeficiency (CD4+ count <200 cells/mm3) or not (CD4+ ≥200 cells/mm3) based on the most recent measurement within a 3 month period [16], and the duration after diagnosis at which a person living with HIV is at risk of soft tissue infection (≥ 4 years) or not (< 4 years) [17]. Also, the disease burden was assessed by the length of hospital stay, need for necrosectomy and eventual amputation.
Ethical considerations
This study was approved by the ethical review board of the Bamenda Regional Hospital and authorisation was granted by the Director of the said institution prior to the study. A signed informed consent was required from participants ≥18 years or from the guardians of participants <18 years after detailed explanation about the study in their language of choice: English, French or Pidgin (the local lingua franca), before they were enrolled into the study. All information collected from our study was treated as confidential.
Statistical analysis
All data were entered using Epi info version 7.0.8.3 and analysis was done with STATA version 12.1. Categorical variables were presented as frequencies and proportions. Continuous variables were presented as means and medians where applicable. Categorical variables were compared using a Chi square test and Odds ratios (OR) were used to assess the degree of associations. Variables with a p value <0.1 on univariate analysis were included in the multivariate logistic regression analysis model in a stepwise fashion and removed if the p value was >0.05, to identify risk factors for cellulitis. Statistical significance was considered at the 5% level for the multivariable analysis.
A sensitivity analysis was done excluding paediatric patients (age ≤ 18 years [18]) by repeating the above procedures.
Finally, we calculated the population-attributable risk (PAR) for factors which were associated with cellulitis to estimate the proportion of preventable disease if these risk factors were reduced or eliminated. Based on the assumption that the controls are representative of the general population in which cellulitis is an infrequent outcome, we calculated the PAR using the formula [19]:
$$ PAR=\frac{P\left( RR-1\right)}{P\left( RR-1\right)+1} x100\% $$
P = prevalence of the given risk factor in the control group and RR = Risk ratio.