Viral aetiology, clinical features and antibiotic use in children under five years old in The Gambia presenting with influenza-like illness

Abstract

Background

Knowledge regarding the prevalence, clinical features and aetiology of paediatric influenza-like illness (ILI) remains limited in African settings. Furthermore, it is likely that many children presenting with ILI receive antibiotics unnecessarily. More data are required to develop antimicrobial stewardship practice and guide effective vaccine strategies. We undertook a one year prospective study of ILI in The Gambia.

Methods

Children under five years of age presenting with ILI were recruited. Clinical and antibiotic prescribing data were collected. Nasopharyngeal swabs (NPS) were collected and analysed for 12 respiratory viruses using a multiplex PCR.

Results

From a total of 735 ILI episodes, 530 (72.1%) NPS were positive for at least one virus. Of these, 36.7% were positive for rhinovirus, 14.7% for respiratory syncytial virus (RSV), 8.4% for influenza and 7.2% for human metapneumovirus (HMPV). Compared to children under six months old, influenza was more common in 6 to 23 month old children (odd ratio (OR) 5.68, 95% confidence interval (CI) 1.72-18.76, p=0.004). RSV and HMPV were associated with low peripheral oxygen saturations (OR 2.13, 95% CI 1.23-3.69, p=0.007 and OR 2.44, 95% CI 1.13-5.27, p=0.023 respectively). 78.3% of all ILI cases were prescribed antibiotics.

Conclusions

A broad range of viruses are responsible for paediatric ILI in The Gambia. Refined treatment guidelines, improved diagnostic capacity and vaccines to prevent respiratory viruses will all play a role in reducing antimicrobial use for these cases.

Introduction

Respiratory viruses such as influenza and respiratory syncytial virus (RSV) are major contributors to morbidity and mortality in children under five years of age globally^{[1, 2]}. In 2016, it was estimated that there were 68 million episodes of lower respiratory infections causing a total of five million hospital admissions^[3]. A recent systematic review estimated that 90 million cases of influenza occur each year in children under five^[4]. In addition, 99% of RSV-related deaths globally occur in low and middle-income countries, making such infections a high priority for this setting^[5].

In temperate regions, peaks of influenza and RSV infection are seen in the winter months, whereas in tropical countries, seasonal patterns are less clear^[6,7]. Thirteen countries in sub-Saharan Africa were found to have one annual influenza peak occurring at varying times, five countries had two peaks and four countries showed year round activity^[8]. The seasonality of RSV in Africa is even less clear, although some studies suggest that RSV-related illnesses occur predominately during times of maximal rainfall^[9]. Most studies of influenza-like illness (ILI) in Africa conducted to date have focused on the prevalence of influenza. However, generating RSV prevalence data is also increasingly important, as several vaccines are now in late-stage development^[10–13]. Very few data are available for other respiratory viruses such as human metapneumovirus (HMPV), parainfluenza viruses, adenovirus and coronavirus as causes of ILI in African settings^[14,15]. Diagnostic tests for respiratory viruses are rarely available in low-resource settings, hence unnecessary antimicrobial use (AMU) is frequent. As a consequence, antimicrobial resistance is an increasing problem in Africa, with West Africa showing one of the highest increases in antibiotic consumption between 2000 and 2010^[16].

We conducted a one year prospective study in children under five years of age in The Gambia presenting with ILI, to determine the viral aetiology, seasonality, clinical features and associated AMU.

Materials and methods

Study design and setting

Between 19 March 2018 and 22 March 2019, a prospective study was conducted within the clinical services department (CSD) of the Medical Research Council Unit in The Gambia - part of the London School of Hygiene and Tropical Medicine. The Unit provides primary health care facilities to the local population as well as being an established research centre. The Gambia is a small country in West Africa with a population of approximately 2.2 million and a rainy season between June and October.

All children under five years of age presenting with an ILI to the outpatients department for primary care were identified by study nurses. Based on the World Health Organization (WHO) criteria, ILI was defined as an acute respiratory infection with a history of fever and a cough or sore throat, and with an onset of symptoms within the past 10 days ^[17]. Parents or guardians of children were invited to take part in the study and written informed consent was obtained. Demographic data were collected, along with a history and clinical examination conducted by a trained study nurse. A nasopharyngeal flocked swab (NPS) was collected in Universal Transport Medium (UTM, Copan, USA), which was frozen at -70°c on the same day. Study participants who re-presented with an ILI more than 14 days from their initial presentation were considered to have a new illness and additional data and NPS collected. Nursing staff provided antibiotics according to normal practice in The Gambia. Further investigations and treatment decisions, as well as need for hospitalisation were taken by a doctor based on a request by nursing staff. The study was approved by The Gambia Government/MRC Joint Ethics Committee (SCC1853).

Detection of respiratory viruses

Total nucleic acid was extracted from 180 µL of UTM using the QIAamp Cador Pathogen mini kit (Indical Bioscience Gmbh, Germany) as per manufacturing instructions. All samples were spiked with Equine arteritis virus (EAV) prior to extraction as an RNA internal extraction control. A Streptococcus pneumoniae strain was extracted with each batch as a DNA extraction control. A previously described multiplex reverse transcriptase polymerase chain reaction (RT-PCR)^[18] was performed to detect the following respiratory viruses: influenza A, influenza B, RSV A, RSV B, parainfluenza viruses 1 to 4, HMPV, adenovirus, endemic coronaviruses (229E, OC43, NL63, all detected with the same TaqMan® dye) and human rhinovirus. Briefly, reactions were performed for 15 min at 50°C, 95°C for 2 min and 40 cycles at 95°C for 8 seconds and 60°C for 34 seconds using the Bio-Rad CFX96 Real Time PCR System. Positive controls for all viruses were included on plasmids (Eurofins Genomics, Germany) and amplified with each RT-PCR run. Results from assays performed as part of this study were not available in real-time and treating nurses and physicians were not aware of any viral diagnoses obtained.

Statistical analysis

All summary statistics and analyses were generated using Stata 14.2 (StataCorp. 2015). Univariable logistic regression was performed to identify associations between clinical features and the presence of respiratory viruses. Multivariable analysis was done starting with a model with all predictors used in the univariable analysis. A final model was derived by removing variables with p-value >0.2 in the multivariable model in a stepwise procedure. Two-sample Z tests for proportions were done to compare proportions of males to females in children presenting with ILI, compared with all presentations in children under five years during the same period.

Results

A total of 805 children met the definition for ILI during the 12 month study period. Informed consent for study participation was provided in 735 cases (91.3%). The number of children whose parents declined to participate in the study were evenly distributed throughout the year (Supplementary Table 1). Of those included in the study, 627 children had one ILI episode, 94 children had two episodes, 13 children had three episodes and one child attended with an ILI on four separate occasions. (Figure 1).

Figure 1.

A flow diagram showing sample recruitment, total number of children consented, total number of influenza-like illness (ILI) episodes, cases of repeat attendance for ILI and the number of viruses detected.

Infants under six months of age accounted for 136 (18.5%) ILI episodes, children aged six to 23 months accounted for 363 (49.3%) episodes and children aged 24 to 59 months accounted for 231 (31.4%) episodes (Table 1). In all, 59.0% of participants were male and 41.0% female. This proportion was not significantly different to all children under five years of age attending outpatients during the same period (1830/3434 (53.3% male; p=0.53).

Table 1. Proportion of cases that satisfied the influenza-like illness recruitment criteria during each month of the study, stratified by sex and age group.

Variable		N=735
Period	March 2018	29 (3.9%)
	April 2018	53 (7.2%)
	May 2018	90 (12.2%)
	June 2018	72 (9.8%)
	July 2018	95 (12.9%)
	August 2018	68 (9.3%)
	September 2018	59 (8.0%)
	October 2018	75 (10.2%)
	November 2018	81 (11.0%)
	December 2018	46 (6.3%)
	January 2019	45 (6.1%)
	February 2019	22 (3.0%)
Sex*	Male	434 (59.2%)
	Female	299 (40.8%)
Age	< 6 months	136 (18.6%)
	6-23 months	363 (49.7%)
	24-59 months	231 (31.6%)

^*NB: Two missing data points

A respiratory virus was detected in 530/735 (72.1%) NPS samples. Of these, 370 (69.8%) were positive for only one virus, 125 (23.6%) for two viruses, 34 (6.4%) were positive for three viruses and one sample tested positive for four viruses (Figure 1). Rhinovirus was the most common virus, detected in 36.7% of ILI cases and present throughout the year (Figure 2). Influenza A was found in 7.0% of cases and influenza B in 1.4%. Influenza A was seen predominately between August and November and coinciding with the rainy season, whilst influenza B was mainly seen in November. RSV was detected in 14.7% of ILI cases, with the majority being RSV A (13.3%) and a small number of RSV B positive samples (1.4%). While the months of May to November accounted for most RSV cases, smaller numbers were seen throughout the year. HMPV was identified in 7.2% of ILI cases, mostly between May and August. Parainfluenza type 1 was identified in 5.7% of cases and parainfluenza type 3 in 7.5%, predominately between March and June, and July and November respectively. Coronavirus was found in 7.8% cases and was seen in low number throughout the year.

Figure 2. Number of presentations positive for each virus by month and year.

Distribution of different viruses detected by month. The length of bar chart and colour indicates the number and type of viruses detected for each month. The partial month data from March 2018 and March 2019 were combined together for the purpose of this figure.

Based on multivariable logistic regression, compared to children under six months old, influenza virus-associated ILI was detected more often in 6 to 23 month and 24 to 59 month old children (odds ratio (OR) 5.68, 95% confidence interval (CI) 1.72-18.76) and OR 2.99, 95% CI 0.84-10.68 respectively; Table 2). RSV was less common in children aged 24 to 59 months compared to those less than six months old (OR 0.53 95% CI 0.29-0.98). Rhinovirus was found less commonly in children aged 6 to 23 months compared to those less than six months old (OR 0.57, 95% CI 0.38-0.86).

Table 2. Multilevel multivariable logistic regression analysis showing factors associated with different viruses.

	influenza A/B		RSV A/B		rhinovirus		human metapneumovirus		coronavirus		adenovirus
Factor	OR (95% Cl)	p-value	OR (95% Cl)	p-value	OR (95% Cl)	p-value	OR (95% Cl)	p-value	OR (95% Cl)	p-value	OR (95% Cl)	p-value
Age
< 6 months	Ref		Ref		Ref				Ref
6-23 months	5.68 (1.72, 18.76)	0.004	0.74 (0.44, 1.25)	0.263	0.57 (0.38, 0.86)	0.008			0.49 (0.25, 0.96)	0.038
24-59 months	2.99 (0.84, 10.68)	0.091	0.53 (0.29, 0.98)	0.043	0.78 (0.50, 1.20)	0.252			0.53 (0.25, 1.10)	0.089
Rhinnorhoea
No			Ref
Yes			3.02 (1.06, 8.57)	0.038
Chest indrawing
No	Ref		Ref				Ref		Ref		Ref
Yes	1.57 (0.89, 2.76)	0.116	2.05 (1.33, 3.15)	0.001			0.59 (0.27, 1.30)	0.190	0.58 (0.29, 1.12)	0.106	0.63 (0.32, 1.25)	0.188
Chest wheeze
No					Ref
Yes					1.75 (1.23, 2.49)	0.002
Chest crepitations
No							Ref		Ref		Ref
Yes							0.41 (0.22, 0.76)	0.005	1.70 (0.82, 3.55)	0.156	2.61 (1.06, 6.39)	0.036
Temp>38
No	Ref						Ref				Ref
Yes	3.13(1.46, 6.68)	0.003					2.15 (0.85, 5.47)	0.108			2.44 (0.96, 6.17)	0.06
Oxygen saturation
Normal			Ref				Ref
Low			2.13(1.23, 3.69)	0.007			2.44 (1.13, 5.27)	0.023

Multivariable regression models to identify clinical features associated with each virus using backward selection and including variables with a p-value of >0.2 in multivariable models. Only data from variables included in the multivariable models are shown for each virus.

Abbreviations: OR – odd ratio; CI – Confidence interval; p-value – probability value; Ref – Reference

Detection of influenza and adenovirus was associated with fever at presentation of ≥38°C (OR 3.13, 95% CI 1.46-6.68 and OR 2.44, 95% CI 0.96-6.17 respectively; Table 2). RSV was associated with rhinorrhea (OR 3.02, 95% CI 1.06-8.57), chest in-drawing (OR 2.05, 95% CI 1.33 to 3.15) and low peripheral oxygen saturations (OR 2.13, 95% CI 1.23-3.69). HMPV was also associated with low oxygen saturations (OR 2.44, 95% CI 1.13-5.27). The presence of rhinovirus was associated with wheeze on auscultation of the chest (OR 1.75, 95% CI 1.23-2.49). Of the total number of 735 enrolled in the study, 8 (1.1%) fulfilled the case definition for severe acute respiratory illness (SARI)^[17]. Of these, two cases were associated with RSV A alone, one each with rhinovirus, HMPV, adenovirus, and parainfluenza 3. One case was associated with multiple viruses (RSV A, HMPV, rhinovirus) and one had no virus identified on RT-PCR.

Details on whether antibiotics were prescribed were available for 708 ILI cases, of which 554 (78.2%) consultations received antibiotics (83.6% amoxicillin, 6.0% co-trimoxazole, 5.0% erythromycin, 1.3% cloxacillin, 0.5% chloramphenicol, 3.6% no name recorded). Children from whom a respiratory virus was subsequently detected whereas likely to have had antibiotics prescribed as those from whom the NPS was negative (79.4% and 75.4% respectively).

Discussion

We describe the prevalence of respiratory viruses and the associated clinical features in children under five years of age presenting with ILI to a primary healthcare setting in The Gambia. The high percentage of cases in which a virus was detected (72.1%) in our study is consistent with similar studies in Madagascar (75.1%) and Cameroon (65.1%)^[10-13,19] but somewhat higher than found in Senegal (58.5%)^[20].These differences could be due to methodological factors such as assay sensitivity and viruses targeted, as well as due to true differences in the epidemiology of ILI.

Although rhinovirus was the most prevalent virus detected, its role in ILI disease aetiology is less clear as increasing data suggest the presence of rhinovirus is also high in asymptomatic children^[21,22]. In contrast to rhinovirus, influenza and RSV are usually associated with clinical symptoms when detected^[23].

Influenza viruses were detected in 8.5% cases, which was similar to studies in Uganda (10.4%) and Gabon (11.4%)^[24,25], but relatively lower than reports from other African countries including Togo (24.7%), Senegal (25.3%) and Cameroon (18.4%) ^{[19, 26, 27]}. As the intensity of influenza virus circulation can vary significantly from year to year, ongoing ILI surveillance across sites is required to confidently conclude a difference in disease burden between these countries. Differences in study design such as using a history of fever rather than documented fever may also have resulted in finding a lower percentage of influenza as a cause of ILI than in other studies. Influenza virus was less commonly identified in children under 6 months of age, which may be explained by protection from maternal antibodies^[37,38].

Compared to influenza, our study demonstrated a higher number of RSV-associated ILI cases (14.7%), which is in keeping with our recent data showing RSV was far more commonly detected than influenza in cases of lower respiratory tract infection (LRTI) in Gambian children^[28]. RSV was significantly more common in children under 6 months of age when compared to older age groups and suggests that in contrast to influenza, there is limited protective impact of naturally occurring maternal antibody during this early period. This represents the burden of illness that could be prevented with an effective maternal RSV vaccine.

Detection of influenza and RSV both peaked during the rainy months of the year, but RSV was present at lower levels throughout the year, confirming our recent findings of RSV-associated LRTI in a different part of the country throughout the year^[28]. Of note, other investigators have noted that an ILI definition may underestimate the true burden of RSV due to the absence of fever in many cases^[29]. Establishing ongoing robust RSV surveillance in African countries will be important for the evaluation of RSV vaccines if and when rollout commences.

We observed that HMPV, parainfluenza virus, endemic coronaviruses and adenovirus were associated with ILI to a similar degree to influenza. There are limited data from sub-Saharan Africa looking at this broader range of viruses. HMPV in particular is increasingly associated with more severe respiratory illness^{[30, 31]}, which is in keeping with our finding that HMPV was associated with low oxygen saturations. Many respiratory viruses detected were present throughout the year and did not demonstrate any clear seasonal pattern. Data from several years are required to be confident about seasonality.

Finally, around three quarters of ILI cases were prescribed antibiotics, which was similar in cases that we later confirmed to have been associated with a respiratory virus. Thus, it is likely that many of these prescriptions were unnecessary.

Antibiotic use in such a high proportion of ILI cases is strikingly similar to the limited data from other low and middle-income countries (e.g. 83.7% of all influenza-associated ILI in Sri Lanka, 82% of all ILI in Thailand)^{[32, 33]}. The use of point-of-care testing for a broader range of viruses than just influenza and RSV could have a significant impact on antibiotic use. Unfortunately, such tools are not currently widely available in The Gambia or many parts of Africa. In addition, the future availability of vaccines against respiratory viruses could reduce inappropriate AMU^[34]. Influenza vaccines are rarely used in sub-Saharan Africa at present despite WHO recommendations for use in high risk groups such as children under five years of age^[35]. Several RSV vaccines are in clinical development though none are market approved currently^[36]. However, our data shows that RSV and influenza only account for a small proportion of the viruses associated with ILI. Therefore, broader antibiotic stewardship strategies including the availability of point-of-care tests, and robust treatment guidelines will also be essential if unnecessary AMU is to be reduced.