Introduction
Ancylostomatids (hookworms) together with ascarids (roundworms) are the most ubiquitous parasitic worms of companion animals (Traversa, Reference Traversa2012). Four species of ancylostomatids are considered as canine hookworms: Ancylostoma caninum, A. braziliense, A. ceylanicum and Uncinaria stenocephala (Bowman et al., Reference Bowman, Montgomery, Zajac, Eberhard and Kazacos2010). Hookworm infections may cause anaemia and hypoproteinemia (Loukas & Prociv, Reference Loukas and Prociv2001); the severity depends on factors such as the age of the animal, intensity of infection and species of hookworm involved (Traversa, Reference Traversa2012; Ng-Nguyen et al., Reference Ng-Nguyen, Hii, Nguyen, Van Nguyen, Van Nguyen and Traub2015).
The geographical distribution of different ancylostomatids is not clear, mostly because the eggs are morphologically indistinguishable (Palmer et al., Reference Palmer, Traub, Robertson, Hobbs, Elliot, While, Rees and Thompson2007; Bowman et al., Reference Bowman, Montgomery, Zajac, Eberhard and Kazacos2010). Overall, U. stenocephala is adapted to cold, temperate and subarctic regions, whereas Ancylostoma spp. occur in warmer areas worldwide (Kalkofen, Reference Kalkofen1987; Traversa, Reference Traversa2012). Ancylostoma caninum is the most common of all species of canine hookworms, having a more global distribution. Ancylostoma braziliense is restricted to the southern hemisphere, and A. ceylanicum is endemic in Asia and the Pacific, and is widespread in this region (Bowman et al., Reference Bowman, Montgomery, Zajac, Eberhard and Kazacos2010; Traub, Reference Traub2013; Traversa et al., Reference Traversa, di Regalbono, Di Cesare, La Torre, Drake and Pietrobelli2014).
Canine hookworms can infect or penetrate human skin, leading to hookworm infection, localized dermatological lesions (cutaneous larva migrans) and/or intestinal distress. However, the specific species responsible for clinical manifestations in humans are not always identified, and therefore it is not known whether all species of canine hookworms are of health concern for the human population. There is evidence for the zoonotic importance of A. braziliense, A. caninum and A. ceylanicum; however, the zoonotic potential of U. stenocephala remains unclear (Prociv & Croese, Reference Prociv and Croese1990, Reference Prociv and Croese1996; Croese et al., Reference Croese, Loukas, Opdebeeck, Fairley and Prociv1994; Landmann & Prociv, Reference Landmann and Prociv2003; Traub et al., Reference Traub, Inpankaew, Sutthikornchai, Sukthana and Thompson2008; Bowman et al., Reference Bowman, Montgomery, Zajac, Eberhard and Kazacos2010; Traversa, Reference Traversa2012).
Ancylostoma ceylanicum is the only species of canine hookworm known to cause natural patent infections in human hosts. Previously, human infections with A. ceylanicum have been categorized as uncommon, but this assumption has been questioned following studies that have identified A. ceylanicum as the second most common hookworm infecting human populations in South-east Asia (Traub et al., Reference Traub, Inpankaew, Sutthikornchai, Sukthana and Thompson2008; Sato et al., Reference Sato, Sanguankiat, Yoonuan, Pongvongsa, Keomoungkhoun, Phimmayoi, Boupa, Moji and Waikagul2010; Jiraanankul et al., Reference Jiraanankul, Aphijirawat and Mungthin2011; Conlan et al., Reference Conlan, Khamlome, Vongxay, Elliot, Pallant, Sripa, Blacksell, Fenwick and Thompson2012; Ngui et al., Reference Ngui, Ching, Kai, Roslan and Lim2012a, Reference Ngui, Lim, Traub, Mahmud and Mistamc; Traub, Reference Traub2013; Inpankaew et al., Reference Inpankaew, Schär and Dalsgaard2014).
Traditionally, egg morphology, and especially egg size, has been used to differentiate between species. However, this is not a reliable method, given the similar morphology and egg sizes between hookworm species (Ehrenford, Reference Ehrenford1953; Lucio-Forster et al., Reference Lucio-Forster, Liotta, Yaros, Briggs, Mohammed and Bowman2012). Species identification based on post-mortem recovery of adult worms or after anthelmintic expulsion is time-consuming and requires highly skilled staff, and there is a high possibility of misidentification (Traub et al., Reference Traub, Robertson, Irwin, Mencke and Thompson2004).
To overcome this obstacle several studies in recent decades have used molecular tools to identify species of canine hookworms. Zhu & Gasser (Reference Zhu and Gasser1998) used polymerase chain reaction–single-strand conformation polymorphism (PCR–SSCP), and Gasser et al. (Reference Gasser, Stewart and Speare1996), Traub et al. (Reference Traub, Robertson, Irwin, Mencke and Thompson2004, Reference Traub, Hobbs, Adams, Behnke, Harris and Thompson2007), e Silva et al. (Reference e Silva, Miranda, Santos and Rabelo2006), Palmer et al. (Reference Palmer, Traub, Robertson, Hobbs, Elliot, While, Rees and Thompson2007) and Liu et al. (Reference Liu, Zheng and Zhang2015) used PCR-linked restriction fragment length polymorphism (PCR-RFLP). Ngui et al. (Reference Ngui, Lim and Chua2012b) used PCR-high resolution melting (PCR-HRM), and Conlan et al. (Reference Conlan, Khamlome, Vongxay, Elliot, Pallant, Sripa, Blacksell, Fenwick and Thompson2012), Mahdy et al. (Reference Mahdy, Lim, Ngui, Fatimah, Choy, Yap, Al-Mekhlafi, Ibrahim and Surin2012), Ngui et al. (Reference Ngui, Lim, Traub, Mahmud and Mistam2012c), Liu et al. (Reference Liu, Zheng and Alsarakibi2014) and Oliveira-Arbex et al. (Reference Oliveira-Arbex, David, Oliveira-Sequeira, Katagiri, Coradi and Guimarães2017) used PCR and DNA sequencing. The objective of the present study was to identify the hookworm species present in canine faecal samples from Morogoro, Tanzania using PCR-RFLP and species-specific PCR coupled with Sanger sequencing.
Materials and methods
Study site and sampling
Non-randomized faecal samples were collected from local dogs within Morogoro municipality, approximately 195 km west of Dar es Salaam. In total, 160 samples were collected between September and December 2015 from four locations within Morogoro municipality: a dipping well at Sokoine University of Agriculture (SUA) main campus, a veterinary clinic at SUA main campus, private houses in Kilakala Ward, and a dipping well in Kihonda Ward. Most of the samples (78%) were from the two dipping facilities, where owners could have their dogs treated for ectoparasites. All samples were obtained per rectum except eight that were collected fresh from the floors of kennels. Samples were stored at 4°C for further analysis. To avoid duplicates, we recorded the name of the owner, the date and location of sampling, and distinctive features of each dog.
Parasitological procedures
Hookworm positive samples were identified using the McMaster technique (Roepstorff & Nansen, Reference Roepstorff and Nansen1998). As electrical power cuts were frequent in the Morogoro laboratory, we were not able to preserve the faecal samples at −20°C and had to rely on ethanol preservation at room temperature. Therefore we decided to produce and recover hookworm larvae (L3) as we expected this approach to generate parasite DNA of a higher quality compared to DNA extracted from faeces stored in ethanol. This would also enable us to perform single larva PCR, if needed. Of the 160 samples, 92 were found to be positive for strongyle-type eggs by McMaster, but only 68 had enough faecal material to set up a coproculture (Roepstorff & Nansen, Reference Roepstorff and Nansen1998). Larvae from those 68 samples were recovered after 10–15 days by the Baermann technique, stored in 70% ethanol and transported to the University of Copenhagen, Denmark for further molecular analysis.
Ethical considerations
A research associate permit was granted by the Directorate of Research and Postgraduate Studies, Sokoine University of Agriculture, Morogoro. At the sample setting and prior to any interaction with the dogs, the owners were informed in Swahili about the aim of the study and were asked for permission to take faecal samples from their dogs. All dogs were treated with ivermectin 1% (4 mg per 10–15 kg body weight) administered subcutaneously. Treatment was free of charge, with the exception of puppies younger than three months old and/or when interfering with the facility business (veterinarian care or deworming service). Toxicosis caused by macrocyclic lactones (ivermectin) might occur, especially in some collie-related breeds (Mealey, Reference Mealey2008). Based on phenotypic characteristics, there were no collie-type dogs among the dewormed animals.
Molecular procedures
Hookworm controls
Uncinaria stenocephala L3 larvae recovered from naturally infected Danish dogs were kindly supplied by Dr Helena Mejer (University of Copenhagen). A pool of A. ceylanicum L3 larvae was graciously provided by Professor Jennifer Keiser (Swiss Tropical and Public Health Institute), obtained after experimental infections in mice. Adult specimens of A. caninum and A. braziliense recovered from necropsied local dogs from Morogoro, Tanzania were generously provided by Mr Lufunyo Msalilwa (Sokoine University of Agriculture). The controls were confirmed by PCR-RFLP and Sanger sequencing (see below).
DNA extraction
Prior to DNA extraction, the larva pellet was incubated at 95°C to evaporate the ethanol used for storage. The DNeasy Blood & Tissue Kit spin column (QIAGEN, Germany) was used to extract DNA from the larvae and worms according to the manufacturer's instructions, except that the DNA was eluted with 50 μl of AE buffer.
PCR-RFLP
PCR. The set of primers used for amplifying the ITS-1, 5.8S and ITS-2 regions of the rDNA was a modification of the primers RTGHFI/RTABCR originally designed by Traub and co-workers (Traub et al., Reference Traub, Robertson, Irwin, Mencke and Thompson2004, Reference Traub, Hobbs, Adams, Behnke, Harris and Thompson2007). Sequences from the following species and their GenBank accession numbers were used to identify suitable target regions for primers: A. caninum (KP844730, JQ812694, DQ438077, DQ438074, DQ438075, DQ438079), A. ceylanicum (LC036567, KF279132, KF279134, DQ381541), A. braziliense (JQ812692, JQ812693, DQ438062, DQ438061, Q438050), U. stenocephala (AF194145, HQ262052–HQ262055), as well as sequences of the human hookworms Ancylostoma duodenale (EU344797) and Necator americanus (LC036565) and the cat hookworm Ancyclostoma tubaeforme (JQ812691). The forward primers CDCTSF (5′-GTGCTAGTCTTCACGACTTTG-3′) and the reverse CDCTSR (5′-CATGTTGCAATATATTCTGATCT-3′) amplifying a 570 bp region were designed using Primer3 (http://primer3.wi.mit.edu) (Rozen & Skaletsky, Reference Rozen, Skaletsky, Misener and Krawetz1999) and synthesized by TAG Copenhagen A/S, Denmark. We conducted the PCR in a total volume of 30 μl containing 3 μl of PCR 10X ammonium buffer, 20 mM MgCl2, 1.5 μl of deoxynucleotide triphosphate (dNTP) (4 mM), 1.2 pmol of each primer, 1.5 units of hot-start DNA polymerase and 1 μl of DNA as template. The PCR conditions were an initial denaturation of 95°C for 15 minutes followed by 40 cycles of 95°C for 30 s, 58°C for 40 s and 72°C for 60 s, and the final extension of 72°C for 10 minutes. The products were stained with GelRed (Biotium, USA) and visualized after gel electrophoresis (1.5% agarose) under UV light using Gel Doc (Bio-Rad, USA) and Quantity One software, version 4.3.1 (Bio-Rad, USA).
Restriction Fragment Length Polymorphism. The PCR products were subjected to digestion by three different restriction enzymes: HinfI, RsaI and Psp1406I (AclI) (Thermo Fisher Scientific, USA). Palmer et al. (Reference Palmer, Traub, Robertson, Hobbs, Elliot, While, Rees and Thompson2007) found that HinfI and RsaI could be used to distinguish between U. stenocephala, A. caninum, A. ceylanicum and A. tubaeforme because of their distinct cutting profiles (table 1) (Gasser et al., Reference Gasser, Stewart and Speare1996; Palmer et al., Reference Palmer, Traub, Robertson, Hobbs, Elliot, While, Rees and Thompson2007). AclI was included because A. caninum and A. duodenale are undistinguishable by HinfI/RsaI digestion. Only A. duodenale shows a cutting site for AclI (George et al., Reference George, Kaliappan, Kattula, Roy, Geldhof, Kang, Vercruysse and Levecke2015). We used three units of endonuclease to digest 2 μl of the PCR product at 37°C for three hours; however, in case of weak PCR amplification up to 7 μl of product was digested. The band profiles were visualized by gel electrophoresis as described above. To confirm the species identity obtained by the PCR-RFLP procedure, PCR products from ten samples were bidirectional sequenced by Macrogen Inc. (Seoul, South Korea). The sequences were analysed and assembled using the software Vector NTI (Lu & Moriyama, Reference Lu and Moriyama2004), aligned and trimmed with BioEdit version 7.2.5 (Hall, Reference Hall1999) and compared to sequences in the NCBI database using BLAST (National Center for Biotechnology Information, Maryland, USA).
Approximate fragment lengths (bp) generated for various hookworm species after enzymatic digestion of the 570 bp PCR product using three restriction enzymes: HinfI, RsaI and Psp1406I (AclI).
a For U. stenocephala the RsaI digestion region contains a polymorphic site, which may result in two distinct band profiles (Nadler et al., Reference Nadler, Lyons and Pagan2013).
b For A. tubaeforme allelic polymorphism may result in two distinct band patterns (Palmer et al., Reference Palmer, Traub, Robertson, Hobbs, Elliot, While, Rees and Thompson2007).
Species-specific PCR and sequencing
As the samples potentially contained DNA from larvae of different species, we aimed to develop species-specific primers to identify the canine hookworm species with lower infection intensities, which might be overlooked by the PCR-RFLP technique. As A. caninum was identified in all samples by sequencing the CDCTSF/CDCTSR PCR products, we only designed primers to target A. braziliense, A. ceylanicum and U. stenocephala (table 2). For primer design, we used Primer-Blast (Ye et al., Reference Ye, Coulouris, Zaretskaya, Cutcutache, Rozen and Madden2012) and the previously mentioned reference sequences in addition to the following sequences of the cox1 region: A. caninum (AJ407961, AJ407962, AJ407964, AJ407966, AJ407941), A. ceylanicum (KP072070-1, KP072079-0, KP072075-7, KM066109, AF225917, LC036568, KC247734, KC247737, KC247740, KC247743-5), U. stenocephala (AJ407939), A. duodenale (AJ407942, AJ407952, AJ407953, AJ407954, AJ407956, AJ407958, AJ407959, AJ407960, AJ407968), N. americanus (AB793562, AM980929, AM980900, AJ407963) and A. tubaeforme (FR846510.1, FR846503.1, FR846511.1, FR846505.1, AJ407940.1).
The three species-specific primer sets used, with species targeted, nucleotide sequence, and the size and region of the product amplified.
The PCR volume was adjusted to 20 μl per reaction, containing 2 μl of PCR 10X ammonium buffer, 20 mM MgCl2, 1 μl of dNTP (4 mM), 0.8 pmol forward and reverse primers, 1 unit of hot-start DNA polymerase and 1 μl of DNA as template. The PCR running conditions were the same as for the PCR-RFLP but using annealing temperatures of 61, 62 and 63°C for A. braziliense, A. ceylanicum and U. stenocephala, respectively. The PCR products were visualized by gel electrophoresis as described above.
To test the specificity of the species-specific PCR, ten A. braziliense, five A. ceylanicum and five U. stenocephala PCR-positive samples were sequenced in both directions (Macrogen Inc., Seoul, South Korea) and analysed as previously described.
Results
Copromicroscopy examination
Eggs of three species of gastrointestinal helminths were identified in 94 of 160 samples (59%). Hookworms were the most frequently found parasite, present in 92 samples (58%), followed by Toxocara spp. in nine samples (6%). Dipylidium caninum proglottids were detected macroscopically and confirmed microscopically in two samples (1%). Of the 92 hookworm positive samples we were able to recover larvae from only 68 samples because of the limited amount of faecal material collected from some dogs (fig. 1). The median and mean numbers of larvae stored were 137 and 1993, respectively (range 3–44,150). The number of larvae recovered varied greatly from sample to sample, mostly related to differences in number of eggs per gramme of faeces (epg) and weight of faecal material.
Process used for analysis of faecal samples from 160 local dogs in Morogoro, Tanzania, to identify the hookworm species present.
PCR-RFLP
We extracted DNA successfully from 66 of the 68 samples. For the PCR-RFLP all 66 samples showed the same band profile and were identified as A. caninum. Ten of the CDCTSF/CDCTSR PCR products were sequenced successfully and confirmed as A. caninum with 100% identity with previously published GenBank sequences (KP844730.1, JQ812694.1 and DQ438078.1). Sequences are deposited in GenBank with the accession numbers MG890369–MG890378.
Species-specific PCR and sequencing
For A. braziliense, 56 of the 66 samples were positive, and ten samples were randomly selected and all sequenced successfully and confirmed as such with a similarity of 98–100% with previously published sequences of this species (i.e. DQ438069.1, JQ083592.1 and DQ438054.1). Sequences are deposited in GenBank with the accession numbers MG890379–MG890388.
Five samples were PCR positive for A. ceylanicum and the five were sequenced successfully and identified as such with a similarity of 99–100% with A. ceylanicum sequences in GenBank (KF896598 and AF225917). Sequences are deposited in GenBank with the accession numbers MG914068–MG914072.
For U. stenocephala five samples were PCR positive but only two were sequenced successfully and deposited in GenBank, with the accession numbers MG890389–MG890390. They showed a similarity of 99% with all five previously published U. stenocephala sequences (HQ262055.1, AF194145.1, HQ262053.1, HQ262052.1 and HQ262054.1).
Discussion
We identified four canine hookworm species in dog faeces from Morogoro, Tanzania. To our knowledge this is the first report documenting the presence of U. stenocephala and A. ceylanicum in Africa using molecular techniques.
The identification of A. caninum and A. braziliense was to be expected, as they are ubiquitous species among dogs in the southern hemisphere, adapted to warm and humid climates. However, the presence of U. stenocephala, which is assumed to be distributed in colder areas, as well as A. ceylanicum, thought to be confined to Asia and the Pacific, was unexpected (Gibbs & Gibbs, Reference Gibbs and Gibbs1959; Bowman et al., Reference Bowman, Montgomery, Zajac, Eberhard and Kazacos2010; Traub, Reference Traub2013).
Nevertheless, there have been previous reports of A. ceylanicum in Africa, based on morphological identification (Yoshida, Reference Yoshida1971), and A. ceylanicum has been morphologically identified in South African cats (Baker et al., Reference Baker, Lange, Verster and van der Plaat1989). However, identification of A. ceylanicum from reports prior to 1951 may not be reliable, as it was in that year that Biocca resolved that A. braziliense and A. ceylanicum were different species (Biocca, Reference Biocca1951).
Our results support the idea that A. ceylanicum has been overlooked and that its distribution is more geographically widespread than formerly reported (Traub, Reference Traub2013). This is important for both veterinary and public health. Ancylostoma ceylanicum has been considered to be a rare or non-relevant species regarding human hookworm infection (Ngui et al., Reference Ngui, Ching, Kai, Roslan and Lim2012a; Traub, Reference Traub2013); however, recent studies in Asia using molecular techniques have shown that A. ceylanicum was present in between 5 and 51% of people infected with hookworms (Traub et al., Reference Traub, Inpankaew, Sutthikornchai, Sukthana and Thompson2008; Sato et al., Reference Sato, Sanguankiat, Yoonuan, Pongvongsa, Keomoungkhoun, Phimmayoi, Boupa, Moji and Waikagul2010; Jiraanankul et al., Reference Jiraanankul, Aphijirawat and Mungthin2011; Conlan et al., Reference Conlan, Khamlome, Vongxay, Elliot, Pallant, Sripa, Blacksell, Fenwick and Thompson2012; Ngui et al., Reference Ngui, Lim, Traub, Mahmud and Mistam2012c; Traub, Reference Traub2013; Inpankaew et al., Reference Inpankaew, Schär and Dalsgaard2014; George et al., Reference George, Kaliappan, Kattula, Roy, Geldhof, Kang, Vercruysse and Levecke2015; Bradbury et al., Reference Bradbury, Hii, Harrington, Speare and Traub2017). Therefore, the presence of A. ceylanicum in dogs from Morogoro suggests that this hookworm species may also be present in people living in the area.
The presence of U. stenocephala in Morogoro could be explained by the close relationship of Sokoine University of Agriculture with Denmark. The Division of Veterinary Sciences was established at SUA in 1976, supported by Danida and followed up by professor exchange programmes, during which Danish professors visited for extended periods of time and may have travelled with their dogs infected with U. stenocephala.
Applying the PCR-RFLP method, we overlooked A. ceylanicum and U. stenocephala. This could be explained by the exponential amplification of DNA of the most abundant species during the PCR (A. caninum in this case). Consequently, the PCR products subjected to enzymatic digestion were either just A. caninum or were outnumbered by this species. This has important implications, as this method cannot be used reliably to estimate the prevalence of specific hookworm species in an area and may overlook the presence of certain species if few in numbers. Furthermore, for the PCR-RFLP it should be noted that intraspecific and intra-individual variation in the target region can lead to complex cutting profiles and therefore dubious species characterization (Chilton, Reference Chilton2004). This was most likely the case for our samples, as we consistently observed an undigested band (570 bp) after RsaI digestion. This band remained intact despite the enzyme increment (3 times) and was also observed in the A. caninum control extracted from a single adult worm. Thus, this pattern may be related to intra-individual variation in the rDNA repeats, an observation that was reported previously by Gasser et al. (Reference Gasser, Stewart and Speare1996). Importantly, this polymorphism can lead to misidentifications with U. stenocephala, as this species’ PCR product can also be left undigested by the RsaI enzyme. However, this can be excluded in our case as we did not observe the U. stenocephala band profile by HinfI digestion.
In addition, the in silico analysis showed that the cutting site for the Psp1406I (AclI) enzyme is located in a polymorphic region. Some of the reference sequences of A. duodenale deposited in GenBank do not present the specific restriction site (AA^CGTT) in the amplified ITS region (accession numbers KC896800, AB504714-15, AB501348-51). Hence, it may not be a reliable method for discerning between A. caninum and A. duodenale. However, despite the ambiguity when analysing reference sequences, experimental studies have distinguished between these two species by the use of Psp1406I (AclI), which were also confirmed by DNA sequencing (George et al., Reference George, Kaliappan, Kattula, Roy, Geldhof, Kang, Vercruysse and Levecke2015).
It is important to emphasize that this study aimed to identify which canine hookworm species were present, and not their prevalences. To achieve that, we developed a species-specific PCR, which cannot currently be used as a diagnostic method as it was not validated with other hookworm species (A. duodenale, N. americanus and A. tubaeforme) nor other species of helminths. Therefore, any positive sample should be sequenced to verify the expected identity of the PCR product. Considering that we could confirm the presence of A. caninum by sequencing the CDCTSF/CDCTSR PCR product, we did not develop A. caninum-specific primers. Nevertheless, we designed specific primers for A. ceylanicum, A. braziliense and U. stenocephala for PCR followed by sequencing to reliably confirm their identity. As 56 samples were positive for A. braziliense we sequenced only ten randomly selected samples, as we regarded this as sufficient to prove their presence among our samples.
Further studies on species prevalence and distribution are needed for improved control of hookworms in the future. To achieve this, it is of paramount importance to develop sensitive and specific diagnostic tools that are able to distinguish between human, feline and canine hookworm species. A cross-sectoral ‘one health’ approach is not only essential when addressing diagnoses but also regarding epidemiology. Current control of hookworms in the human population is based on treatment of the individual (pregnant women and children under five years of age) or as an added advantage in control programmes for schistosomiasis and lymphatic filariasis using mass drug administration (WHO, 2002, 2006). However, if zoonotic transmission is ignored, control programmes are unlikely to be fully successful with regards to reduction of reinfection among humans (Thompson & Conlan, Reference Thompson and Conlan2011; Traub, Reference Traub2013).
In conclusion, local dogs from Morogoro are reservoirs of zoonotic hookworm species. Of special interest is the finding of A. ceylanicum, previously thought to be confined to Asia, where it causes hookworm infection in high numbers of people. Nevertheless, we were able to identify this species (along with U. stenocephala) only by the use of species-specific primers coupled with Sanger sequencing, and not by the PCR-RFLP method, which may question its reliability as a diagnostic tool. Our results dispute the assumed geographical distribution or confinement of the various hookworm species and call for further research.
Acknowledgements
We acknowledge the veterinarians and veterinary students from SUA without whose generous help this study would not have been possible. We also thank the welcoming staff of the Department of Veterinary Microbiology and Parasitology and the Department of Veterinary Medicine and Public Health, SUA, with special thanks to Dr Helena A. Ngowi.
Financial support
This work was kindly supported by the Danida Fellowship Centre, the Oticon Foundation, and Kong Christian den Tiendes Fond.
Conflict of interest
None of our funders have influenced the conduct or reporting of this study. The authors have no conflict of interest.
