Streptococcus pneumoniae (pneumococcus) is the predominant bacterial respiratory pathogen during childhood. Despite the increasing use of pneumococcal conjugate vaccines (PCV), S. pneumoniae is still responsible for more than 500,000 child deaths each year.1 Colonization of the nasopharynx precedes respiratory infections and invasive disease caused by S. pneumoniae.2 Variations in the local environment of the nasopharynx affect whether pneumococcal colonization occurs and, once colonization is established, whether infection occurs.3 , 4 The nasopharyngeal microbiota may provide a barrier to colonization by S. pneumoniae through competition for nutrients, production of antimicrobial substances or modulation of the local immune response.5 However, the bacteria within the nasopharyngeal microbiota that modify the risk of pneumococcal colonization have not been systematically identified.
In this study, we used nasopharyngeal samples from children in Botswana to identify clinical risk factors for pneumococcal colonization and to identify associations between the bacterial composition of the nasopharyngeal microbiota and S. pneumoniae colonization.
MATERIALS AND METHODS
Botswana is a landlocked country in southern Africa that has a semiarid climate, with a short rainy season that typically occurs from November to March. The country’s under-five child mortality rate was estimated to be 41 per 1,000 live births in 2016.6 Haemophilus influenzae type B (Hib) and 13-valent pneumococcal conjugate vaccine (PCV-13) were introduced in November 2010 and July 2012, respectively. Gaborone, the capital and largest city of Botswana, is located in the country’s South-East District.
We conducted a case-control study of pneumonia between April 2012 and June 2016, as described elsewhere in detail.7 For the purposes of these analyses, we utilized the control group comprised of children 1–23 months of age receiving well child or acute care services at one of 18 public clinics in the Gaborone area. Controls were matched to pneumonia cases 1:1 by primary care clinic and date (≤ 2 weeks from the case enrollment). Exclusion criteria included World Health Organization clinical pneumonia (“cough or difficult breathing with lower chest wall indrawing”), a chronic medical condition predisposing to pneumonia (other than HIV), hospitalization in the prior 14 days and asthma.8 Children with upper respiratory symptoms—defined as “cough or nasal congestion or discharge”—were eligible for inclusion. Sociodemographic and clinical data were recorded from a face-to-face interview with the child’s caregivers and review of infant and maternal medical records.
Nasopharyngeal specimens were collected using flocked swabs and universal transport medium (Copan Italia, Brescia, Italy). These samples were placed into a cooler with ice packs immediately after collection and transported within 6 hours to the National Health Laboratory in Gaborone where they were transferred to cryovials for storage at -80°C. Samples were shipped on dry ice at 6-month intervals to St. Joseph’s Healthcare (Hamilton, ON, Canada). Total genomic DNA was extracted from nasopharyngeal specimens using a protocol involving mechanical and enzymatic lysis.9 Polymerase chain reaction (PCR) was used to amplify the variable V3 region of the bacterial 16S ribosomal RNA (rRNA) gene using broad-range primers and an Illumina MiSeq instrument (BioProject accession number PRJNA423191).9 , 10 These same nasopharyngeal specimens were tested for respiratory syncytial virus, influenza viruses A and B, parainfluenza virus types 1–4, human metapneumovirus, adenovirus, coronaviruses (NL63, 229E, OC43, HKU1), bocavirus and rhinovirus/enterovirus using nucleic acid extracted on the NucliSens easyMAG platform (bioMérieux, Marcy l’Etoile, France) and amplified using a Rotor-Gene 6500 real-time PCR thermocycler (Qiagen, Hilden, Germany). This nucleic acid extract was also tested for S. pneumoniae using a quantitative PCR assay targeting the autolysin (lytA) gene.9
Raw sequences were analyzed through a pipeline that included Cutadapt and PANDAseq tools.9 Sequences were clustered at 97% identity using AbundantOTU+, and given taxonomic assignments using the Ribosomal Database Project classifier against the Greengenes 2011 database.9 Each read was grouped into an operational taxonomic unit (OTU), a cluster of similar 16S rRNA sequences corresponding roughly to a species or a group of closely related species. We excluded 3 nasopharyngeal samples with less than 5,000 reads. Using the R package phyloseq, OTUs with fewer than 10 counts across all samples or that did not appear in at least 10% of samples were removed.11 Sample reads were then normalized to the cumulative 75th percentile using the metagenomeSeq package in R.12 Because a single OTU accounted for > 75% of reads for 82% of genera, we aggregated OTUs at the genus level for compositional analyses.
Characteristics of children colonized and noncolonized with S. pneumoniae were compared using χ2 or Fisher exact tests (for categorical variables) and 2-sample t tests (for continuous variables). Species richness was measured by the Chao1 index, and microbial diversity was assessed using the Shannon diversity index.13 , 14 After verifying that these measures approximated normal distributions in the study population, we used multivariable linear regression to compare the Chao1 and Shannon diversity indices in nasopharyngeal samples from colonized and noncolonized children. We considered genera with a mean relative abundance of ≥ 0.5% in the overall population to be highly abundant. We included these genera in logistic regression models to evaluate whether the proportion of each genus was associated with S. pneumoniae colonization. These analyses were adjusted for age, given the substantial changes in the nasopharyngeal microbiota during infancy that we previously reported in this study population.15 We used random forest models implemented with the R package caret to identify the bacterial genera that were most important in discriminating S. pneumoniae–colonized children from noncolonized children.16 For these analyses, the study population was divided into training (50%) and test (50%) sets and the genera Streptococcus and unclassified Streptococcaceae were excluded to ensure that S. pneumoniae was not assigned to genera within the model. The model was initially fit on the training set, and the sensitivity and specificity of the model were then calculated based on fitting of test set data. To further evaluate the association between S. pneumoniae colonization and upper respiratory symptoms, we included the density of S. pneumoniae colonization (measured as copies/mL by quantitative PCR) in a multivariable logistic regression model adjusting for the following potential confounders: age, sex, season, enrollment year, maternal HIV infection, current breastfeeding, electricity in the home, household use of wood as a cooking fuel, number of household members, PCV-13 doses, receipt of antibiotics within the prior 7 days and detection of one or more respiratory viruses. Statistical analyses were conducted using R version 3.4.2.
One hundred seventy children were included in these analyses. Mean age was 8.3 months, and 51% were female. Nasopharyngeal colonization by S. pneumoniae was detected in 96 (56%) children. Sixty (35%) children had one or more respiratory viruses, the majority (76%) of which were rhinovirus/enterovirus infections. Characteristics of the study population are shown in Table 1. Age, electricity in the home, household use of wood as a cooking fuel and the prevalence of upper respiratory symptoms differed between S. pneumoniae-colonized and noncolonized children.
Nasopharyngeal Microbiota Diversity
A total of 9,016,655 high-quality 16S rRNA sequences (mean of 53,039 sequences per sample) were obtained from the 170 nasopharyngeal samples included in this study. Rarefaction curves were constructed to ensure coverage of the bacterial diversity present (Supplemental Figure 1, http://links.lww.com/INF/D256). All samples had sufficient sequencing depth to obtain a high degree of sequence coverage. The mean (SD) Chao1 and Shannon diversity indices in the overall population were 250.5 (106.5) and 1.30 (0.61). Species richness, as measured by the Chao1 index, and microbial diversity, assessed using the Shannon diversity index, did not differ in S. pneumoniae–colonized and noncolonized children after adjustment for age (Supplemental Figure 2, http://links.lww.com/INF/D257).
Nasopharyngeal Microbiota Composition
Sequences were clustered into 336 bacterial OTUs (after filtering), representing 155 genera from 10 phyla. Principal component analysis using Bray-Curtis distances demonstrated that overall nasopharyngeal samples clustered by detection of S. pneumoniae and that S. pneumoniae–colonized and noncolonized microbial communities differed significantly (Fig. 1; P < 0.001 by PERMANOVA, adjusting for age). The abundances of highly abundant genera in children by S. pneumoniae colonization status are shown in Figures 2, 3. Adjusting for age, S. pneumoniae colonization was associated with higher relative abundances of Moraxella (P = 0.001) and Streptococcus (P < 0.0001) and lower relative abundances of Corynebacterium (P = 0.001) and Staphylococcus (P = 0.03). Respiratory virus infection was associated with a higher relative abundance of Streptococcus (P = 0.01). The most abundant genera in nasopharyngeal samples contained typical commensal bacteria (Corynebacterium, Dolosigranulum) and common potential respiratory pathogens (Haemophilus, Moraxella, Staphylococcus and Streptococcus). Each of the highly abundant genera containing a common respiratory pathogen was comprised almost entirely of a single dominant OTU (Supplemental Figure 3, http://links.lww.com/INF/D258).
To further evaluate for confounding by age, we performed analyses in a subset of S. pneumoniae–colonized and noncolonized children matched on age. Mean (SD) ages of the 57 noncolonized and 57 colonized children in these matched analyses were 7.9 (4.8) and 8.2 (4.8) months, respectively. Analyses comparing nasopharyngeal microbial diversity and composition in S. pneumoniae–colonized and noncolonized children were substantively unchanged in this matched cohort.
Bacterial Genera Predictive of Pneumococcal Colonization
A random forest analysis including nonstreptococcal bacterial genera was performed to identify the genera that were most useful in discriminating S. pneumoniae–colonized and noncolonized children (Fig. 4). We calculated a sensitivity of 81% and a specificity of 85% for the determination of S. pneumoniae colonization status by this decision tree model. The most discriminative genera were Moraxella, unclassified Lactobacillales, Haemophilus, unclassified Bacillaceae and Klebsiella. Among the top 20 predictive genera that were not highly abundant, S. pneumoniae colonization was associated with a higher relative abundance of unclassified Lactobacillales (P = 0.005) and lower relative abundances of Chryseobacterium (P = 0.03), Granulicatella (P = 0.04), Halomonas (P = 0.04), Rothia (P = 0.03) and unclassified Bacillaceae (P = 0.046) after adjustment for age.
Upper Respiratory Symptoms
We performed additional analyses to investigate the association between S. pneumoniae colonization and upper respiratory symptoms. First, even when children with respiratory viruses were excluded, we found that the prevalence of upper respiratory symptoms was higher in children colonized with S. pneumoniae than in noncolonized children (49% vs. 20%; P =0.002). Moreover, we found that the density of S. pneumoniae colonization was an independent predictor of the presence of upper respiratory symptoms (P = 0.001) in a multivariable logistic regression model. The composition of the nasopharyngeal microbiota of children with upper respiratory symptoms also differed significantly from that of children without upper respiratory symptoms (P < 0.001 by PERMANOVA, adjusting for age). The presence of upper respiratory symptoms was associated with a higher relative abundance of Streptococcus (P = 0.0004) and a lower relative abundance of Corynebacterium (P = 0.04) in analyses adjusted for age.
In this study of infants and young children in Botswana, we observed marked alterations in nasopharyngeal microbial composition in children colonized with S. pneumoniae. These alterations included changes in the relative abundances of genera containing potential respiratory pathogens (Moraxella, Staphylococcus) and a genera that contain common nasopharyngeal commensals (Corynebacterium).
The nasopharynx is the site of dynamic interactions between diverse communities of commensal bacteria and potential respiratory pathogens. Our findings indicate that colonization of the nasopharynx by S. pneumoniae is associated with distinct alterations in these microbial communities. Specifically, we observed a higher relative abundance of Moraxella and lower relative abundances of Corynebacterium and Staphylococcus in children colonized with S. pneumoniae. Although the sequencing methodology that we used did not enable determination of bacterial composition at the species level, the genera containing common respiratory pathogens (Moraxella, Staphylococcus, Streptococcus) were dominated by a single OTU, and prior culture-based studies suggest that Moraxella catarrhalis, Staphylococcus aureus and S. pneumoniae are the predominant species from these genera within the nasopharynxes of children.17 , 18 The observed associations between the relative abundances of Moraxella and Staphylococcus and detection of S. pneumoniae by species-specific PCR are consistent with previous studies that examined density interactions between these bacteria in the upper respiratory tract.19–22 These studies generally reported a synergistic relationship between S. pneumoniae and M. catarrhalis and an antagonistic relationship between S. pneumoniae and S. aureus.19–23 A prior study that used high-throughput sequencing methods also reported an inverse association between S. pneumoniae and Corynebacterium, as was observed among children in our cohort.24 Using a decision tree model, we also identify several other bacterial genera that accurately discriminate S. pneumoniae–colonized from noncolonized children. While these findings should be interpreted with caution, given the cross-sectional study design, the associations suggest that these genera may modify the risk of pneumococcal colonization. The identification of bacteria that inhibit colonization by S. pneumoniae and other respiratory pathogens could form the basis for novel pneumonia prevention strategies, including the use of a probiotic cocktail to selectively modify the nasopharyngeal microbiota such that it more effectively resists pathogen colonization.
We found that the biodiversity of nasopharyngeal microbial communities was similar in children with and without S. pneumoniae colonization. Given the cross-sectional design of this study, we cannot exclude the possibility that nasopharyngeal microbial diversity influences S. pneumoniae colonization risk, or that nasopharyngeal colonization by S. pneumoniae alters microbial diversity. Laufer et al24 reported that culture-identified S. pneumoniae colonization was associated with lower microbial diversity among children with and without acute otitis media in the United States. Notably, nasopharyngeal microbial diversity in this study was substantially higher than in our cohort, which could be related to differences in age, socioeconomic status or genetic and environmental factors between these populations.24 Conversely, higher nasopharyngeal microbial diversity was associated with the establishment of colonization in a pneumococcal challenge study conducted in adults, breaking with the traditional dogma that a higher diversity microbial community is more effective in resisting colonization by exogenous pathogens.25 We may not have found an association between S. pneumoniae colonization and nasopharyngeal microbial diversity because colonization is associated with only a transient loss of diversity, or because this effect is strain-dependent. In a mouse model, pneumococcal challenge led to a reduction in microbial diversity, but diversity had returned to baseline within 1 week.26 Moreover, in the previously described pneumococcal challenge study conducted in adults, microbial diversity did not change following challenge with a serotype 6B strain, while challenge with a serotype 23F strain actually transiently increased diversity.25
S. pneumoniae colonization was associated with upper respiratory symptoms in our cohort, even after adjusting for respiratory virus infection and other measured confounding factors. This raises the possibility that S. pneumoniae colonization may be a symptomatic event in some infants and young children. Although historically upper respiratory symptoms have been attributed to respiratory viruses, 2 prior studies in children also suggest a potential role for S. pneumoniae. In a longitudinal study of 234 Australian children, a nasopharyngeal microbiota profile dominated by Streptococcus was associated with acute respiratory infection symptoms independent of respiratory virus infection.27 Moreover, in a study of 2,840 children in Kenya, S. pneumoniae colonization was associated with upper respiratory symptoms in the preceding 2 weeks, although testing for respiratory viruses was not performed.28 The potential mechanisms for this association have not been defined, but mild nasal mucosal inflammation is observed following exposure to S. pneumoniae. In a study conducted among adults in the United States, experimental pneumococcal challenge resulted in recruitment of neutrophils and mononuclear cells to the nasal mucosa in some subjects.29 Moreover, it is possible that S. pneumoniae colonization results in upper respiratory symptoms through alterations of the nasopharyngeal microbiota. In particular, we found that the presence of upper respiratory symptoms was associated not only with enrichment of the nasopharyngeal microbiota with Streptococcus but also with a lower relative abundance of Corynebacterium, a genus that includes common nasopharyngeal commensals.
Respiratory viruses are postulated to play a key role in the pathogenesis of S. pneumoniae infections.20 Although the synergistic interactions between influenza viruses and S. pneumoniae are perhaps the most well described, S. pneumoniae colonization has been associated with infection by most common respiratory viruses, including rhino-enteroviruses, respiratory syncytial virus and parainfluenza viruses.20 One mechanism by which respiratory viruses have been proposed to influence the risk of S. pneumoniae infection is through alteration of the composition of the nasopharyngeal microbiota. In our cohort, respiratory virus infection was associated with a higher relative abundance of Streptococcus but was not associated with the relative abundances of other highly abundant genera. Prior studies evaluating the impact of respiratory virus infection on nasopharyngeal microbial composition suggest that this effect may be virus-specific. In a study of 740 infants hospitalized for bronchiolitis, respiratory syncytial virus (RSV) and rhinovirus infections were associated with specific alterations of the nasopharyngeal microbiota.30 Similarly, higher relative abundances of Streptococcus and Haemophilus and a lower relative abundance of Staphylococcus were observed in a cohort of previously healthy children with RSV infection.31 Finally, among pneumonia patients, influenza A (H1N1) virus infection was associated with enrichment of the oropharyngeal microbiota by Pseudomonas, Bacillus and Ralstonia species possessing certain signaling and motility genes.32
Our study has several limitations. First, given the cross-sectional design, we were unable to assess whether the alterations of the nasopharyngeal microbiota observed in children with S. pneumoniae predisposed to or were the result of pneumococcal colonization. Moreover, because the median duration of pneumococcal colonization may exceed 1 month in infants and young children, we are unable to determine the timing of S. pneumoniae acquisition relative to nasopharyngeal sampling in the study population.33 In addition, the method for 16S rRNA gene sequencing that we used did not permit determination of bacterial composition at the species level. Alternative approaches, such as shotgun sequencing, could provide a more detailed understanding of bacterial density interactions in the upper respiratory tract. Although we tested for the most common respiratory viruses in children, it remains possible that the association between S. pneumoniae colonization and upper respiratory symptoms is related to undetected respiratory virus infection or other unmeasured factors. Moreover, given that the majority of viral infections in our cohort were rhino-enteroviruses, we were unable to evaluate for associations between infection with other respiratory viruses and the composition of the nasopharyngeal microbiota. Finally, we excluded Streptococcus and unclassified Streptococcaceae from the microbial decision tree model. Although this was necessary to ensure that S. pneumoniae was not assigned to genera within the model, the exclusion of nonpneumococcal streptococci may have reduced the accuracy of this microbial decision tree model. In particular, previous studies observed an inverse relationship between S. pneumoniae colonization and the density of viridans group streptococci in the nasopharynx.18 , 34
In conclusion, pneumococcal colonization was associated with characteristic alterations of the nasopharyngeal microbiota of children that reflect synergistic and antagonist interactions of S. pneumoniae with commensal bacteria and other potential respiratory pathogens. Prospective studies are needed to determine if the nasopharyngeal microbiota provides a barrier to S. pneumoniae colonization and thus could be modified as a novel pneumonia prevention strategy.
The authors would like to thank Copan Italia (Brescia, Italy) for their generous donation of the universal transport media and flocked swabs used for the collection of nasopharyngeal specimens. The authors offer their sincere thanks to the children and families who participated in this study.
2. Bogaert D, De Groot R, Hermans PWStreptococcus pneumoniae colonisation: the key to pneumococcal disease. Lancet Infect Dis. 2004;4:144–154.
3. Wolter N, Tempia S, Cohen C, et alHigh nasopharyngeal pneumococcal density, increased by viral coinfection, is associated with invasive pneumococcal pneumonia. J Infect Dis. 2014;210:1649–1657.
4. García-Rodríguez JA, Fresnadillo Martínez MJDynamics of nasopharyngeal colonization by potential respiratory pathogens. J Antimicrob Chemother. 2002;50:59–73.
5. de Steenhuijsen Piters WA, Sanders EA, Bogaert DThe role of the local microbial ecosystem in respiratory health and disease. Philos Trans R Soc Lond B Biol Sci. 2015;370.
7. Kelly MS, Smieja M, Luinstra K, et alAssociation of respiratory viruses
with outcomes of severe childhood pneumonia in Botswana. PLoS One. 2015;10:e0126593.
8. World Health Organization. Pocket Book of Hospital Care for Children
: Guidelines for the Management of Common Childhood Illnesses. 2013.Geneva, Switzerland: WHO;
9. Whelan FJ, Verschoor CP, Stearns JC, et alThe loss of topography in the microbial communities
of the upper respiratory tract in the elderly. Ann Am Thorac Soc. 2014;11:513–521.
10. Bartram AK, Lynch MD, Stearns JC, et alGeneration of multimillion-sequence 16S rRNA gene libraries from complex microbial communities
by assembling paired-end illumina reads. Appl Environ Microbiol. 2011;77:3846–3852.
11. McMurdie PJ, Holmes Sphyloseq: an R package for reproducible interactive analysis and graphics of microbiome census data. PLoS One. 2013;8:e61217.
12. Paulson JNmetagenomeSeq: Statistical analysis for sparse high-throughput sequencing. Bioconductor package. 2014.
13. Gotelli NJ, Colwell RKMagurran AE, McGill BJEstimating species richness. In: Biological Diversity: Frontiers in Measurement and Assessment. 2011:Oxford, United Kingdom: Oxford University Press; 39–54.
14. Magurran AEMeasuring Biological Diversity. 2004.2nd ed. Oxford, United Kingdom: Blackwell Science Ltd;
15. Kelly MS, Surette MG, Smieja M, et alThe nasopharyngeal microbiota
with respiratory infections in Botswana. Pediatr Infect Dis J. 2017;36:e211–e218.
16. Kuhn M, Wing J, Weston S, et alCaret: Classification and Regression Training. R Package Version 6.0–21. 2015.Vienna, Austria: CRAN;
17. Gehanno P, Lenoir G, Barry B, et alEvaluation of nasopharyngeal cultures for bacteriologic assessment of acute otitis media in children
. Pediatr Infect Dis J. 1996;15:329–332.
18. Faden H, Stanievich J, Brodsky L, et alChanges in nasopharyngeal flora during otitis media of childhood. Pediatr Infect Dis J. 1990;9:623–626.
19. Chien YW, Vidal JE, Grijalva CG, et alDensity interactions among Streptococcus pneumoniae, Haemophilus influenzae and Staphylococcus aureus in the nasopharynx of young Peruvian children
. Pediatr Infect Dis J. 2013;32:72–77.
20. Bosch AA, Biesbroek G, Trzcinski K, et alViral and bacterial interactions in the upper respiratory tract. PLoS Pathog. 2013;9:e1003057.
21. Jacoby P, Watson K, Bowman J, et alKalgoorlie Otitis Media Research Project Team. Modelling the co-occurrence of Streptococcus pneumoniae with other bacterial and viral pathogens in the upper respiratory tract. Vaccine. 2007;25:2458–2464.
22. Bogaert D, van Belkum A, Sluijter M, et alColonisation by Streptococcus pneumoniae and Staphylococcus aureus in healthy children
. Lancet. 2004;363:1871–1872.
23. Pericone CD, Overweg K, Hermans PW, et alInhibitory and bactericidal effects of hydrogen peroxide production by Streptococcus pneumoniae on other inhabitants of the upper respiratory tract. Infect Immun. 2000;68:3990–3997.
24. Laufer AS, Metlay JP, Gent JF, et alMicrobial communities of the upper respiratory tract and otitis media in children
. MBio. 2011;2:e00245–e00210.
25. Cremers AJ, Zomer AL, Gritzfeld JF, et alThe adult nasopharyngeal microbiome as a determinant of pneumococcal acquisition. Microbiome. 2014;2:44.
26. Krone CL, Biesbroek G, Trzciński K, et alRespiratory microbiota
dynamics following Streptococcus pneumoniae acquisition in young and elderly mice. Infect Immun. 2014;82:1725–1731.
27. Teo SM, Mok D, Pham K, et alThe infant nasopharyngeal microbiome impacts severity of lower respiratory infection and risk of asthma development. Cell Host Microbe. 2015;17:704–715.
28. Abdullahi O, Karani A, Tigoi CC, et alThe prevalence and risk factors for pneumococcal colonization of the nasopharynx among children
in Kilifi District, Kenya. PLoS One. 2012;7:e30787.
29. Wright AK, Ferreira DM, Gritzfeld JF, et alHuman nasal challenge with Streptococcus pneumoniae is immunising in the absence of carriage. PLoS Pathog. 2012;8:e1002622.
30. Mansbach JM, Hasegawa K, Henke DM, et alRespiratory syncytial virus and rhinovirus severe bronchiolitis are associated with distinct nasopharyngeal microbiota
. J Allergy Clin Immunol. 2016;137:1909–1913.e4.
31. de Steenhuijsen Piters WA, Heinonen S, Hasrat R, et alNasopharyngeal microbiota
, host transcriptome, and disease severity in children
with respiratory syncytial virus infection. Am J Respir Crit Care Med. 2016;194:1104–1115.
32. Leung RK, Zhou JW, Guan W, et alModulation of potential respiratory pathogens by pH1N1 viral infection. Clin Microbiol Infect. 2013;19:930–935.
33. Heinsbroek E, Tafatatha T, Chisambo C, et alPneumococcal acquisition among infants exposed to HIV in rural Malawi: a longitudinal household study. Am J Epidemiol. 2016;183:70–78.
34. Johanson WG Jr, Blackstock R, Pierce AK, et alThe role of bacterial antagonism in pneumococcal colonization of the human pharynx. J Lab Clin Med. 1970;75:946–952.