Skip Navigation

Symptomatic Predictors of Influenza Virus Positivity in Children during the Influenza Season

  1. Suzanne E. Ohmit and
  2. Arnold S. Monto
  1. University of Michigan School of Public Health, Ann Arbor, Michigan
  1. Reprints or correspondence: Dr. Arnold S. Monto, Dept. of Epidemiology, University of Michigan School of Public Health, 109 Observatory St., Ann Arbor, MI 48109-2029 (asmonto{at}umich.edu).
 
Next Section

Abstract

Background. Symptomatic predictors of the etiology of infectious diseases are necessary when quick action is required in treatment, as with cases of influenza or anthrax, or for when patient isolation is required, as with cases of severe acute respiratory syndrome (SARS). Predictors of influenza virus infection during the influenza season have previously been evaluated in adult studies of the antiviral agent zanamivir; cough and temperature α37.8°C predicted influenza virus positivity in 79% of those evaluated.

Methods. Fever and other respiratory symptoms were examined to determine their value in predicting influenza virus–positive status in children. Data analyzed were from a clinical trial of zanamivir in children 5–12 years of age and from a trial of oseltamivir in children 1–12 years of age.

Results. In the pediatric study of zanamivir, as in the adult zanamivir study, cough and fever were the best predictors of influenza virus infection; a temperature α38.2°C plus cough predicted 83% (95% CI, 79%–88%) of illnesses that were determined to be influenza virus positive. Cough (positive predictive value, 70%; 95% CI, 64%–75%), but not fever, was the best predictor of influenza virus–positive status in children aged 5–12 years in the oseltamivir trials, but neither cough nor fever were successful predictors in young children 1–4 years of age. The latter findings appeared to be the result of less symptomatic diversity among patients recruited for this trial, such that subjects who were determined to be influenza virus positive and subjects who were influenza virus negative had similar symptoms.

Conclusions. The results of these studies suggest that, during the influenza season, symptomatic predictors of influenza virus infection are applicable to identification of cases in children, although confirmation of predictive values in subjects 1–4 years of age may require further study of additional signs/symptoms.

Symptomatic predictors of the etiology of an infectious illness are particularly valuable when actions need to be taken quickly, but when a specific diagnosis cannot be made rapidly. Such situations vary historically, from severe acute respiratory syndrome (SARS), in which isolation and quarantine measures needed to be implemented quickly to render control effective, to influenza, in which specific antiviral treatment must begin within 48 h of onset of illness for maximum effectiveness [1, 2]. Symptomatic predictors of infectious illness may also be helpful in distinguishing between illnesses with different implications, as was the case during the inhalational anthrax episode in 2001, when comparison with influenza symptoms was examined [3]. Symptomatic predictors of influenza have been examined in a number of situations, typically involving surveillance activities, practice settings, or clinical trials. Although, in early studies, there was pessimism with regard to the ability of the use of symptoms as indicators of virus positivity, recent data suggest the opposite [4–6]. The difference in study conclusions arises from the realization that the use of symptomatic predictors needs to be limited to periods of known influenza virus circulation, thereby tying laboratory surveillance to illness detection. In the treatment trials of the neuraminidase inhibitors (zanamivir and oseltamivir), symptoms were recorded for, and specimens for virus identification were collected from, individuals with influenza-like illnesses, before the individuals were randomized to drug or placebo groups [7–10]. Data from these trials provided a systematic method of relating symptoms that are present early in an illness to confirmation of influenza etiology. In a previous report, the presence of cough and fever in adults who were recruited in studies of zanamivir was found to accurately predict 79% of influenza virus–positive cases [6]. This article extends those observations to similar pediatric studies, of either zanamivir or oseltamivir, that involve children <13 years of age.

Previous SectionNext Section

Methods

Source of data and recruitment criteria. Data from single pediatric clinical trials of zanamivir and oseltamivir were provided by GlaxoSmithKline and Roche, respectively. For the purpose of this analysis, it was only necessary to have data on subjects' signs and symptoms at the time of subject recruitment, and not subjects' subsequent assignments to treatment or placebo groups [9, 10]. Only those participants for whom complete data on variables of interest were collected consistently across the studies were included in this analysis. Items of interest included virus identification results, data on fever, and the presence and severity of the following symptoms: cough, nasal symptoms/congestion, sore throat, myalgia, headache, fatigue, feverishness, and loss of appetite. Inclusion and exclusion criteria for each trial varied; however, in both studies, documented transmission of influenza virus at each study site was required before recruitment could begin. This was typically defined as 2 local influenza virus identifications in a particular week, either in a surveillance program or from a local health department.

In the single pediatric zanamivir study, only fever (temperature, 37.8°C) and duration of illness of 36 h was required for inclusion in the study, regardless of other specific symptoms. Complete data from this analysis were available for 468 participants, aged 5–12 years [9].

In the single oseltamivir pediatric study, fever (temperature, 37.8°C) and 1 respiratory symptom (cough or nasal symptoms) were required for inclusion into the study [10]. Illness had to be 48 h in duration, and all participants were screened for respiratory syncytial virus infection with a rapid test and were excluded if infection was present. There were 476 records available for analysis of children 1–12 years of age; 221 children were aged 1–4 years, and 255 were 5–12 years of age.

Identification of virus positivity. Methods used to identify those infected with influenza virus varied somewhat between the 2 studies. In both cases, a subject was considered to be positive for influenza virus infection if there was isolation of an influenza virus from sample culture and/or a 4-fold increase in influenza antibody titer between blood specimens collected at the time of recruitment and 3 weeks thereafter. In addition, in the zanamivir trial, specimens for virus identification were tested by standard RT-PCR techniques [11]; those specimens that were found to be positive by RT-PCR were considered to be influenza virus positive.

Analytic procedures. Fever was recorded by degree, and symptoms were reported on a 4-point scale of severity (none, mild, moderate, or severe). Logistic regression was performed to determine the independent contribution of fever and each symptom in prediction of influenza virus–positive status. In addition, each symptom and combination of symptoms deemed to be of importance in logistic regression was analyzed, independently or in combination, to determine sensitivity and positive predictive value (PPV). This was initially done for various levels of fever and either no symptoms versus any symptoms or no and mild symptoms versus moderate and severe symptoms. PPV was calculated as the probability of a laboratory-confirmed influenza virus infection when a symptom alone or in combination was present; sensitivity of a symptom was defined as the probability that an individual with laboratory-confirmed influenza virus positivity has that symptom. Analyses were carried out using the SAS statistical software, version 8.2 (SAS Institute).

Previous SectionNext Section

Results

Pediatric studies in older children: zanamivir study. The study that evaluated zanamivir was conducted among children aged 5–12 years [9]. Respiratory illness with a temperature 37.8°C was the only criterion required for inclusion into the study; as in other studies, recruitment took place when influenza virus transmission was documented. Using only clinical judgment to determine eligibility, 74% of participants tested positive for influenza virus infection; 35% of influenza virus–positive cases involved influenza type B. In a logistic regression model (table 1), a temperature of 38.2°C and cough independently predicted (P < .001) influenza virus positivity; the PPV of combined fever and cough was 83% (95% CI, 79%–88%). Sore throat was a significant negative predictor (P = .001) as was myalgia (P = .047). In a separate model of a subset of 399 children with available WBC counts, those with low WBC counts (<4500 cells/mm3) had significantly increased odds of influenza virus positivity (OR, 4.32; 95% CI, 2.21–8.44; P < .001), compared with those with normal or high WBC counts (PPV, 89%; 95% CI, 83%–94%). The PPV of combined fever, cough, and low WBC count was 94% (95% CI, 89%–99%); however, this combination of signs/symptoms was infrequently present, with a sensitivity of only 27%.

Table 1
View larger version:
Table 1

Logistic regression analysis of symptomatic predictors of laboratory-confirmed influenza virus–positive status among 468 children aged 5–12 years recruited to a multicenter clinical trial of the influenza antiviral agent zanamivir.

Pediatric studies in children: oseltamivir study. The study of treatment with oseltamivir included children aged 1–12 years (mean age, 5.4 years). To be eligible for recruitment, a subject had to demonstrate fever (temperature, 37.8°C) and either cough or nasal symptoms of 48 h in duration. Influenza virus circulation (almost exclusively type A) in the area was confirmed, and each eligible subject had a negative rapid test result for respiratory syncytial virus [10].

For comparison to the zanamivir pediatric study, data were considered separately for children 5–12 years of age and children aged <5 years. In the group aged 5–12 years (n = 255), 66% of subjects recruited tested positive for influenza. In the logistic regression analysis (table 2), cough (P < .001) and headache (P = .017), but not fever (P = .32), were significant independent predictors of influenza virus–positive status. The PPV for combined cough and headache (73%; 95% CI, 67%–80%) was not significantly improved by the addition of fever (PPV, 74%; 95% CI, 68%–81%). The PPV for combined cough and fever only was 71% (95% CI, 65%–77%).

Table 2
View larger version:
Table 2

Logistic regression analysis of symptomatic predictors of laboratory-confirmed influenza virus–positive status among 255 children aged 5–12 years recruited to a multicenter clinical trial of the influenza antiviral agent oseltamivir.

There has always been a view that symptomatic predictors might not be of great value in younger children. However, with the simple inclusion criteria that were used in the original trial, 67% of the 221 children aged <5 years were determined to be influenza virus positive. In the logistic regression model (table 3), myalgia (P = .010) was the only symptom to independently predict influenza virus–positive status; presence of fever or cough did not. The PPV of myalgia alone (PPV, 73%; 95% CI, 66%–81%) was not improved by the additions of fever and cough (PPV, 72%; 95% CI, 64%–81%), and both provided only modest increases in the PPV beyond that produced by the inclusion criteria. The PPV for combined cough and fever only was 64% (95% CI, 58%–71%). With the exception of myalgia (none or mild vs. moderate or severe), there was little difference in the prevalence of symptoms between those children who tested positive for influenza virus and those who tested negative for influenza virus (data not shown).

Table 3
View larger version:
Table 3

Logistic regression analysis of symptomatic predictors of laboratory-confirmed influenza virus–positive status among 221 children aged 1–4 years recruited to a multicenter clinical trial of the influenza antiviral agent oseltamivir.

Previous SectionNext Section

Discussion

It was previously believed that it was impossible to use symptomatic predictors to identify which respiratory infections were actually caused by influenza viruses, because the signs and symptoms of many illnesses were too similar. To study this belief, data from surveillance studies were initially examined to determine whether individuals from whom influenza viruses were isolated had different characteristics of illness than those who were not infected by influenza viruses. A number of signs and symptoms were significantly associated with influenza virus positivity. Fever and cough were the variables most consistently identified; other symptoms, such as “red eyes,” were less frequently reported [4]. However, the value of these observations to predict infection positivity was relatively low (∼40%), a result of the fact that the surveillance from which the data were obtained was conducted over long periods, during much of which time no influenza viruses were identified [5]. This is reflected in the relationship of prevalence to PPV, at fixed sensitivity and specificity.

The aim in the above mentioned clinical trials of neuraminidase inhibitors was to recruit the largest numbers of subjects with true influenza virus illnesses. Therefore, enrollment was limited to periods of documented influenza virus transmission. This basic requirement, common to all studies, resulted in an overall frequency of influenza virus positivity ranging from 64% to 74% [7–10]. This was the case in spite of the use of different inclusion and exclusion criteria in the various studies. The question, then, is whether it is possible to identify symptomatic predictors that would further improve on the ability to identify influenza virus illnesses. In the previous analysis of the zanamivir adult study, the PPV increased to 79% when a temperature 37.8°C and cough of any level were used as predictors [6]. There was little advantage to measuring other symptoms. Fever and cough were important predictors in this study because there was diversity in those cases included; for example, some component trials required only reported feverishness with inclusion of individuals who were afebrile at recruitment.

Similarly, in the zanamivir children's study, the requirement of documented influenza virus transmission with few other inclusion criteria also resulted in the recruitment of a diverse group of cases. The requirement of a temperature 38.2°C and cough resulted in a PPV of 83% (table 4). This indicates that symptomatic predictors of influenza virus infection are as valid in older children as in adults. The oseltamivir studies presented a different situation, because the included cases were less diverse because of the requirement of at least 1 respiratory symptom (cough or nasal symptoms) for inclusion. In the older children, cough and a temperature 38.2°C resulted in a PPV of 71%, a small increase from the overall PPV of 66%; however, with the addition of headache, the PPV increased somewhat to 74%. This was more extreme in younger children, in which the PPV of combined fever and cough (64%) was less than that demonstrated with the inclusion criteria of 67%; in the younger children, PPV increased surprisingly with the inclusion of the symptom myalgia (table 4). Interestingly, results from identical analyses (not shown) of data from trials of oseltamivir in adults [7, 8] demonstrated a PPV of 69% (95% CI, 65%–73%) for combined fever and cough. This was slightly higher than the 64% who tested positive for influenza virus achieved by the inclusion criteria. This strengthens the conclusion that the difference between the results from the zanamivir and oseltamivir studies in older children is, at least in part, a result of differences in symptomatic diversity of the included populations.

Table 4
View larger version:
Table 4

Summary of positive predictive values (PPV) for fever, cough, and fever plus cough across studies, as well as single and combination signs/symptoms shown to be important in analysis-specific logistic models, from trials of the influenza antiviral agents zanamivir and oseltamivir, in children <13 years.

Taken together, it can be concluded that it is possible to predict influenza virus infection on the basis of clinical presentation. A limitation may be that the data analyzed came from clinical trials and not from population-based surveillance. However, this conclusion is sustained by a study conducted as part of surveillance activities, in which the PPV of cough and fever in combination was 87% [12]. It appears that this observation extends to children 5 years. Although it is not clear whether this applies to younger children, in the oseltamivir clinical trial, 65% of those recruited were considered to be influenza virus positive on the basis of inclusion criteria of fever plus at least 1 other symptom (cough or nasal symptoms). One other study has also suggested that symptomatic predictors of influenza virus infection are applicable to young children [13].

Rapid diagnostic tests have varied greatly in their ability to detect influenza virus infections; in a recent population-based study, test sensitivity was found to be only 41% [14–18]. Therefore, these tests may be of little added value in deciding whom to treat when influenza viruses are known to be circulating. Also, because the PPV of the tests decreases when prevalence of influenza decreases, rapid diagnostic tests will be less useful during periods of sporadic influenza virus occurrence. However, rapid tests may be more sensitive in the very young, who are known to shed more virus [19]; a study evaluating rapid testing in parallel with standard virology would need to be performed for a definitive conclusion in that population. There are tests of high sensitivity and specificity, particularly RT-PCR assays, that are becoming generally available and are essential in special situations, such as when a patient of any age is hospitalized. Rapid diagnostic tests also serve to identify other respiratory agents that might be involved and will thus be little affected by seasonality considerations [20]. However, within the defined influenza virus season and in most population groups, clinical characteristics seem sufficient to identify those cases to be considered for antiviral therapy.

Previous SectionNext Section

Acknowledgments

Acknowledgments

Financial support. State of Michigan Department of Community Health (20050854).

Potential conflicts of interest. A.S.M. has received research support and has been an ad hoc consultant to Roche and GlaxoSmithKline. S.E.O.: no conflicts.

  • Received January 27, 2006.
  • Accepted May 28, 2006.
Previous Section
 

References

  1. ↵
    1. Scales DC,
    2. Green K,
    3. Chan AK,
    4. et al
    . Illness in intensive care staff after brief exposure to severe acute respiratory syndrome. Emerg Infect Dis 2003;9:1205-10.
    MedlineWeb of Science
  2. ↵
    1. Call SA,
    2. Vollenweider MA,
    3. Hornung CA,
    4. Simel DL,
    5. McKinney WP
    . Does this patient have influenza? JAMA 2005;293:987-97.
    Abstract/FREE Full Text
  3. ↵
    1. Fine AM,
    2. Wong JB,
    3. Fraser HS,
    4. Fleisher GR,
    5. Mandl KD
    . Is it influenza or anthrax?. A decision analytic approach to the treatment of patients with influenza-like illnesses. Ann Emerg Med 2004;43:318-28.
    CrossRefMedlineWeb of Science
  4. ↵
    1. Brown LE,
    2. Hampson AW,
    3. Webster RG
    1. Monto AS,
    2. Ohmit SE
    . The evolving epidemiology of influenza infection and disease. In: Brown LE, Hampson AW, Webster RG, editors. Options for the control of influenza III. Amsterdam, The Netherlands: Elsevier Science BV; 1996. p. 45-9.
  5. ↵
    1. Carrat F,
    2. Tachet A,
    3. Rouzioux C,
    4. Housset B,
    5. Valleron AJ
    . Evaluation of clinical case definitions of influenza: detailed investigation of patients during the 1995–1996 epidemic in France. Clin Infect Dis 1999;28:283-90.
    Abstract/FREE Full Text
  6. ↵
    1. Monto AS,
    2. Gravenstein S,
    3. Elliott M,
    4. Colopy M,
    5. Schweinle J
    . Clinical signs and symptoms predicting influenza infection. Arch Intern Med 2000;160:3243-7.
    Abstract/FREE Full Text
  7. ↵
    1. Treanor JJ,
    2. Hayden FG,
    3. Vrooman PS,
    4. et al
    . Efficacy and safety of the oral neuraminidase inhibitor oseltamivir in treating acute influenza: a randomized controlled trial. JAMA 2000;283:1016-24.
    Abstract/FREE Full Text
  8. ↵
    1. Nicholson KG,
    2. Aoki FY,
    3. Osterhaus ADME,
    4. et al
    . Efficacy and safety of oseltamivir in treatment of acute influenza: a randomized controlled trial. Lancet 2000;355:1845-50.
    CrossRefMedlineWeb of Science
  9. ↵
    1. Hedrick JA,
    2. Barzilai A,
    3. Behre U,
    4. et al
    . Zanamivir for treatment of symptomatic influenza A and B infection in children five to twelve years of age: a randomized controlled trial. Pediatr Infect Dis J 2000;19:410-17.
    CrossRefMedlineWeb of Science
  10. ↵
    1. Whitley RJ,
    2. Hayden FG,
    3. Reisinger KS,
    4. et al
    . Oral oseltamivir treatment of influenza in children. Pediatr Infect Dis J 2001;20:127-33.
    MedlineWeb of Science
  11. ↵
    1. Claas EC,
    2. van Milaan AJ,
    3. Sprenger MJ,
    4. et al
    . Prospective application of reverse transcriptase polymerase chain reaction for diagnosing influenza infections in respiratory samples from a children's hospital. J Clin Microbiol 1993;31:2218-21.
    Abstract/FREE Full Text
  12. ↵
    1. Boivin G,
    2. Hardy I,
    3. Tellier G,
    4. Maziade J
    . Predicting influenza infections during epidemics with use of a clinical case definition. Clin Infect Dis 2000;31:1166-9.
    Abstract/FREE Full Text
  13. ↵
    1. Friedman MJ,
    2. Attia MW
    . Clinical predictors of influenza in children. Arch Pediatr Adolesc Med 2004;158:391-4.
    Abstract/FREE Full Text
  14. ↵
    1. Hamilton MS,
    2. Abel DM,
    3. Ballam YJ,
    4. et al
    . Clinical evaluation of the ZstatFlu-II test: A chemiluminescent rapid diagnostic test for influenza virus. J Clin Microbiol 2002;40:2331-4.
    Abstract/FREE Full Text
    1. Chan KH,
    2. Maldeis N,
    3. Pope W,
    4. et al
    . Evaluation of the Directigen Flu A+B test for rapid diagnosis of influenza virus type A and B infections. J Clin Microbiol 2002;40:1675-80.
    Abstract/FREE Full Text
    1. Schultze D,
    2. Thomas Y,
    3. Wunderli W
    . Evaluation of an optical immunoassay for the rapid detection of influenza A and B viral antigens. Eur J Clin Microbiol Infect Dis 2001;20:280-3.
    CrossRefMedlineWeb of Science
    1. Boivin G,
    2. Hardy I,
    3. Kress A
    . Evaluation of a rapid optical immunoassay for influenza viruses (FLU OIA test) in comparison with cell culture and reverse transcription-PCR. J Clin Microbiol 2001;39:730-2.
    Abstract/FREE Full Text
  18. ↵
    1. Effler PV,
    2. Ieong M-C,
    3. Tom T,
    4. Nakata M
    . Enhancing public health surveillance for influenza virus by incorporating newly available rapid diagnostic tests. Emerg Infect Dis 2002;8:23-8.
    MedlineWeb of Science
  19. ↵
    1. Uyeki TM
    . Influenza diagnosis and treatment in children: a review of studies on clinically useful tests and antiviral treatment for influenza. Pediatr Infect Dis J 2003;22:164-77.
    CrossRefMedlineWeb of Science
  20. ↵
    1. Weinberg GA,
    2. Erdman DD,
    3. Edwards KM,
    4. et al
    . Superiority of reverse-transcription polymerase chain reaction to conventional viral culture in the diagnosis of acute respiratory tract infections in children. J Infect Dis 2004;189:706-10.
    Abstract/FREE Full Text
| Table of Contents

This Article

  1. Clin Infect Dis. (2006) 43 (5): 564-568. doi: 10.1086/506352
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 54 (5)
  1. Clinical Infectious Diseases
  1. Alert me to new issues

Most