Three Part Question
In [an infant with clinical suspicion of pertussis], how [useful is a lymphocyte count] to [determine the likelihood of a positive or negative diagnosis]?
A 3-week-old baby presents with cough and episodes of apnoea. Nasopharyngeal aspirate is negative for common respiratory viruses. You consider the diagnosis of pertussis and take a full blood count to assess the lymphocyte count. You wonder what the sensitivity and specificity of the lymphocyte count is for pertussis in infants.
A Medline search was performed using the following terms: (MeSH: Lymphocytosis OR MeSH: Lymphocyte Count OR “lymphocyte” OR “lymphocytosis” OR MeSH: Leukocytosis OR MeSH: Leukocyte count OR “leucocyte” OR “leukocyte” OR “leucocytosis” OR “leukocytosis” OR “white cell count” OR “full blood count” OR “complete blood count”) AND (MeSH: Whooping Cough OR pertussis OR “whooping cough”). Limits were: “Humans” and “English Language”.
A search of The Cochrane Library using the search term “pertussis” yielded five reviews, none of which were relevant.
The search yielded 414 individual articles. Eleven full text articles were examined, of which four were relevant; one further article was found in the reference lists of included papers
|Author, date and country
||Study type (level of evidence)
|Guinto-Ocampo et al,|
|98 infants (≤12 months) tested for pertussis (culture, PCR or direct fluorescent antibody test)
18 patients were pertussis positive
2001–2005, USA ||Exploratory cohort with inconsistently applied reference standard (level 3b)||Diagnostic usefulness of lymphocyte count >9400/µl (Cut-off chosen by ROC analysis to maximise sensitivity and specificity)||Sensitivity 89% (95% CI 65% to 99%) Specificity 75% (95% CI 64% to 84%) PPV 44% (95% CI 28% to 61%) NPV 97% (95% CI 92% to 100%) ||All pertussis positive patients had a lymphocyte count measured, but only 65% of pertussis negative patients had a lymphocyte count measured |
|Lin et al,|
|36 children (0–19 years) with pertussis (culture, PCR or direct fluorescent antibody test)
1997–2001, Taiwan ||Prospective cohort (level 3b)||Diagnostic usefulness of lymphocyte count >10 000 cells/µl Mean lymphocyte count 13 000 cells/µl in children <6 months old and 6900 cells/µl in children >6 months old ||Sensitivity 44% (95% CI 28% to 62%) Significant difference, p=0.008 ||Cohort of cases only, no controls, so therefore cannot calculate specificity or predictive values. Only 36 of the total of 139 patients in the study had a lymphocyte count measured |
|Heininger et al,|
|482 unimmunised children (0–16 years) with cough of ≥7 days duration and subsequent culture positive for pertussis
1990–1996, Germany ||Prospective cohort (level 3b)||Diagnostic usefulness of lymphocyte count >95% of the upper limit according to age 38.7% of children <6 months old had lymphocyte count >95% of the upper limit of normal vs 34.6% of children >6 months old ||Sensitivity 35% (95% CI 31% to 39%) Not a significant difference ||Full data on pertussis negative children not given, so therefore cannot calculate specificity or predictive values. Only 482 of the total of 2592 patients in the study had a lymphocyte count measured |
|Gordon et al,|
|173 children (0–17 years) with pertussis (culture or direct fluorescent antibody test)
1980–1990, Canada ||Retrospective cohort (level 3b)||Diagnostic usefulness of lymphocyte count >10×109/l in children <6 months old Diagnostic usefulness of lymphocyte count >10×109/l in children >6 months old 56% of children <6 months old had lymphocyte count >10×109/l vs 46% of children >6 months old ||Sensitivity 56% (95% CI 46% to 65%) Sensitivity 46% (95% CI 34% to 58%) Not a significant difference ||Cohort of cases only, no controls, so therefore cannot calculate specificity or predictive values. 173 of the total of 975 patients in the study had a lymphocyte count measured |
|Strebel et al,|
|88 children (<5 years) with specified clinical suspicion cultured for pertussis
33 patients were pertussis positive ||Prospective cohort with inconsistently applied reference standard (level 3b)||Diagnostic usefulness of lymphocyte count >10 000/mm3||Sensitivity 88% (95% CI 69% to 97%) Specificity 57% (95% CI 37% to 75%) PPV 63% (95% CI 45% to 79%) NPV 85% (95% CI 69% to 100% ||76% of pertussis positive patients had a lymphocyte count measured, but only 55% of pertussis negative patients had a lymphocyte count measured |
There are no good quality studies addressing the diagnostic strength of lymphocytosis as a marker for pertussis in infants. The sensitivity of lymphocytosis ranged from 39% to 89% over the five studies (Guinto-Ocampo, Lin, Heininger, Gordon, Strebel).Usually, poorer quality studies show inflated sensitivity. However, in this case the higher quality studies, comparing children with and without pertussis, showed higher sensitivities (88–89%)(Guinto-Ocampo, Strebel).than lower quality studies looking only at children with pertussis (39–56%) (Lin, Heininger, Gordon). Specificity was fairly low throughout (57–75%)(Guinto-Ocampo, Lin, Heininger, Gordon, Strebel).Overall, the data suggest that a lack of lymphocytosis means that pertussis is an unlikely diagnosis, but the presence of lymphocytosis is not very specific.
There are several important limitations to the review. There is no gold standard for pertussis diagnosis, with most studies using culture, direct fluorescent antibody test and PCR, and diagnosing pertussis if any one of these are positive. Two older studies used culture only (Heininger, Strebel).which is likely to underestimate pertussis. It is interesting to note the UK Health Protection Agency guidelines (Amirthalingam)which state that PCR should be used for hospitalised infants (up to and including 1 year of age). This test is offered free of charge in England and Wales, with a same day result for samples received by 10:00 h. Either nasopharyngeal aspirates or pernasal swabs can be used. In contrast, pertussis culture is recommended for infants not requiring hospitalisation, and in children over 12 months of age.
The studies are heterogeneous with regard to their populations. Patient characteristics were poorly described, but some studies clearly focus on a community setting (Heininger) while others are hospital based (Lin, Gordon).The ethnicity of participants was not described, although it is known to affect some white cell lines (Haddy) and so may be relevant.
In all studies, assessment of the lymphocyte count may have affected the decision to test for pertussis, which would artificially increase the sensitivity and decrease the specificity. There is evidence to support this hypothesis as more children with a pertussis diagnosis had a lymphocyte count measured than those with a negative diagnosis (Guinto-Ocampo, Strebel).
The definition of lymphocytosis varied from 9400 to 13500 cells/µl for infants(Guinto-Ocampo, Heininger)although the most common cut-off was 10 000 cells/µl (Lin, Gordon, Strebel). The one study that explored which cut-off value maximised sensitivity and specificity (in infants) found that 9400 cells/µl was optimal (Guinto-Ocampo).The clinical question concerned a child under 6 months of age. Three studies compared the lymphocyte count in infants under 6 months of age with those over 6 months of age (Lin, Heininger, Gordon).Two found no difference in the percentage of pertussis positive children with a count above the normal range (Heininger, Gordon)and one found that the mean lymphocyte count was higher in the younger age group (Lin). Therefore, measuring lymphocyte count has the same utility in infants as in older children.
NPV, negative predicted value; PPV, positive predictive value; ROC, receiver operating characteristic.
Clinical Bottom Line
A child with a normal lymphocyte count is unlikely to have pertussis. (Grade B)
Approximately half of children with suspected pertussis and a raised lymphocyte count will not have pertussis. (Grade B)
There is no apparent effect of age on the utility of a lymphocyte count in the diagnosis of pertussis. (Grade B)
- Whiting P, Rutjes AW, Reitsma JB, et al . The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25.
- Review Manager (RevMan) [Computer program]. Version 5.0. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2008.
- Guinto-Ocampo H, Bennett JE, Attia MW . Predicting pertussis in infants. Pediatr Emerg Care 2008;24:16–20.
- Lin PY, Chiu CH, Wang YH, et al . Bordetella pertussis infection in northern Taiwan, 1997-2001. J Microbiol Immunol Infect 2004;37:288–94.
- Heininger U, Klich K, Stehr K, et al . Clinical findings in Bordetella pertussis infections: results of a prospective multicenter surveillance study. Pediatrics 1997;100:E10.
- Gordon M, Davies HD, Gold R . Clinical and microbiologic features of children presenting with pertussis to a Canadian pediatric hospital during an eleven-year period. Pediatr Infect Dis J 1994;13:617–22.
- Strebel PM, Cochi SL, Farizo KM, et al . Pertussis in Missouri: evaluation of nasopharyngeal culture, direct fluorescent antibody testing, and clinical case definitions in the diagnosis of pertussis. Clin Infect Dis 1993;16:276–85.
- Amirthalingam G the Pertussis Guidelines Group. HPA Guidelines for the Public Health Management of Pertussis. Health Protection Agency, London, 2011.
- Haddy TB, Rana SR, Castro O . Benign ethnic neutropenia: what is a normal absolute neutrophil count? J Lab Clin Med 1999;133:15–22.