Surveillance of measles in England and Wales: implications of a national saliva testing programme.

Research/Recherche Surveillance of measles in England and Wales: implications of a national saliva testing programme M. Ramsay,' R. Brugha,2 & D. Brown3 The impact of a mass vaccination campaign against measles-rubella in England and Wales was assessed using the results of a saliva test for measles-specific IgM, which was offered to all notified cases of measles. By means of clinical data supplied by the reporting doctors, we estimated the sensitivity and specificity of various clinical case definitions in predicting a confirmed case. A saliva sample was obtained within the appropriate time period for 3442/7574 (45.4%) of notified cases; the proportion confirmed by saliva testing was low and fell from 67/681 (9.8%) at the start of the campaign to 1/373 (0.3%) after 35 weeks. The specificity of all clinical case definitions was low, but was highest at 41.1% (95% confidence interval, 39.2- 43.0%°) for the definition similar to that recommended in the USA. Extrapolating from cases observed during ongoing surveillance permits estimation of an annual notification rate for nonmeasles rash and fever illness. For countries with good measles control, a clinical case definition is too nonspecific for accurate surveillance. Laboratory confirmation of suspected measles cases is required and saliva testing is an acceptable method. The continued sensitivity of measles surveillance in England and Wales requires that the incidence of notified rash and fever illness and the proportion of cases in which measles infection was excluded by laboratory testing be monitored. Introduction An effective surveillance system is required to ensure the adequate control and the eventual elimi- nation of a vaccine-preventable disease (1). In No- vember 1994 a mass campaign was conducted in England and Wales, whereby all schoolchildren aged 5-16 years were offered combined measles-rubella vaccine (2, 3). Prior to the campaign, surveillance of measles in England and Wales was based upon the statutory notification of clinically diagnosed cases of measles to the Office of National Statistics (ONS). Over the period 1991-94, a pilot study of notified measles indicated that less than 40% of suspected cases could be confirmed by laboratory testing and I Consultant Epidemiologist, Immunisation Division, PHLS Com- municable Disease Surveillance Centre, 61 Colindale Avenue, London NW9 5EQ, England. Correspondence should be sent to this author. 2 Senior Registrar, Immunisation Division, PHLS Communicable Disease Surveillance Centre, London, England. 3 Consultant Virologist, PHLS Enteric and Respiratory Virus Labo- ratory, London, England. Reprint No. 5807 that a saliva assay was over 90% specific and sensi- tive for the detection of measles IgM (4). Since the start of the campaign, therefore, the Public Health Laboratory Service (PHLS) has enhanced the sur- veillance of measles by offering the saliva test for all cases of suspected measles notified via the statutory system (4). Confirmed cases of measles detected through this scheme are defined as those cases with salivary or serological evidence of recent infection (the presence of measles-specific IgM in the absence of measles vaccination in the previous 6 weeks) (5). Cases ascertained by other means include those with evidence of recent infection, as reported by laborato- ries (positive measles serum IgM test, a fourfold rise in serum IgG or isolation of measles virus) in the absence of recent vaccination, and cases with rash and fever which are epidemiologically linked (a his- tory of contact in the 2 weeks prior to onset) to a confirmed case. At the start of the campaign, clinical case defini- tions were issued to encourage the prompt notifica- tion of suspected measles and to guide the diagnosis. Rash and fever in a child was defined as a suspect case, whereas fever of at least 38.5°C, a macular rash, and at least one of cough, coryza or conjunc- Bulletin of the World Health Organization, 1997, 75 (6): 515-521 © World Health Organization 1997 515 M. Ramsay et al. tivitis was defined as a probable case. Similar clinical guidelines have been recommended by WHO (6, 7) and used for measles surveillance in the USA (8) and Canada (9). This article describes the results obtained from the first 35 weeks of the enhanced measles surveillance programme in England and Wales and reviews the potential usefulness of clinical case definitions for measles. Criteria and standards for the evaluation of measles surveillance systems in countries near to eliminating this disease are proposed. Methods Because notifications of measles are initially re- ceived by consultants in communicable disease con- trol in each health district, these individuals were asked to inform local general practitioners of the importance and availability of salivary testing and to request the prompt notification of any suspected case of measles. The consultants were supplied with a stock of kits for the collection of saliva specimens, each containing instructions, a request form, a saliva swab, packaging and a reply-paid box. On receipt of a notification of measles the consultants sent a kit to the notifying doctor, asking for a saliva sample to be collected 1-6 weeks after the onset of illness. Saliva samples were taken by the doctor, nurse, parent or patient by wiping a specially designed sponge swab around the patient's gum margin (10). The notifying doctor was asked to complete a brief questionnaire, reporting the presence or absence of each of five clinical features (rash, fever, conjunctivitis, cough or coryza) and the duration in days of any rash. Samples were sent for analysis by first-class post to the PHLS Enteric and Respiratory Virus Laboratory (ERVL), and the results were returned by post to the notifying doctors, with copies to the appropriate consultant in communicable disease control. Positive cases were reported by phone or fax to the consultant con- cerned. The results were collated at the PHLS Com- municable Disease Surveillance Centre (CDSC). Aggregated results were published in the Communi- cable Disease Report and regular newsletters gave feedback on the progress and outcome of the surveil- lance programme. Saliva specimens were tested for measles-spe- cific IgM by antibody-capture radioimmunoassay. This assay, based on semipurified, tissue-culture- grown virus and a measles-specific monoclonal antibody, has been reported previously to be both sensitive and specific in community based surveys provided samples are collected within 6 weeks of onset (4). Using clinical information supplied by the re- porting doctors, we estimated the sensitivity and specificity of clinical case definitions in predicting IgM-positive cases for cases where the sample was taken during the appropriate time period. The case definitions in England include a suspect case (with rash and fever) and a probable case (with rash, fever :38.5 °C, and one of cough, coryza or conjunctivitis) (5). The latter definition is similar to that currently used by WHO (7). A similar definition, previously recommended by Strassburg (6) and used in the USA and Canada (8, 9), requires the presence of fever -38°C, rash for -3 days plus one of cough, coryza or conjunctivitis. Since a maximum tempera- ture was not available from the questionnaire and because only notified cases were followed by a saliva test, the following case definitions were examined: - notified cases of measles with rash and fever; - notified cases of measles with rash, fever plus either cough, coryza or conjunctivitis; and - notified cases of measles with rash for -3 days, fever plus either cough, coryza or conjunctivitis. IgM-positive cases with a history of vaccination in the 6 weeks prior to onset of illness and cases where neither the presence nor absence of a symp- tom or sign was reported were excluded from the analysis. Because information on contacts was not routinely collected, cases that would fulfil the defini- tion of an epidemiologically linked case were not examined separately. To assess the potential for true measles cases to be misdiagnosed and therefore not reported to the statutory system, we performed measles-specific IgM assays on saliva samples from 108 notified cases of rubella with a rash for -3 days, fever, and either cough, coryza or conjunctivitis. These cases were notified as rubella during the period from week 44 of 1994 (94/44) to week 16 of 1995 (95/16); these sam- ples were obtained in exactly the same way as those for notified measles cases. Results In the first 35 weeks of the enhanced surveillance programme (between weeks 94/44 and 95/26) a pro- visional total of 7574 cases of suspected measles were notified to ONS (Table 1). Saliva samples were re- ceived and tested from 4253 (56%) notified cases, and for 3442 (45%) the sample was taken within the appropriate time period (1-6 weeks) after the onset of illness. The number of notifications declined over the survey period but the proportion of cases tested remained constant. Based upon the corrected totals WHO Bulletin OMS. Vol 75 1997516 Surveillance of measles in England and Wales Table 1: Numbers of notified and confirmed measles cases in the enhanced surveillance programme, by 5-week periods, 1994-95 No. saliva IgM-positive: No. of appropriate saliva Perioda No. of notifications samples obtained Total IgM-positive Recently vaccinated Confirmed cases 94/44 to 94/48 1 859 681 (36.6)b 79 12 67 (9.8) 94/49 to 95/01 1 205 568 (47.1) 39 3 36 (6.3) 95/02 to 95/06 1 067 506 (47.4) 14 3 11 (2.2) 95/07 to 95/11 1 037 525 (50.6) 10 3 7 (1.3) 95/12 to 95/16 920 436 (47.4) 4 2 2 (0.5) 95/17 to 95/21 696 353 (50.7) 2 0 2 (0.6) 95/22 to 95/26 790 373 (47.2) 4 3 1 (0.3) Total 7574 3442 (45.4) 152 26 126 (3.7) a The first and last dates of each 5-week period are abbreviated by the last two digits of the year and the week number (e.g. 94/44 indicates the 44th week of 1994). b Figures in parentheses are percentages. of notifications available by age, similar proportions of notified cases were tested from each age group (Table 2). Specimens were received at ERVL within 1 week of sampling in 2501 (73%) cases and results were dispatched to the reporting doctor within 2 weeks of sampling in 2390 (69%) and within 3 weeks in 3162 (92%) cases. Measles-specific IgM was detected in 152 (4.4%) cases from whom samples had been taken within the appropriate time period. A total of 26 of these cases had received measles vaccine in the 6 weeks before the sample had been taken, leaving 126 cases fulfilling the definition for a confirmed case (72 males and 54 females). The number and proportion of confirmed cases declined rapidly over the 35-week period following the start of the campaign (Table 1), and a lower proportion of cases was confirmed among under-10-year-olds compared with older chil- dren and adults (Table 2). The sensitivity of the clinical case definitions for predicting a confirmed case was high for all defini- tions and highest for suspect cases (fever and rash) (Table 3). Specificity was low for all potential clinical definitions, but was highest for the definition based on that recommended in the USA and Canada. Be- cause of incomplete data on the questionnaire, how- ever, application of the more specific definitions resulted in more cases being excluded from the data set. If any of the stated case definitions had been used to select cases for saliva testing, a substantial proportion of tests could have been avoided. Use of the suspect case definition (rash and fever), would have resulted in 510/3154 (16%) tests being avoided and only 4/124 (3.2%) confirmed cases would have been missed. Use of the probable case definition would have resulted in 832/3003 (28%) cases not being tested, and 7/123 (5.7%) confirmed cases Table 2: Numbers of notified and confirmed measles cases, by age, in the enhanced surveillance programme, 1994-95 Saliva results: No. of appropriate saliva Age group No. of notifications samples obtained Total IgM-positive Recently vaccinated Confirmed cases <1 year 1656 862 (52.1)8 14 0 14 (1.6) 1-4 years 3042 1479 (48.6) 34 12 22 (1.5) 5-9 years 1 086 497 (45.8) 19 4 15 (3.0) 10-14 years 473 169 (35.7) 45 7 38 (23.5) 315 years 1 010 406 (40.2) 40 3 37 (9.2) Not known 140 29 (20.7) 0 0 0 (0) Total 7407b 3442 (46.5) 152 26 126 (3.7) a Figures in parentheses are percentages. b The discrepancy in the number of notified cases between Tables 1 and 2 is because the weekly totals represent provisional data (Table 1) whereas the age breakdown is only available for corrected data on a quarterly basis. WHO Bulletin OMS. Vol 75 1997 517 M. Ramsay et al. Table 3: Sensitivity and specificity of clinical case definitions for the detection of measles Case definition No. of confirmed cases/Total Sensitivity (%) Specificity (%) Notified case of measles Yes 126/3 290 (3.83)a No NAb Not classified NA Suspect casec Yes 120/2644 (4.54) 96.77; 91.9-99.1d 16.16; 14.9-17.5 No 4/510 (0.78) Not classified 2/262 (0.76) Probable casee Yes 116/2171 (5.34) 94.31; 88.6-97.7 27.71; 26.1-29.3 No 7/832 (0.84) Not classified 3/415 (0.72) USA/Canada case definition' Yes 91/1 529 (5.95) 85.85; 77.7-91.9 41.08; 39.2-43.0 No 15/1 066 (1.41) Not classified 20/824 (2.43) a Figures in parentheses are percentages. b NA = not available. c Notified case with fever and rash. d Figures in italics are the 95% confidence interval. e Notified case with fever (maximum fever temperature was not recorded on the request form) + rash + (cough or coryza or conjunctivitis). f Notified case with fever (maximum fever temperature was not recorded on the request form) + rash ¢3 days + (cough or coryza or conjunctivitis). would have been excluded. The case definition based on that used in the USA and Canada (fever, rash -3 days, and cough or coryza or conjunctivitis) would have reduced the number of cases tested by 1066/ 2595 (41%) and would have missed 15 (14%) con- firmed cases. In addition, using any of the case defi- nitions would have excluded further confirmed cases for whom information on clinical features was miss- ing. A further 20 confirmed cases would have been missed using the definition from the USA and Canada, 3 with the probable case definition and 2 using the suspect case definition. Of the 108 cases notified as rubella and which fulfilled the USA/Canada case definition for mea- sles, only one (0.9%) was confirmed as measles; this compares with 88/1221 (7.2%) of notified measles cases that fulfilled the same case definition (rash ¢3 days, fever, and cough, coryza or conjunctivitis) during this same period (from 94/44 to 95/16). Discussion Where a vaccine-preventable disease is close to elimination, the aim of a surveillance system is to detect all cases and to facilitate a rapid response to prevent a potential outbreak (11-13). Prior to the availability of an acceptable method for the routine laboratory confirmation of measles, clinical case definitions were used for measles surveillance. The ideal clinical case definition is sensitive (to detect a high proportion of true cases) and specific (to limit the number of false cases). Even if the sensitivity and specificity of a case definition remain unchanged, however, a reduction in the prevalence of the disease concerned will result in a fall in the predictive value of a "positive" case (11). The results of the present survey have shown that the use of clinical case defini- tions would have resulted in substantial numbers of nonmeasles rashes being misclassified as measles, while a substantial amount of resources would have been expended on further follow-up or control measures (14). In addition, because different pro- portions of cases were confirmed in different age groups (4, 15) false conclusions could have been made about the groups at highest risk of measles, possibly leading to incorrect policy decisions (11). In England and Wales, surveillance based on notified cases or on any of the clinical case definitions exam- ined here would have underestimated the effect of the mass vaccination campaign. Between November 1994 and June 1996 the number of notified cases fell by less than 50%, whereas notified cases which were confirmed by laboratory testing fell by over 95%. Demonstration of measles control, therefore, requires both a sensitive reporting system for sus- pected cases and the widespread use of a specific method for laboratory confirmation. In England, the Department of Health recommend that all suspected cases of measles (fever and rash) be tested for sali- vary IgM (5). In practice, cases notified as measles are offered testing, whether or not they meet one of WHO Bulletin OMS. Vol 75 1997518 Surveillance of measles in England and Wales the clinical case definitions. This article indicates that the application of any clinical case definition would have slightly reduced the sensitivity of a surveillance system based upon notified cases. The use of the suspect case definition (fever and rash) would have resulted in four confirmed cases being missed, com- pared with seven or more cases had the more specific case definitions been applied. Because all notified cases were offered testing, however, the number of tests performed was much greater than if only prob- able or suspect cases had been tested. The use of a case definition to select cases for laboratory testing would therefore have substantially reduced the costs of the surveillance system. Since the completeness of clinical information was variable, however, addi- tional resources would be required to follow all cases to obtain complete data for classification. In view of the low incidence of true measles in England and Wales, however, many confirmed cases may repre- sent false-positive saliva IgM tests (16) and a propor- tion of the cases "missed" by selection on the basis of a clinical case definition may not be true measles infections. In a country with good measles control, where the number of suspected cases far exceeds confirmed cases, the costs of obtaining full clinical information and/or performing laboratory testing on all suspected cases will be high for a low yield of true cases. Since the incidence of true measles varies over time and between countries, different approaches to the use of clinical and laboratory case definitions will, therefore, be required. Restriction of the use of laboratory testing to those cases which fulfil a clinical definition would also be favoured if the investigation had low accept- ability (such as the collection of paired sera). The large number of notified cases for which saliva speci- mens were submitted indicates that salivary sur- veillance is a feasible and acceptable system for confirming measles infection in the community. Therefore, if resources permit, this noninvasive ap- proach is ideal for maximizing the sensitivity and specificity of the surveillance system. It is not known, however, how many cases of rash and fever illness, which are not diagnosed or notified as measles (and therefore not eligible for saliva testing), are attribut- able to measles infection. The very low proportion of confirmed measles among cases notified as measles, and the even lower proportion among cases notified as rubella suggest that, in view of the current epide- miology of this infection, this surveillance pro- gramme does not grossly underestimate the true number of measles cases. The eventual global elimination of measles will depend on achieving high vaccine coverage and a sustained surveillance programme to identify high- risk groups and to inform policy-makers. As the inci- dence of true measles infections declines and the proportion of health care workers who have not en- countered measles increases, it will be important to maintain and monitor the established surveillance systems. In the years leading up to the campaign for global eradication of poliomyelitis, strict criteria were established to evaluate poliomyelitis surveil- lance (17). We propose that similar criteria be used to evaluate measles surveillance programmes in a country nearing elimination of this disease. Cases of paralytic poliomyelitis caused by wild poliovirus are currently detected by maintaining sur- veillance for acute flaccid paralysis (AFP) (16). An expected incidence for nonpoliomyelitis AFP (1 case per 100000 children under 15 years of age) has been established and acts as a benchmark for evaluating the sensitivity of a surveillance programme for polio- myelitis caused by wild virus. If the reported AFP incidence exceeds the expected minimum rate, the surveillance system is sufficiently sensitive. A mini- mum expected notification rate for nonmeasles rash and fever illnesses could be used to monitor measles surveillance. The first 35 weeks of experi- ence with the PHLS surveillance programme leads us to expect approximately 7000 notifications of sus- pected measles (excluding those confirmed by the measles saliva assay) annually in England and Wales. This represents an annual rate of approximately 20 notifications per 100000 persons; after continued surveillance a more accurate baseline will be avail- able. Once the baseline is established, a fall in notifi- cation rates to below this level could indicate a fall in the sensitivity of the surveillance scheme. This rate of suspected measles must reflect the incidence of nonmeasles rash and fever illnesses. Preliminary work suggests that rubella, parvovirus B19 (4) and human herpesvirus 6 (HHV6) infections (18) can lead to a notification of suspected measles; a detailed study of the infectious causes of rash and fever ill- nesses in England is planned. The true causes and incidence of such illnesses are unclear and are likely to vary over time and to differ in countries where other causes of rash and fever illness, such as dengue fever, are common. An accurate baseline of the incidence of nonmeasles rash and fever illnesses, suitable for use in many countries, will be available when the surveillance scheme has been running for longer in England and Wales and when other coun- tries establish similar schemes. Since the start of the measles programme, saliva samples have been obtained from over 50% of noti- fied cases, and for over 45% of cases, samples were taken during the appropriate period. For samples taken outside the time period recommended, the sensitivity of the test falls and therefore real measles cases may be missed (4). The proportion of cases WHO Bulletin OMS. Vol 75 1997 519 M. Ramsay et al. appropriately tested must be maintained or im- proved to ensure maximum sensitivity of the surveil- lance scheme. In addition to the need to obtain appropriate specimens from suspected cases, a rapid turnaround for laboratory testing and for feedback of results is required. Over 70% of saliva specimens were received within 1 week of sampling and over 90% of results were reported within 3 weeks. Con- firmed cases were reported even sooner, usually by telephone or fax to the consultants in communicable disease control, permitting early follow-up. Because of the nature of the immune response, however, the saliva assay does not consistently detect IgM in the first week of the illness (19), and saliva samples should not be taken until some time after disease onset. The ideal laboratory test would be able to confirm the diagnosis within hours so that rapid re- sponse to control the outbreak could be instituted. Even with a rapid turnaround for the saliva test, there is still the possibility that secondary cases may develop. Despite this limitation, it is desirable that confirmation should be obtained as early as possible, and rapid feedback maintained. This will encourage the appropriate investigation of subsequent cases and help to ensure that control measures are insti- tuted as soon as possible. We have demonstrated that the PHLS salivary surveillance programme is a sensitive, specific, ac- ceptable and timely surveillance scheme for measles. We intend to use these baseline measures to monitor the ongoing programme. Criteria need to be devel- oped to evaluate such methods of measles surveil- lance. This programme could form a model for other countries hoping to establish enhanced surveillance as measles nears elimination. Acknowledgements We gratefully acknowledge the contribution of all the con- sultants in communicable disease control, general practi- tioners, patients, and parents who have contributed to this surveillance. We also acknowledge the laboratory and administrative staff who helped to run the surveillance scheme. Resume Surveillance epidemiologique de la rougeole en Angleterre et au Pays de Galles: implications du programme national de test sur la salive L'impact d'une campagne de vaccination systema- tique contre la rougeole et la rubeole en Angleterre et au Pays de Galles a 6t6 evalu6 en faisant appel aux r6sultats donnes par un test qui recherche dans la salive les IgM sp6cifiques de la rougeole et a 6t6 offert a tous les cas notifies. Grace aux donn6es cliniques fournies par les m6decins notificateurs, nous avons estime la sensibilit6 et la sp6cificit6 de diverses d6finitions du cas clinique en fonction de la confirmation ult6rieure. Un 6chantillon de salive a 6t6 obtenu dans le temps requis pour 3 445/7 574 (45,4%) cas notifi6s; la proportion des cas con- firm6s par le test sur la salive a 6t6 faible et elle est pass6e de 67/681 (9,8%) au d6but de la campagne a 1/373 (0,3%) apres 35 semaines. La sp6cificit6 de toutes les definitions du cas clinique est faible, mais elle est la plus forte, a 41,1% (intervalle de con- fiance a 95%, 39,2-43%), pour la definition simi- laire a celle recommandee par les Etats-Unis d'Amerique. L'extrapolation a partir des cas ob- serves pendant la surveillance en cours permet une estimation du taux annuel de notification pour des rashs et des fievres qui ne sont pas caus6s par la rougeole. Dans les pays ou la lutte contre cette maladie est bonne, la definition du cas clinique manque trop de specificite pour permettre une sur- veillance exacte et la confirmation par le laboratoire des cas suspects est requise. Le test sur la salive constitue alors une methode acceptable. Pour que la surveillance de la rougeole garde une bonne sensibilite en Angleterre et au Pays de Galles, il faut que soient suivies l'incidence des rashs et des fievres notifi6s ainsi que la proportion des cas pour lesquels le test de laboratoire permet d'exclure une infection par la rougeole. References 1. Cutts F, Waldman RJ, Zoffman H. Surveillance for the expanded programme on immunization. Bulletin of the World Health Organization, 1993, 71: 633-639. 2. Ramsay M et al. The epidemiology of measles in England and Wales: rationale for the 1994 national vaccination campaign. Communicable disease report review, 1994, 4: R141-R146. 3. Surveillance of measles since the vaccination cam- paign. Communicable disease report weekly, 1995, 5: 35. 4. Brown DWG et al. Salivary diagnosis of measles: a study of notified cases in the United Kingdom. British medical journal, 1994, 308: 1015-1017. 5. Chief Medical Officer. Measles and Rubella Immuni- sation Campaign. London, HMSO, 1994 (PL CMO (94) 12). 6. Strassburg MA. Guidelines for the investigation and control of outbreaks of EPI target diseases. Geneva, World Health Organization, 1984 (unpublished docu- ment EPI/GEN/84/7 available upon request from Glo- bal Programme for Vaccines and Immunization, World Health Organization, 1211 Geneva 27, Switzerland). 520 WHO Bulletin OMS. Vol 75 1997 Surveillance of measles in England and Wales 7. Using surveillance data and outbreak investigations to strengthen measles immunization programmes. Geneva, World Health Organization, 1996 (unpub- lished document WHO/EPI/GEN/96.02 available upon request from Global Programme for Vaccines and Immunization, World Health Organization, 1211 Geneva 27, Switzerland). 8. Centers for Disease Control. Case definitions for public health surveillance. Morbidity and mortality weekly report, 1990, 39: RR-13. 9. Implementation of the ACE Guidelines for measles control in Canada. Canada diseases weekly report, 1991, 17: 35-40. 10. Mortimer PP, Parry JV. Non-invasive virological diagnosis: are saliva and urine specimens adequate for blood? Reviews in medical virology, 1991, 1: 73- 78. 11. Cutts F, Brown DWG. The contribution of field tests to measles surveillance and control: a review of avail- able methods. Reviews in medical virology, 1995, 5: 35-40. 12. Hutchins SS et al. A school-based measles out- break: the effect of a selective revaccination policy and risk factors for vaccine failure. American journal of epidemiology, 1990, 132: 157-168. 13. Wassilak SGF et al. Continuing measles transmis- sion in students despite a school-based outbreak con- trol programme. American journal of epidemiology, 1985, 122: 208-217. 14. Robertson SE et al. A million dollar measles out- break: epidemiology, risk factors, and a selective revaccination strategy. Public health reports, 1992, 107: 24-31. 15. Adjaye N et al. Measles serology in children with a history of measles in early life. British medical journal, 1983, 286:1478. 16. Ramsay M, Cohen B, Brown D. Serum IgM testing is needed in all cases of suspected measles. British medical joumal, 1996, 313: 231. 17. Report of the First Meeting of the Global Commission for the Certification of the Eradication of Poliomyelitis. Geneva, World Health Organization, 1995 (unpub- lished document WHO/EPI/GEN/95.6 available upon request from Global Programme for Vaccines and Immunization, World Health Organization, 1211 Ge- neva 27, Switzerland). 18. Tait DR et al. Exanthem subitum (roseola infantum) misdiagnosed as measles and rubella (corrected title). British medical journal, 1996, 312: 101-102. 19. Perry KR et al. The detection of measles, mumps and rubella antibodies in saliva using antibody capture radioimmunoassay. Journal of medical virology, 1993, 40: 235-240. WHO Bulletin OMS. Vol 75 1997 521