Diagnosis of fever in Africa

Diagnosis of fever in Africa

CORRESPONDENCE 2 3 4 5 Pfeffer MA, Braunwald E, Moye LA, et al. Effect of captopril on mortality and morbidity in patients with left ventricular d...

56KB Sizes 3 Downloads 38 Views

CORRESPONDENCE 2

3

4

5

Pfeffer MA, Braunwald E, Moye LA, et al. Effect of captopril on mortality and morbidity in patients with left ventricular dysfunction after myocardial infarction. Results of the survival and ventricular enlargement trial: the SAVE Investigators. N Engl J Med 1992; 327: 669–77. Lerman A, Gibbons RJ, Rodeheffer RJ, et al. Circulating N-terminal atrial natriuretic peptide as a marker for symptomless leftventricular dysfunction. Lancet 1993; 341: 1105–09. Motwani JG, McAlpine H, Kennedy N, Struthers AD. Plasma brain natriuretic peptide as an indicator for angiotensinconverting-enzyme inhibition after myocardial infarction. Lancet 1993; 341: 1109–13. McDonagh TA, Robb SD, Murdoch DR, et al. Biochemical detection of left-ventricular dysfunction. Lancet 1998; 351: 9–13.

SIR—Early identification of symptomless left-ventricular dysfunction may be advantageous to prevent the onset of overt heart failure. 1 However, the prevalence of this disorder in the overall population is not known and no recommendations exist for the identification of affected individuals. In this context, Theresa McDonagh and colleagues (Sept 20, p 829)1 should be congratulated for a comprehensive analysis of echocardiographic data on left-ventricular function in the Glasgow MONICA population.2 They report that 7·7% of individuals in the urban population have ejection fractions of less than 35% by biplane Simpson’s rate method, 2 most of whom had evidence of ischaemic heart disease, hypertension, or both. Further, 4·0% of men and 2·0% of women presented with ejection fractions of even less than 30% by this method. These numbers seem fairly high, but the investigators correctly point out that the prevalence of coronary artery disease—the main aetiologic factor for congestive heart failure—is higher in Glasgow than in most European states and North America.3 We compare the data from McDonagh and colleagues with those from the MONICA Centre in Augsburg, Germany, where the prevalence of myocardial infarctions is much lower.3 M-mode echocardiography was done according to the guidelines of the American Society of Echocardiography. Of 1866 individuals studied, technically satisfactory results were obtained in 755 men and 811 women aged 25–75 years (mean 50 [SD 13] years). Left-ventricular volumes were determined with the Teichholz equation. Reduced leftventricular systolic function was defined as an ejection fraction two SDs below the mean of individuals without a history or evidence of cardiovascular disease (<48%).

372

The overall proportion of individuals with an ejection fraction below 48% was 2·7%, with higher rates in men than in women (3·2 vs 2·3%, p<0·001). This rate varied according to age: from 2·1% in people aged 25–35 years to 4·3% among those aged 66–75 years. Of 43 participants with reduced left-ventricular systolic function, 25 were symptomatic or known to have cardiovascular disease. The most frequent diagnoses were hypertension (ten people), coronary heart disease (nine), valvular heart disease (four), and dilated cardiomyopathy (two). With a different formula for echocardiographic calculation of ejection fraction, severely reduced systolic function (ejection fraction 35%) was found in one man with history of myocardial infarction and in one woman who had dilated cardiomyopathy; both received loop diuretics and were thus symptomatic according to the criteria used by McDonagh. The findings of these two demographically similar MONICA studies are important. First, the prevalence of left-ventricular systolic dysfunction is higher in men than in women and increases with age. 2 Second, rates of severe left-ventricular systolic dysfunction might be geographically distinct, as has been shown for myocardial infarctions. 3 The two studies differ with regard to symptomfree patients: in the common sample, all patients with severe left-ventricular systolic dysfunction (ejection fraction <35%) had symptoms, whereas the Scottish investigators report a high rate of symptom-free patients. Thus, symptomless severe left-ventricular dysfunction may be less common in a population with a moderate prevalence of coronary artery disease than in one with a high rate of disease. *Heribert Schunkert, Ulrich Broeckel, Hans W Hense, Ulrich Keil, Guenter A J Riegger Institut für Epidemiologie und Sozialmedizin, University of Münster, Münster, Germany; GSF Forschungzentrum, Institut für Epidemiologie, Munich-Neuherberg; and *Klinik und Poliklinik für Innere Medizin II, University of Regensburg, D-93053 Regensburg, Germany 1

2

3

The SOLVD Investigators. Effect of enalapril on mortality and the development of heart failure in asymptomatic patients with reduced left ventricular ejection fractions. N Engl J Med 1992; 327: 658–91. McDonagh TA, Morrison CE, Lawrence A, et al. Symptomatic and asymptomatic left-ventricular systolic dysfunction in an urban population. Lancet 1997; 350: 829–33. Tunstall-Pedoe H, Kuulasmaa K, Amouyel P, Arveiler D, Rajakangas AM,

Pajak A. Myocardial infarction and coronary deaths in the World Health Organization MONICA Project. Registration procedures, event rates, and case-fatality rates in 38 populations from 21 countries in four continents. Circulation 1994; 90: 583–612.

Diagnosis of fever in Africa SIR—Ellen Einterz and Myra Bates (Sept 13, p 781)1 report a striking unreliability in the assessment of fever by patients or their carers attending an outpatient service in Cameroon, and ask whether patients know when they are hot. Their data suggest that they do. Recalculation of their data shows that a high proportion of respondents correctly identified fever, especially in children (sensitivity; table). The proportion of patients correctly identified without fever (specificity) was low but constant over the agegroups. The proportion of patients who correctly reported fever (positive predictive value) varied with the prevalence of fever as measured by axillary temperature. Hence, the reported low positive predictive values are largely explained by the low prevalence of fever, and to some degree by a low specificity. Can fever be diagnosed with an acceptable level of accuracy without a thermometer? We have carried out a community-based cluster sample survey among children aged 2·5–37 months in Mtito Andei Division, which is an area of seasonal malaria transmission in Kenya. A trained interviewer showed mothers (or carers) how to assess fever by feeling the child’s forehead and comparing this temperature with that of their own forehead. The mother was then asked in her own language if the child had a raised body temperature, the expression for which differs from that for malaria or malaria-like illnesses. The prevalence of fever, defined by concurrent measurement of an axillary temperature (肁37·5°C) by electronic thermometer (Philips HP5316), was 0·06 (19 of 317 examined children). The sensitivity of this test was similar (0·89; 17 of 19 febrile children) to the value reported by Einterz and Bates, but the specificity was much higher (0·80; 238 of 298 non-febrile children). The proportion of mothers who correctly reported the absence of fever (negative predictive value) of this test was 0·99. The positive predictive value was 0·22 (17 of 77 children) but would be expected to be higher in a clinical setting, in which self-selection for fever before presentation would result in a higher prevalence of fever than at the

THE LANCET • Vol 351 • January 31, 1998

CORRESPONDENCE

Age (years)*

Prevalence*†

Fever on asking*

Fever by feeling‡

Sensitivity

Specificity

PPV

NPV

PPV

NPV

<1 1–5 5–15 肁15

0·26 0·42 0·33 0·10

0·90 0·94 0·76 0·74

0·41 0·48 0·55 0·45

0·35 0·56 0·45 0·13

0·92 0·91 0·82 0·94

0·61 0·76 0·68 0·33

0·95 0·91 0·94 0·99

All

0·20

0·84

0·46

0·28

0·92

0·53

0·97

PPV=positive predictive value (proportion of patients or carers who correctly reported fever); NPV=negative predictive value (proportion of patients or carers who correctly reported absence of fever). *Data from Einterz and Bates1 or ‡calculated with prevalence data from Einterz and Bates.1 †Prevalence of axillary temperature 肁37·5°C.

Predictive values of reports of fever

community level. For example, with the prevalence data recorded by Einterz and Bates, the predictive values shown in the above table are obtained. Our findings suggest that mothers can accurately diagnose the absence of fever without a thermometer. The use of clinical signs alone will necessarily result in an unacceptably high proportion of false positives, especially at peripheral levels of health care. Patients or carers should refer to health workers to confirm diagnosis of fever by thermometer before initiating antimalarial treatment, although this should not result in a delay of treatment. For this purpose, we believe that thermometers should be made available at the lowest level of health care—for example, community-health workers. We strongly support the suggestion by Einterz and Bates that in the absence of fever on examination, history of fever should be supported by other clinical signs before antimalarial treatment is prescribed. Funding for this study was received through a grant from the Netherlands Foundation for the Advancement of Tropical Research (WOTRO).

*Hans Verhoef, Elsa Hodgins, Clive E West, Jane Y Carter, Frans J Kok Division of Human Nutrition and Epidemiology, Wageningen Agricultural University, PO Box 8129, 6700 EV Wageningen, Netherlands; and African Medical and Research Foundation, Nairobi, Kenya 1

Einterz EM, Bates ME. Fever in Africa: do patients know when they are hot? Lancet 1997; 350: 781.

SIR—Ellen Einterz and Myra Bates1 find that patients’ and carers’ report of fever is not reliable—only 28% of patients claiming to have fever were in fact febrile. They therefore conclude that “in the absence of fever on examination, history of fever should be supported by clinical signs such as splenomegaly, anaemia, or jaundice before antimalarial treatment is prescribed”. As part of a malaria treatment study conducted by the Bandim Health Project and Laboratório Nacional da Saúde Publica at Bandim Health Centre in the outskirts of Bissau, the capital of Guinea-Bissau, we asked the mother or carer of sick children with symptoms compatible with malaria if

THE LANCET • Vol 351 • January 31, 1998

the child had had fever during the past 12 h. Children with 20 or more parasites per 200 leucocytes had their axillary temperature measured. 70 children were included in each of three treatment groups, two of which had to return to the health centre on day 1 and day 2. Again the same questions were asked and the temperature measured. Of the 210 children, some of the information for day 0 was unavailable for seven children. On day 1 and 2, 18 and 25, respectively, of the expected 140 children did not attend, or some information was missing. The table below shows the percentages of mothers who reported fever (肁37·5°C) in children and the actual temperature. On day 0, 38 (18·7%) of 203 children had a temperature below 37·5°C; on day 1, 110 (90·2%) of 122 children; and on day 2, 109 (94·8%) of 115 children. There was no difference in parasitic density between the febrile and the non-febrile children on inclusion. These results suggest that a substantial proportion of sick children with a maternal report of fever and high parasite levels were afebrile when attending a health facility. We only included children with parasitaemia and cannot therefore estimate the risk of overdiagnosing malaria based on maternal reports of fever. However, our data indicate that relying only on measured temperature may lead to underdiagnosis of malaria. Redd and colleagues2 found a higher specificity but a lower sensitivity when they included splenomegaly, nailbed pallor, and actual temperature in the malariacase definition. However, because of lack of training of health staff “adequate performance should not be assumed”. Therefore, even if new algorithms for diagnosing clinical Temperature (°C) <37·5

肁37·5

Day 0 No Yes

2·6% (1) 97·4% (37)

2·4% (4) 97·4% (161)

Days 1 and 2 No Yes

67·6% (148) 32·4% (71)

66·7% (12) 33·3% (6)

Carers’ and mothers’ responses to question, “Did your child have fever during the last 12 h?” on the basis of actual temperature when attending health centre

malaria are needed, they must be feasible and realistic at the primary-healthcare level in developing countries. Until that has been achieved our results indicate that the mothers’ history of fever is an important index in deciding to treat or not to treat a child for malaria. *Poul-Erik Lund Kofoed, Francisco Dias, Francisco Lopes, Lars Rombo Projecto de Saúde de Bandim, Guinea-Bissau; *Department of Paediatrics, Hospital of Kolding, DK 6000 Kolding, Denmark; Laboratório Nacional da Saúde Publìca, Guinea-Bissau; and Division of Infectious Diseases, Pulmonary Medicine and Dermatology, Mälarsjukhuset, Ekilstuna, Sweden 1

2

Einterz EM, Bates ME. Fever in Africa: do patients know when they are hot? Lancet 1997; 350: 781. Redd SC, Kazembe P, Luby SP, et al. Clinical algorithm for treatment of Plasmodium falciparum malaria in children. Lancet 1996; 347: 223–27.

Authors’ reply SIR—Four additional studies that compare self-assessment of fever with measurement of body temperature have now been reported. In mothers previously trained to diagnose fever by tough, Hans Verhoef and colleagues in Kenya find a declaration of fever to be incorrect in 78% of cases. Dilys Morgan and colleagues (Nov 22, p 1549),1 in a study of HIV-seropositive Ugandan adults and controls, show declarations of fever to be incorrect in 85% of cases. These are close to our own results from Cameroon, where 72% of clinic patients (including 87% of adults) who claimed to be febrile at the time of examination were, in fact, afebrile. The two other studies had lower rates of normothermia in patients who claimed to be febrile. Poul-Erik Kofoed and colleagues in Guinea-Bissau selected only children with important malaria parasitaemia and failed to confirm a reported history (within the previous 12 h) of fever in 39% of patients, whereas Samuel Dunyo and colleagues (Nov 22, p 1550),2 whose data in the table differ slightly from those in the text, in a communitybased study in Ghana report normal temperatures in 24% of those who simultaneously believed they had fever. Combining data from all five studies,

373