Dr B. A. Wills Wellcome Trust Clinical Research Unit, Centre for Tropical Diseases, 190 Ben Ham Tu, Quan 5, Ho Chi Minh City, Viet Nam. Tel.: +84 8 835 3954; Fax: +84 8 835 3904; E-mail: email@example.com
OBJECTIVES A positive tourniquet test is one of several clinical parameters considered by the World Health Organization to be important in the diagnosis of dengue haemorrhagic fever, but no formal evaluation of the test has been undertaken. As many doctors remain unconvinced of its usefulness, this study was designed to assess the diagnostic utility of both the standard test and a commonly employed modified test.
METHODS A prospective evaluation of the standard sphygmomanometer cuff tourniquet test, compared with a simple elastic cuff tourniquet test, was carried out in 1136 children with suspected dengue infection admitted to a provincial paediatric hospital in southern Viet Nam.
RESULTS There was good agreement between independent observers for both techniques, but the sphygmomanometer method resulted in consistently greater numbers of petechiae. This standard method had a sensitivity of 41.6% for dengue infection, with a specificity of 94.4%, positive predictive value of 98.3% and negative predictive value of 17.3%. The test differentiated poorly between dengue haemorrhagic fever (45% positive) and dengue fever (38% positive). The simple elastic tourniquet was less sensitive than the sphygmomanometer cuff, but at a threshold of 10 petechiae (compared with the WHO recommendation of 20) per 2.5 cm2 the sensitivity for the elastic tourniquet rose to 45% (specificity 85%). Other evidence of bleeding was frequently present and the tourniquet test provided additional information to aid diagnosis in only 5% of cases.
CONCLUSION The conventional tourniquet test adds little to the diagnosis of dengue in hospitalized children. The simple, cheap elastic tourniquet may be useful in diagnosing dengue infection in busy rural health stations in dengue endemic areas of the tropics. A positive test should prompt close observation or early hospital referral, but a negative test does not exclude dengue infection.
Infection with dengue virus is one of the leading causes of illness and hospital admission in Southeast Asian children (Halstead 1997; Gubler 1998). Infection may be asymptomatic, or may result in a variety of clinical syndromes ranging from dengue fever (DF), a non-specific febrile illness, to dengue haemorrhagic fever (DHF) and dengue shock syndrome (DSS). The main feature differentiating DHF from DF is an increase in vascular permeability resulting in leakage of plasma from the intravascular compartment to the extravascular space (Halstead 1989; Rigau-Perez et al. 1998). In severe DHF the loss of fluid is critical and hypovolaemic shock develops resulting in potentially lethal DSS. Thrombocytopenia, coagulation abnormalities and clinical bleeding of varying severity are features of DHF and DSS, but may also occur in DF (Srichaikul & Nimmanitya 2000).
Over the last 40 years the incidence and thus the public health importance of dengue virus infections has increased dramatically in many tropical and subtropical countries (Monath 1994). Diagnosis is often difficult, particularly for the less severe manifestations of dengue infection (DF and mild DHF), which cannot easily be differentiated from the many other febrile infections of childhood. In patients with DSS clinical diagnosis is usually more straightforward. However, the possibility of progression from mild DHF to more severe disease requiring prompt resuscitation means that, despite the lack of facilities for serological or virological confirmation in many endemic areas, early and accurate diagnosis is important.
The World Health Organization (WHO) has published guidelines for the diagnosis and management of dengue infections (WHO 1997). One of the features which may be used in the clinical case definition of DHF is a positive tourniquet test. For this, the WHO guidelines stipulate that a blood pressure (BP) cuff should be inflated on the upper arm of the patient to a point mid way between systolic and diastolic pressure for 5 min, and the number of resulting petechiae in 2.5 cm2 on the volar aspect of the forearm just distal to the antecubital fossa should be counted. A test is considered positive when 20 or more petechiae are observed in the 2.5 cm2. The tourniquet test reflects both capillary fragility and thrombocytopenia. In a study of 240 children in Delhi in 1996 (Kabra et al. 1999), in which the WHO case definitions were used, the tourniquet test was positive in 40% of children with DF, 18% of children with DF with unusual bleeding, 62% with DHF and 64% with DSS. In another study involving 172 Thai children (Kalayanarooj et al. 1997), 36% of those with DF and 52% of those with DHF had positive tourniquet tests. In addition, a positive tourniquet test was found in 21% of children with a final diagnosis of viral infection other than dengue, although the investigators comment that a positive tourniquet test was significantly more likely to be found in children with dengue infection.
Many children with suspected dengue infection do not present to a hospital facility but rather to a health clinic or community practitioner where access to a sphygmomanometer may be limited or busy staff may find it difficult to maintain the cuff pressure for the required 5 min. Potentially useful information is therefore not available to the health worker to assist in making a clinical diagnosis and, more importantly, in making the critical decision as to whether it is safe to monitor the child at home or whether referral to a hospital is necessary. A simple elastic tourniquet is usually available even in the simplest of health facilities. This and other alternative `tourniquets' are used regularly to perform a modified, simpler, tourniquet test in rural health centres in Viet Nam, but no validation of the method has ever been carried out. The aim of this study was first to compare prospectively the results of the modified tourniquet test with those of the standard WHO method in a large group of children with suspected dengue infection, and secondly to evaluate the usefulness of both methods in establishing a clinical diagnosis of DF or DHF, using the WHO case definitions supported by confirmatory serology.
The study was conducted on the Infectious Disease Ward and the Intensive Care Unit of Dong Nai Paediatric Hospital, a provincial hospital 40 km north of Ho Chi Minh City. The study was approved by the Ethical and Scientific Committees of Dong Nai Paediatric Hospital and the Centre for Tropical Diseases of Ho Chi Minh City. Informed consent was obtained from the parents or guardians of the patients.
Patients and clinical methods
Children aged between 1 and 15 years, hospitalized between June 1996 and June 1998 with a diagnosis of suspected dengue infection (either a history of fever for <7 days with a negative malaria smear and no clinical evidence of an alternative diagnosis, or presentation with classical DSS) were eligible for entry to the study provided a parent or guardian gave informed consent. A member of the study team recorded the clinical details on a standard form, shortly after admission.
The BP was measured in both arms to ensure that the results were similar. A standard tourniquet test was first performed on the right arm by a nurse (observer 1) specifically trained in this technique, using the WHO protocol as described above. A series of six sites on the flexor and extensor aspects of the forearm, identified by the letters A to F, were displayed pictorially for reference on each study sheet. Observer 1 was asked to record the site at which the maximum number of petechiae were visible, choosing from the six sites A to F and the standard WHO site. In addition, observer 1 recorded both the number of petechiae visible in 2.5 cm2 at this `maximum' site, and the number visible at the standard site if the two were different. For this, a 2.5-cm2 window was cut out from a sheet of coloured plastic; the plastic sheet was laid on the arm and petechiae visible in the window were counted.
Immediately after this, observer 1 performed a modified tourniquet test on the left arm, using a simple elastic tourniquet (Lotto, Viet Nam). New elastic tourniquets measure 25 cm in length and 2.5 cm in width. In a small pilot study we have shown that when new, the elastic cuffs can be stretched by 50% of their initial length, but that after repeated use the cuffs lose their elasticity. Prior to each modified tourniquet test the maximum stretched length of the elastic cuff (EC) was measured and the cuff discarded if this length exceeded 38 cm. On average a new cuff was used for every tenth patient. The EC was applied, at maximal stretch, around the midpoint of the left upper arm for 5 min. In this way the pressure was sufficient to stop venous return but not occlude arterial supply and the forearm became dusky in colour. The site on the forearm at which the maximum number of petechiae were visible after the test, together with the actual number visible at this site and the standard site, were recorded in the same way as for the standard tourniquet test. A second trained observer recorded independently the same set of observations from both arms a few minutes later, without conferring with observer 1. Children presenting with grade IV DSS, that is without a palpable pulse or measurable BP, were resuscitated first with parenteral fluids and the two tourniquet tests were performed once adequate circulation had been established.
All children were treated in accordance with the hospital guidelines for management of dengue infections, based on the current WHO guidelines. Clinical progress, outcome and all laboratory results were recorded in simple research notes and later transferred to a database.
Serum samples were obtained from all children at the time of admission, and from the majority at discharge. All samples were separated immediately and stored at –70 °C. Subsequently, the samples were analysed using conventional capture enzyme-linked immunosorbent assay techniques for dengue and Japanese encephalitis (JE) viruses either at the Institute of Health and Community Medicine, Universiti Malaysia, Sarawak, Malaysia, or at the Department of Virology, Armed Forces Research Institute of Medical Sciences, Bangkok, Thailand. Criteria for serodiagnosis were as described previously for each laboratory (Innis et al. 1989; Cardosa et al. 1992).
After discharge, a panel of three paediatricians reviewed all available information on each patient, including the serology results, and assigned a final diagnosis using the WHO case definitions for both DF and DHF (Table 1). Evidence of increased vascular permeability is necessary for a diagnosis of DHF, inferred either from changes in the haematocrit or by the presence of pleural effusions or ascites. The sensitivity of clinical detection of effusions and ascites is poor unless the volumes of fluid are large. We did not have facilities for X-ray or ultrasound examination for all children and therefore chose to use haemoconcentration as the indicator of increased vascular permeability. The WHO guidelines define the degree of haemoconcentration necessary for a diagnosis of DHF as a rise in the haematocrit ≥20% above the average for age, sex and population. In another study involving more than 1000 medical admissions to Dong Nai Paediatric Hospital during the same time period, the mean haematocrit for children aged 5–10 was 37% (C.X.T. Phuong, personal communication). Thus for this study, if the haematocrit was ≥44% on admission or subsequently, this was accepted as evidence of increased vascular permeability and contributed to the diagnosis of DHF.
Table 1. Criteria for the diagnosis of dengue fever and dengue haemorrhagic fever, adapted from the WHO guidelines. Patients must satisfy all the criteria stipulated for the respective diagnoses
A diagnosis of DHF can be made without serological confirmation, provided all other clinical and laboratory criteria are fulfilled (WHO 1997). Those children who did not satisfy all criteria for a diagnosis of DHF, but had positive serology were classified as DF. Those who satisfied some of the criteria for DHF but had indeterminate serology were considered unclassifiable. Patients with definite negative serology were considered not to have dengue infection even if they fulfilled some of the criteria for DHF or DF.
For identification of the site at which the maximum number of petechiae were noted the kappa statistic (Altman 1991) was used to measure both agreement between observers using a single method and agreement between the two cuff methods within a single observer. For the number of petechiae noted at the maximum and standard sites the mean differences (95% limits of agreement) were calculated to measure inter-observer and inter-method agreement (Bland & Altman 1986). Sensitivity, specificity, positive predictive values and negative predictive values of both methods in the diagnosis of dengue infection were calculated, first using the strict WHO definition of a positive tourniquet test, and later using a series of other defined cut-off points. The optimum cut-off points for the two tourniquet tests were determined from receiver operating characteristic curves. For this the gold standard was serologically confirmed dengue infection. All statistical analyses were performed using the program SPSS for Windows, version 8.0 (SPSS Inc., Chicago, USA).
Between June 1996 and June 1998, 1136 children were admitted to Dong Nai Paediatric Hospital with suspected dengue infection and entered the study. Of these patients, 905 underwent both the standard BP cuff and the modified EC tourniquet test. Further analysis relates to this group of 905 children. Serology confirmed dengue infection in 548 (60.6%) children, with indeterminate results in a further 286 (31.6%), in most cases because the second serum sample was not obtained or was obtained too early for definitive serodiagnosis. Serology was negative in 71 children (7.8%). For the diagnostic classification, 312 of the 905 children (34.5%) fulfilled the case definition for DF and 286 (31.6%) the case definition for DHF. Of this latter group, 50 children fulfilled all the diagnostic criteria for DHF but had indeterminate serology results. The 71 seronegative children were grouped together as `not dengue' and there were 236 children (26.1%) who were unclassifiable. Figure 1 presents the number of subjects within each diagnostic group according to the maximum number of petechiae observed with the BP cuff tourniquet test. The results for the maximum number of petechiae observed have been grouped together as follows: 0–4, 5–9, 10–19, 20–39 and ≥40.
Using the WHO criteria, which stipulate that the test should be considered positive if more than 20 petechiae are found in 2.5 cm2, the BP cuff technique was positive in 294 of the 905 patients (32.5%; 95% CI 29.5–35.7) and in 228 of the 548 seropositive patients (41.6%; 95% CI 37.5–45.9). The distribution of positive tourniquet tests according to the diagnostic groups is shown in Table 2. Comparatively, 38.1% of those with DF had positive tourniquet tests, as did 45.1% of those with DHF. Of the unclassifiable patients 17.8% had positive tests; it is likely that many of these patients had dengue infection but serological confirmation was not possible with the specimens available. In the 71 seronegative children the tourniquet test was positive in 5.6%; most of these children are likely to have had viral infections other than dengue.
Table 2. Results of the standard WHO tourniquet test in various diagnostic categories
Blood pressure cuff
In 11 of the patients one of the observers failed to perform the test or did not document their findings. This gave a total of 894 patients for which agreement between the two observers could be evaluated. With regard to site, the proportion with observed agreement equalled 93.5% (836/894), with very good agreement between the two observers (kappa=0.83). In 77% of the cases the site recorded by both observers as showing the maximum number of petechiae was on the volar aspect of the forearm, immediately proximal to the antecubital fossa, i.e. about 5 cm proximal to the WHO standard site. The total number of petechiae documented at the maximum site differed on average by −0.45 between the two observers [95% limits of agreement (range): −8.86, 7.96 (−33, 30)]. The average difference in the total number of petechiae counted at the standard site was −0.39 [95% limits of agreement (range): −6.14, 5.36 (−30, 19)].
The EC test was not performed or the site not written on the form by either one of the observers for six patients. Thus, the inter-observer agreement of the EC method could be evaluated for 899 patients. The percentage of observed agreement for site was 91.0 (818/899) again with very good inter-observer agreement (kappa=0.82). The site most commonly recorded (67% of the time) to show the maximum number of petechiae was the same as for the BP cuff method, and the number of petechiae differed on average by −0.11 between the two observers [95% limits of agreement (range): −7.73, 7.51 (−35, 35)]. The average difference in the number of petechiae recorded at the standard site on the arm was −0.22 [95% limits of agreement (range): −5.26, 4.82 (−32, 13)].
Agreement between the elastic cuff and blood pressure cuff
Comparing the two methods with regard to the maximum site (Table 3), for observer 1 the percentage of observed agreement was 72.3%, with fair agreement (kappa=0.38). Similar results were noted for observer 2 (percentage of observed agreement=70.9%, kappa=0.36). Also comparing the two methods, there was similar agreement in relation to the numbers of petechiae counted by the observers. At both the maximum and the standard sites the total number of petechiae recorded were statistically significantly less when using the EC compared with the BP cuff. On average the EC observers reported approximately 4–5 petechiae less at the maximum site and approximately 2–3 petechiae less at the standard site. At the maximum site approximately 26% and at the standard site approximately 12% of the individuals differed by more than 10 petechiae for the readings of the two cuff methods.
Table 3. Agreement between the elastic cuff (EC) and blood pressure (BP) cuff methods for the tourniquet test
Predictive value of blood pressure and elastic cuff methods
Using the WHO criteria, the BP cuff technique was positive in 294 patients (32.5%) and the EC technique in 195 patients (21.5%). Table 4 shows the predictive values of the two methods in the diagnosis of dengue infection (DF and DHF together), where the serological diagnosis is taken as the gold standard. Both methods showed good specificity but poor sensitivity.
Table 4. Predictive value of blood pressure cuff and elastic cuff methods for the tourniquet test for diagnosis of dengue infection
To see if less exacting criteria for a positive result might improve the usefulness of the tourniquet test in the diagnosis of dengue infection, we looked at the sensitivity and specificity for varying cut-offs of maximum observed petechiae with both methods (Figure 2). As expected, the sensitivity of the tests improved and the specificity decreased when the cut-off for the number of petechiae was reduced. The BP cuff was more sensitive than the EC at each cut-off point. Using a cut-off of 10 for both the BP cuff and the EC improved the sensitivity from 41.6 to 69.0% and 28.5 to 46.2%, respectively. The respective reductions in specificity were 94.4–73.2% and 97.2–84.5%.
In 286 of the children (31.6%) no definitive serological diagnosis could be made. Table 5 gives the percentage of children with more than 10, 15 or 20 petechiae at the maximum site, determined by both the BP and EC for the three groups, serologically confirmed dengue, indeterminate and negative. In children with indeterminate serology the percentage with petechiae above each cut-off always lies between the percentages observed for the children with a positive or negative dengue diagnosis. This indicates that a bias between those with definitive serological results and those with indeterminate serology is unlikely.
Table 5. Percentage of children with more than the specified cut-off for number of petechiae observed at the maximum site, using the two tourniquet test methods
In the assessment of a child with suspected dengue infection we have demonstrated that both the conventional BP cuff tourniquet test and the simpler EC technique show good agreement between observers for both site and maximum number of petechiae, but that the BP cuff method results in the development of consistently greater numbers of petechiae. The most common site for the greatest number of petechiae to occur is just proximal to the antecubital fossa, about 5 cm proximal to the recommended WHO site.
Using the WHO criteria the BP test gave positive results in about 40% of patients with confirmed dengue infection, but the test was positive in those with DF almost as frequently as in those with DHF. In part, this may reflect the exacting nature of the WHO criteria for dengue diagnosis. Some children have established capillary leak but unless repeated platelet counts are performed they may not fulfil the criteria for DHF and are classified as having DF. Similarly, a few children have evidence of capillary leak and positive serology but never show any haemorrhagic manifestations, an absolute requirement for the diagnosis of DHF. Thus, a number of the children classified as having DF are likely actually to have had DHF and the proportions of positive tourniquet tests may be misleading. However, it should be noted that the test was also positive in 5.6% of children who definitely did not have dengue infection. This pattern of results is very similar to that seen in several previous smaller studies looking at the tourniquet test in dengue infection.
Overall, a positive standard tourniquet test is reasonably specific and has a high positive predictive value for dengue infection, if performed on children suspected to have dengue from an endemic area. It differentiates poorly between DF and DHF as classified by the WHO criteria. Unfortunately, the sensitivity of the tourniquet test is low using the accepted cut-off point of 20 petechiae per 2.5 cm2, although it is possible that this would improve if repeated tests were carried out in the days following admission. A positive test is only one of several markers of bleeding that are acceptable for the case definition of DHF. However, in only 43 of the 263 children (16.3%) without any other evidence of bleeding on admission, was the test positive; thus, the test provided additional information to aid diagnosis in <5% of children in the study. It is important to recognize, however, that the population is highly selected, in that all children were thought to have dengue infection and were considered ill enough to warrant hospital admission. Medical and nursing staff are likely to have more time to evaluate children fully in hospital, including careful examination to detect skin petechiae or mucosal bleeding, and to have access to laboratory tests in the form of a haematocrit and platelet count.
In a busy clinic or rural health facility both time and equipment are often limited, and the focus is likely to be on triage to identify those patients with serious or potentially serious disease amongst the many children seen each day. Despite its limitations the EC tourniquet test may have a role in these circumstances. The cuff is cheap, readily available and simple to use; no specialized equipment or maintenance are required. If a cut-off point of 10 petechiae per 2.5 cm2 is chosen the test had a sensitivity of 46.2% comparable with the standard WHO test using a BP cuff (sensitivity 42%), although the specificity was only 84.5%. In a child with fever of short duration and non-specific viral symptoms a positive test might prompt referral from a distant rural site to a clinic with facilities for observation and basic intervention during the danger period for shock. Alternatively, parents might be taught to look out for specific warning signs and asked to return daily or twice daily for review. However, there are many children with dengue infection, at risk of DSS, who do not have a positive tourniquet test, so it is critical that a negative test should not be interpreted as signifying that the child does not have dengue infection.
In conclusion, it seems that in a hospital setting the BP cuff tourniquet test adds little to the diagnosis of dengue infection. Used in the community, a positive test would be helpful in predicting dengue infection, provided it is accepted that a negative test does not exclude dengue infection. Both the conventional test with a cut-off of 20 petechiae, and the modified test using a cut-off of 10 petechiae give similar results, but the EC method is simpler to perform, cheaper and likely to be much more widely available.
Dr Cao Xuan Thanh Phuong, who planned and supervised the study, died tragically at the end of October 2000. Her dedication to the care of sick children and enthusiasm for clinical research are greatly missed by friends, colleagues and the local community.
The authors are grateful to the Director and staff of the Dong Nai Paediatric Hospital, in particular the doctors and nurses of the Infectious Diseases ward. We are also grateful to Dr M.J. Cardosa and the staff of the Institute of Health and Community Medicine, Universiti Malaysia, Sarawak, Malaysia, and to Dr T. Endy and the staff of the Department of Virology, Armed Forces Research Institute of Medical Sciences, Bangkok, Thailand, for their generous help with serological confirmation of dengue infection.
The study was supported by The Wellcome Trust of Great Britain.
Dong Nai Paediatric Hospital Study Group: Pham Thi Thu Thuy, Nguyen Thi Tuyet Anh, Tran Thanh Thuy, Truong Dinh Luat, Nguyen Van My, Nguyen Thi Que Phuong, Chu Van Thien, Nguyen Thi Thuy Nga.