This review investigated the accuracy and usefulness of the haemoglobin colour scale (HCS) for the diagnosis of anaemia. The authors concluded that the HCS is potentially accurate and useful in the diagnosis of mild and moderate anaemia in places where there is no laboratory-based method, but further research is required. The conclusions appear reliable.

To assess the accuracy and usefulness of the haemoglobin colour scale (HCS), compared with a reference standard and clinical diagnosis, for the diagnosis of anaemia.

MEDLINE, EMBASE, CINAHL and the Science Citation Index were searched for relevant studies until May 2005; the search terms were reported. The titles and abstracts of studies listed in the World Health Organization Haemoglobin Colour Scale bibliography and the bibliographies of all included studies and reviews were checked for additional studies. Personal communication with experts was also used to identify further studies. No language restrictions were applied.

Study designs of evaluations included in the review

Any diagnostic accuracy study was eligible for inclusion. The review did not report the types of design included.

Specific interventions included in the review

Studies assessing the HCS against a reference standard for the diagnosis of anaemia were eligible for inclusion. Some studies also assessed the accuracy of clinical examination, copper sulphate method, Sahli method and the Tallquist paper scale against the reference standard. The included studies generally used a haemoglobin level of less than 11 or less than 12 g/dL as an indication of anaemia, and a level of less than 6 or less than 7 g/dL as severe anaemia; one study used a haemoglobin level of less than 5 g/dL to define very severe anaemia. Some studies used a venous blood sample for the measurement whereas others used a finger prick sample.

Reference standard test against which the new test was compared

Studies that used haemiglobincyanide (cyanmethaemoglobin) as the reference standard to determine the presence or absence of anaemia were eligible for inclusion. HemoCue was also considered as a reference standard in poorer tropical countries. The included studies used HemoCue, automated analyser, photometer and copper sulphate methods as the reference standard.

Participants included in the review

Inclusion criteria for the participants were not specified. The included studies selected the participants from population-based or hospital-based samples. The study participants were either pregnant women, blood donors, children, adult out-patients, adult and child in-patients, or hospital out-patients. The prevalence of anaemia where reported) ranged from 5 to 85%.

Outcomes assessed in the review

Inclusion criteria for the outcome measures were not specified. The sensitivity and specificity of the HCS were calculated for anaemia and severe anaemia (where data were available), for each included study. Diagnostic odds ratios (DORs) were also estimated for individual studies, and used to compare diagnostic performance between subgroups.

How were decisions on the relevance of primary studies made?

Both authors independently assessed the studies for inclusion using a pre-structured piloted form, but the authors did not state how any possible disagreements were resolved.

The studies were assessed for: clear description of selection criteria, methods of testing and the reference standard used; whether blood for the HCS test and the reference standard were collected and analysed at the same time; whether the HCS test results were interpreted without knowledge of the results of the reference standard; whether the same clinical data were available when the HCS results were interpreted as would be available when the test is used in practice; whether intermediate HCS results were reported; explanation of withdrawals; whether results from all those applying the test were used; and whether the spectrum of patients was representative of those who will receive the test in practice. Further information about how these criteria were used was not provided.

Both authors independently assessed the quality of the studies, but the authors did not state how any possible disagreements were resolved.

Both authors independently extracted the data using a pre-structured data extraction form, but the authors did not state how any possible disagreements were resolved. Data were extracted on participant characteristics, study characteristics (e.g. health care setting and operator training) and results.

How were the studies combined?

The studies were stratified into three groups. Six studies were laboratory based with adequate training, four were real-life with short training, and the remaining four had elements of both (e.g. laboratory based with inadequate training or real-life with extensive training). The study results, including the sensitivity and specificity of the HCS, were reported for each study group. The precision of the estimates (e.g. 95% confidence interval, CI) was reported where it was available or could be calculated from the raw data. Pooled estimates of sensitivity and specificity were calculated from those laboratory-based studies that were homogeneous and reported true-positive, true-negative, false- negative and false-positive rates. DORs were reported for each study group.

How were differences between studies investigated?

The authors presented the findings of studies separately for each study group, and discussed the possible reasons for differences among individual studies and between study groups. The possible effect of patient characteristics, test operators (e.g. level of training and experience) and threshold effect on the estimate of sensitivity and specificity were discussed. The chi-squared test was used to assess statistical heterogeneity between estimates of sensitivity and specificity in laboratory-based studies. Spearman's rho was used to assess the correlation between study-level estimates of sensitivity and specificity.

Fourteen studies were included. The study designs and total number of participants were not reported.

Thirteen studies reported the sensitivity and specificity of the HCS. Apparently, in all of these studies the results of the HCS were interpreted without knowledge of the results of other methods. Overall, laboratory studies were less heterogeneous than other studies and reported a higher sensitivity and specificity.

The sensitivity and specificity in the five laboratory-based studies ranged from 0.85 to 0.99 and from 0.91 to 1.0, respectively. A pooled estimate from four of these studies (those reporting 2x2 data) showed an overall sensitivity of 0.87 (95% CI: 0.86, 0.89) and a specificity of 0.97 (95% CI: 0.97, 0.98). The authors did not find any statistically significant heterogeneity between estimates of sensitivity and specificity in laboratory-based studies (P<0.25 for both), and all laboratory-based studies used the same diagnostic threshold.

The sensitivity in the real-life studies ranged from approximately 0.44 to 0.88 and the specificity ranged from approximately 0.45 to 0.88. Pooled estimates were not calculated for these studies. Real-life studies reported significantly lower DORs for anaemia (between 2.5 and 7.5) compared with laboratory-based studies (between 58 and 48,000)(P=0.003).

Six studies reported results for severe anaemia. The sensitivity was over 0.84 in three of these studies and 0.5 (95% CI: 0.42, 0.58) in one study (the only 'real-life' study), and in all of these studies the specificity was between 0.86 (95% CI: 0.84, 0.88) and 0.99 (95% CI: 0.99, 1.0).

The five studies (two 'real-life', two 'unclear' and one laboratory based) that compared the accuracy of both the HCS and clinical examination with a reference standard reported a sensitivity of 0.33 (95% CI: 0.29, 0.38) to 0.57 (95% CI: 0.34, 0.79) and a specificity of 0.79 (95% CI: 0.73, 0.79) to 0.84 (95% CI: 0.79, 0.88) for clinical examination. They found that the HCS generally had better sensitivity. Some included studies compared the accuracy of other methods with the HCS. One study showed that the HCS had a similar sensitivity and specificity to the Sahli method, while three studies showed that the HCS had a similar sensitivity and specificity to the copper sulphate method.

There was no statistically significant association between study-level estimates of sensitivity and specificity (Spearman's rho 0.027, P=0.29), suggesting that factors other than variation in the diagnostic threshold account for between-study heterogeneity.

The HCS may be a reasonably accurate and useful tool for the diagnosis of mild and moderate anaemia in places where there is no laboratory. Further investigation is warranted.

The review question was clear and the inclusion criteria, though broad, were appropriate to the question. The search strategy was reasonably comprehensive, although it was not clear if efforts were made to identify unpublished studies. Two reviewers independently performed the study selection, data extraction and quality assessment processes, but it was unclear how they resolved any possible disagreements. Appropriate criteria were used to assess the methodological quality of the studies, but poor reporting of study methods hindered the quality assessment. In addition, the results were not discussed with adequate consideration of study quality. Many studies did not provide adequate information on the criteria used to select the participants or health workers, thereby hindering the usefulness of comparisons across studies.

The methods of data synthesis were generally appropriate, although the accuracy results of individual studies were only partially reported. Potential sources of heterogeneity were discussed; appropriate statistical methods were used to assess the presence of a threshold effect and to test for residual heterogeneity. Given the differences among the studies, the synthesis of only laboratory-based studies with similar findings and which used the same diagnostic threshold appeared appropriate, though there was some heterogeneity in participants and settings. Since the participants and settings varied among the studies, the results may not be representative of routine practice. Overall, the study conclusions seem to be supported by the data presented.

Practice: The authors stated that the HCS may improve diagnosis of anaemia where laboratory tests are not available but, owing to the lack of robust evidence, it is, as yet, difficult to draw a conclusion about the accuracy and usefulness of the HCS in routine practice.

Research: The authors stated that further research is required to assess the role of the HCS in the diagnosis of anaemia and its usefulness when combined with clinical diagnosis.

Department of International Development in the UK (Effective Health Care Research Consortium and Malaria Knowledge Programme).

This is a critical abstract of a systematic review that meets the criteria for inclusion on DARE. Each critical abstract contains a brief summary of the review methods, results and conclusions followed by a detailed critical assessment on the reliability of the review and the conclusions drawn.