CORRESPONDENCE 491 rapidly, might have led to more rapid clearance of HRP- 2. With chloroquine, which acts less rapidly, antigenae- mia may have persisted longer, and this could account for the disparity in findings. The utility of the Par- aSighP-F test as a diagnostic tool for the detection of treatment failures would thus depend on the antimalar- ial drug being used and also probably on the country of use, since the kinetics of HRP-2 during the decline phase of malaria and the effect of the immune status of the individual are not yet completely understood. Although our sample was small, our own feeling is that treating all patients with a positive PuruSightTM-F test on day 15 may result in a large number of patients receiving ‘rescue’ medication when it is not warranted. s. vakharia’ N. GopinathanZ N. A. Kshirsagarl Departments of ‘Clinical Pharmacology and 2Pharmacology K. E. M. Hospital Pare1 Mumbai 400 012 India 20 February 1997 The ParaSightTM-F test for detecting treatment failure: a reply We thank Dr Vakharia and colleagues for their com- ments [above] on our recent article (Karbwang et al., 1996: Transactions, 90, 5 13). The aim of our study was to shorten the follow-up period after the treatment of malaria, which is usually 28 d for short half-life drugs such as artemether and 42 d for those with a longer half- life such as mefloquine. If any drug can clear parasitae- mia as well as sequestered parasites within the first week after treatment (i.e., curative therapy), HRP-2 is not likely to be detected beyond 6 d after the last parasite has been killed (i.e., 14 d after the treatment). With a drug which did not clear the sequestered parasites with- in the first week, and was no longer present beyond one week due to its short half-life (e.g., artemether), HRP-2 would certainly be detected beyond 6 d and the out- come will be a recrudescence. This is what happened in our study. On the other hand, if the drug is not very po- tent but has a long half-life, the parasites would be ex- pected to be cleared only slowly and sequestered parasites would be present beyond the first week after treatment. In patients with a sensitive response, the se- questered parasites are eventually cleared but HRP-2 would be circulating longer than 14 d after treatment (due to slow clearance of the parasites). It is not surpris- ing that the PuraSightTM-F test remained positive be- yond 14 d in patients with a sensitive response after chloroquine treatment in the study by Dr Vakharia and colleagues [above]. Their study was carried out in an area with increasing chloroquine resistance of Plasmodi- urn falciparum; thus, the sequestered parasites would be cleared slowly and HRP-2 would be expected to remain circulating for a few days after 14 d (depending on how long the parasites were still present after the first week). In this situation the PuraSighP-F test should be per- formed even after day 14 to determine the day when it becomes negative in patients with a sensitive response, which would be the best day to predict recrudescence in patients with chloroquine-resistant malaria who have been treated with chloroquine. Clinical Pharmacology Unit Faculty of Tropical Medicine Mahidol University 42016 Rajvithi Road Bangkok 10400 Thailand Juntra Karbwang 7 March 1997 Hepatitis B and C viruses and hepatocellular carcinoma Olubuyide et al. (1997: Transactions, 91, 38) present- ed recently the results of a study on hepatocellular car- cinoma (HCC) in Nigeria. The authors performed a case-control study in an attempt to evaluate the role of hepatitis B virus (HBV) and hepatitis C virus (HCV) in the development of HCC. Hospital cases of HCC were matched with hospital controls who did not have HCC. Serum samples were taken from both groups and tested for HBsAg (a marker for carriage of HBV) and antibody against HCV (anti-Ho. The authors investigated the association between these markers and the presence of HCC in this sample of patients. Unfortunately, howev- er, there were errors in their statistical analyses which draw into question the validity of their conclusions. We have re-analysed the data presented in Table 2 of Olubuyide et al. (lot. cit.) and contrasted our findings with those of the authors. As will be evident, we not only found discrepancies between the analyses but we also found a number of inconsistencies in the paper, e.g. lack of agreement between the text and their Table 3.* All our statistical analyses were performed using the commercial software SAS@ Release 6.03 (SAS Institute Inc., Cary, North Carolina, USA). Our estimates of confidence intervals on the odds ratios were derived by the standard method. At the bottom of the first column on p. 39, the au- thors report a value of P<O.OOl for the association be- tween the prevalence of HBsAg and HCC. That is, they claim that there is a highly significant association be- tween the prevalence of HBsAg and the presence of HCC. Simple inspection of their data suggests other- wise. As the authors state: ‘Of the 64 cases with HCC, 38 (59.3%) were HBsAg positive compared with 32 (50%) of the controls (P<O.OO1)‘. In fact, the x2 statistic for the 2x2 contingency table of HCOtHBsAg is 1.135. With one degree of freedom, this yields a P value of 0.287. Thus, as common sense would suggest, there is no statistically significant association. The authors go on to report a similar (P<O*OOl) asso- ciation between the prevalence of anti-HCV and HCC (12 of the 64 cases and 7 of the 64 controls were anti- HCV positive). Once again, their claim of statistical sig- nificance is unfounded. The x2 statistic for this associa- tion equals 1.545. With one degree of freedom, this yields a P value of 0.214. Again, there is no statistically significant association between the prevalence of anti- bodies to HCV and HCC. Further, the unadjusted odds ratio calculated for the above association is 1.88. The 95% confidence interval (95% CI) calculated for this odds ratio is 0.69-513. Thus, while the odds ratio is correctly supplied in Table 3, the confidence interval reported by the authors in this table (0.63-9.77) is erroneous-although it correctly in- cludes unity, indicating that there is no statistically sig- nificant additional risk of HCC in anti-HCV positive individuals. More worrying than the error in the calcu- lation of this confidence interval is the way that this re- sult is reported in the text and abstract. Twice in the text and once in the abstract this odds ratio is reported as be- ing 6.88, larger by 5 than the correct value of 1.88. Fur- ther, the lower bound of their (erroneous) confidence interval, when reported in the text and abstract, has be- come 1.63. rather than 0.63. Thus their conclusion that ‘the risk for developing HCC increased with the pres- ence of anti-HCV (odds ratio=6*88)’ is erroneous and misleading. *The odds ratio of 6.88, which is the discrepancy referred to here, was in fact corrected to 1.88 by the author during the processing of this paper, and the 95% CI was changed to 0.63-9.77. The values were corrected inTable 3 but, owing to an oversight on my part, the corresponding changes were not made in the text. Unfortunately this omission was not noticed by the author or myselfwhen the proofs were checked. -Editor.