HIV antibody assay that gave false negative results: multicentre collaborative study

BMJ 1997; 315 doi: (Published 27 September 1997) Cite this as: BMJ 1997;315:772
  1. Barry G Evans (bevans{at}, consultant epidemiologista,
  2. John V Parry, clinical scientist, grade Ca,
  3. Philip P Mortimer the multicentre Collaboratice Study Group, consultant virologista
  1. a Public Health Laboratory Service, London NW9 5EQ
  1. Correspondence to: Dr Evans
  • Accepted 14 March 1997


Objective: To identify false negative results arising from the use of a commercial kit to detect antibody to HIV-1 and HIV-2 between July 1995 and March 1996.

Design: The 56 laboratories in the United Kingdom that were using the assay were asked to retrieve and retest specimens with an alternative assay for HIV-1 and HIV-2. Details of false negative results were obtained and these serum samples further investigated.

Subjects: 24 181 patients tested with the assay who were reported as being negative for HIV antibody. An additional 497 patients were confirmed as HIV positive with the assay.

Results: Serum samples of 20 973 of the patients were retested, and four patients were found to have had false negative results with the kit; three further patients were found to have had false negative results in the course of other laboratory testing. The seven patients with false negative results with the kit were of diverse risk group and HIV-1 subtype. Four had evidence of recent HIV infection.

Conclusion: The commercial kit had a sensitivity of 99.2% (497/501), or less if the additional three patients with false negative results were taken into account.

Key messages

  • Dual screening should be considered for patients known to be at increased risk of HIV infection

  • Although assay kits for HIV antibody generally show high sensitivity, false negative results may still arise during routine use and go undetected

  • Tested serum samples should be stored so that they may, if necessary, be checked.

  • Any kit failures should be reported to the Medical Devices Agency or Adverse Incident Centres, as well as to the manufacturer

  • All laboratories should be informed of test kit failures, whether or not they have used the suspect kit


Early in 1995 Abbott Laboratories modified its assay kit for HIV antibody (IMx HIV 1/2 3rd Generation Plus) to increase the sensitivity for the recently described outlier HIV-1 variants, collectively designated subtype O. This assay (code 8B32) was sold in the United Kingdom from July 1995 after it had been evaluated on 785 serum samples containing HIV antibody (S Dooley et al, fifth joint meeting of the European Group for Rapid Viral Diagnosis and the European Society against Virus Diseases, Prague, September 1995) and several panels of sequential specimens from plasma donors undergoing HIV antibody seroconversion.

On 29 March 1996 Abbott wrote to all laboratories that had been supplied with the assay, stating: “Information obtained from users of this product has identified a small set of high titre positive samples yielding some discrepant results with other assays.” Customers were advised to “discontinue use of [the assay] or evaluate each specimen both undiluted and at a 1:4 dilution.” This followed a report by a public health laboratory to Abbott on 5 February of a false negative result with the kit and, reputedly, similar findings in France and Spain. At the end of March 1996 it also emerged that two other public health laboratories had just found false negative results with the kit in the course of confirmatory testing for HIV infection. Considerable alarm was created among those who had been screened for HIV infection since the previous July, and an urgent retesting programme was therefore begun at the start of April 1996.


By 8 April all user laboratories in England and Wales had been sent a questionnaire, and those with stored specimens that had been tested by the assay and found to be negative had agreed either to test them by another assay or to send them to a reference laboratory for retesting. Similar action was taken in Scotland and Northern Ireland.

Telephone contact was maintained to monitor progress, and on 17 and 18 April a second questionnaire was sent by facsimile to determine how many specimens had been found to be positive for HIV antibody with the kit, whether retesting of stored specimens that had negative results with the assay had been completed, and for how many patients stored specimens could not be traced. Laboratories were asked to refer specimens that were previously negative with the kit but positive in another assay to the Virus Reference Division of the Central Public Health Laboratory in London.


An initial analysis was published on 12 April 1996,1 and this is updated here. Of the 66 laboratories in the United Kingdom cited by Abbott Laboratories as users of its assay, 56 used the kit as their sole screening assay. Twenty eight were NHS laboratories, 18 public health laboratories, and 10 private laboratories. Five laboratories had tested over 1000 patients with the kit; 17 had screened fewer than 100 patients.

Of over 24 000 patients originally found to be negative for HIV antibody with the kit, almost 21 000 had stored specimens retested with an alternative assay (table 1). Four of these patients' specimens were confirmed as positive for HIV antibody.

Table 1

Retesting of patients who had been tested for HIV antibody with Abbott assay between July 1995 and March 1996

View this table:

Clinicians treating these four patients were advised that an erroneous result had been issued. Table 2 shows details of these four patients and of three others whose serum samples had given false negative results in the Abbott assay (all originally tested between January and March 1996). The serum samples available for study from these seven patients were positive in all other screening enzyme immunoassays by which they were tested. One patient (case 1) had serological evidence of recent infection (positive for HIV IgM antibody, with only p24 and p160 bands on western blotting). Another patient (case 3) seemed to have had an acute illness related to HIV infection. Two other patients (cases 5 and 7) were known to have seroconverted within the previous three months. Apart from the specimen in case 1, which was not followed up further, all the specimens with false negative results and further specimens from each of these patients gave strong reactions to all major HIV-1 proteins on western blotting. Serotyping showed the presence of five different subtypes of HIV-1.2

Table 2

Laboratory data on seven patients with false negative results in Abbott assay

View this table:

The questionnaires established that 497 confirmed HIV antibody specimens had been identified after screening antibody with the kit between July 1995 and March 1996. This allows the sensitivity of the assay to be estimated as 99.2%—497/501 (497 true positive specimens plus 4 false negative specimens). It would be less if the other three false negative results are taken into consideration.

The specimens of 2519 patients had not been stored (table 1). As many as possible of these patients were recalled, bled, and retested by a different assay.


Within three weeks of the defect in the kit being acknowledged by the manufacturer the retesting of over 20 000 specimens had been completed and four negative results retracted. Many more patients were made anxious because few knew which assay had been used at screening. Laboratories and clinics that had not used the test were contacted by concerned patients, and many patients had to be recalled because their serum specimen had not been stored.

The absence of false negative reactions in early trials, the failure of the retesting programme to identify false negative results with the Abbott kit among patients screened between July and December 1995, and the occurrence of all seven false negative results between January and March 1996 suggest that the defect was present only in the batches of the assay distributed in the United Kingdom after January 1996. If so, the working sensitivity of the kit between January and March 1996 was considerably lower than the 99.2% estimate. A previous study of 12 kits assaying antibody to HIV-1 and HIV-2 in specimens in the United Kingdom found that their sensitivities were 99.9% or greater.3

Most of the NHS and public health laboratories had stored all of their patients' serum samples, but less than 30% of the samples tested in private laboratories had been stored. Regrettably, not all of the patients who had been tested by private laboratories could be traced, so their serostatus remains unknown.

We failed to identify any specific feature common to all patients whose specimens gave false negative results with the Abbott kit—for example, low titre of HIV antibody or reactivity to HIV IgM antibody. The samples were of several subtypes, and an unexpectedly high proportion (four of seven) had evidence of recent infection. Abbott Laboratories have since attributed the defect to the complement activity of some fresh undiluted serum samples. Its modified assay, launched in the United Kingdom in September 1996, incorporates edetic acid (EDTA) in the diluent to inhibit this activity.

This incident has stimulated debate about means of preventing future false negative results in screening assays for HIV antibody. Firstly, before release or modification of such kits, they should be evaluated using many known positive specimens both by the manufacturer and by an independent laboratory. To be confident statistically that the sensitivity of a kit significantly exceeds even the estimated sensitivity of the defective kit described here (99.2%) would, however, entail examining several thousand positive serum samples.

Secondly, specimens from cases in which the clinical suspicion of HIV infection is raised should be tested by two screening assays, not just one. Unfortunately, such a suspicion is not always communicated to the laboratory.

Thirdly, laboratories screening specimens for HIV antibody must retain samples.

Finally, even single HIV assay failures should be reported to the Adverse Incident Centre of the Medical Devices Agency in England (tel: 0171 972 8223 or 8273; fax: 0171 972 8113) or to the Adverse Incident Centres in the health departments in Northern Ireland, Scotland, or Wales, as well as to the manufacturer. Then, if a pattern of false reactivity emerges, rapid action can be taken.

This incident shook public confidence in HIV testing, which is expected to be virtually 100% sensitive. Vigilance is needed to avoid false negative results.


K Osborne and J McMenamin coordinated the data collection and helped in drafting the report. The retesting exercise relied on the excellent cooperation of many epidemiologists and laboratory scientists throughout England, Wales, Scotland, and Northern Ireland. J Booth, I Chrystie, P Morgan-Capner, E M Ndawula, J Paul, G Underhill, and M Zuckerman kindly provided specimens and results for further analysis. B Davis, D Goldberg, V King, and C Roberts also gave valuable personal time to the exercise. C-P Pau generously supplied reagents for HIV serotyping.

Funding: No additional funding.

Conflict of interest: None.


  1. 1.
  2. 2.
  3. 3.
View Abstract