Factors Associated with False-Positive Results from Fingerstick OraQuick ADVANCE Rapid HIV 1/2 Antibody Test

Abstract

Objective:

Identify factors associated with false-positive rapid HIV antibody tests.

Design:

This retrospective cohort study with nested case–controls involved patients tested for HIV by Boston Medical Center (BMC) affiliates.

Methods:

Cases had a reactive fingerstick OraQuick ADVANCE rapid HIV 1/2 antibody test and a negative Western blot. Controls had nonreactive rapid tests. We compared the prevalence of HIV risk factors between cases and the total nonreactive population and the prevalence of other clinical factors between cases and controls.

Results:

Of the 15 094 tests, 14 937 (98.9%) were negative and 11 (0.07%) were false positives (specificity of 99.9%). Cases were more likely to have had an HIV-infected sex partner and to be tested at certain sites compared to true negatives. More cases than controls had O-negative blood type.

Conclusion:

O-negative blood type and sex with an HIV-infected person may increase false-positive HIV fingerstick results. More targeted studies should examine these risk factors.

Keywords

rapid HIV antibody test case–control analysis Western blot false positive true negative specificity

Introduction

Over the past several years, reports from the New York City and Minnesota State Health Departments identified a spike in false-positive results with the OraQuick ADVANCE HIV 1/2 oral fluid test. These results caused both organizations to employ fingerstick testing.^1,2 While false-positive HIV tests have been reported with fingerstick samples undergoing OraQuick ADVANCE HIV 1/2 testing as well, the rate appears to be much lower.³

The Centers for Disease Control and Prevention’s (CDC) 2003 initiative, “Advancing HIV Prevention,” recommended expanded routine HIV testing in clinical and other settings.⁴ Testing a larger population with lower disease prevalence will lead to an increase in false-positive tests. These developments, therefore, merit further efforts to identify risk factors for testing falsely positive on a rapid fingerstick HIV antibody test.

Previous studies examining risk factors primarily evaluated Western blots and enzyme-linked immunosorbent assay (ELISA) tests. Novick et al tested patients with injection drug abuse using ELISA tests and found a 7% false-positive rate.⁵ Several ELISA-based investigations have reported an association between false-positive results and autoimmune diseases, possibly due to cross-reactivity of the HIV antigen epitope and immune factors.⁶ Other conditions reported to be associated with false-positive ELISA results include pregnancy, transplantations, infections (visceral leishmaniasis, leprosy, malaria, mycobacterial infections, cytomegalovirus), recent vaccinations, transfusions, malignancies, renal disease, and alcoholic liver disease.^{7

–17} These links may be due to the subsequent creation of alloantibodies that cause a false-positive result.^8,9,18

A paucity of studies has examined the risk factors associated with obtaining rapid HIV antibody test false-positive results. Premarketing studies performed by the manufacturer of OraQuick ADVANCE HIV 1/2 testing in HIV-negative patients linked false positives to multiparity, the presence of rheumatoid factor, Epstein Barr virus, hepatitis A (HAV), and hepatitis B (HBV) viruses.³ The above-mentioned outbreak of false-positive HIV results from oral fluid tests propelled the Minnesota Department of Public Health (DPH) to investigate factors associated with false-positive results. It found that patients aged 37 years or older were 4 times more likely to have a false-positive oral test than people under 37 years old.² The San Francisco DPH noted a relationship between false positives in oral fluid tests and expiration date within 1 month of use.¹⁹ Reports from tropical countries with rapid fingerstick HIV test screening programs show false positives in people with malaria, sleeping sickness, dengue fever, and schistosomiasis.^20

–23 In addition, false positives have been found among pregnant women (younger and nulliparous women were more likely to be false positives) and a patient with angioimmunoblastic T-cell lymphoma.^24,25

To further clarify the risk factors associated with false-positive results in patients undergoing rapid HIV tests using fingerstick samples, we performed a retrospective cohort study with nested case–controls among individuals tested by BMC’s affiliated rapid HIV testing services.

Methods

Our study sample consisted of patients receiving HIV counseling and testing services through Project TRUST and Project Umbrella, 2 BMC-affiliated programs working out of 10 sites between 2008 and 2009. Sites included a sexually transmitted infection (STI) clinic, an urgent care center, a drug rehabilitation center, a tuberculosis clinic, a methadone clinic, and an acute care hospital. The BMC institutional review board approved this project.

We performed 2 primary analyses, a retrospective cohort to assess the association between false-positive results and patient characteristics (demographic and traditionally defined HIV risk factors) using the Projects TRUST and Umbrella databases and a case–control analysis to evaluate clinical factors using hospital and ambulatory medical record abstraction. Case patients were defined as those with a reactive test on the fingerstick OraQuick ADVANCE HIV 1/2 test but a negative result on the confirmatory Western blot. Controls were defined as patients who tested negative on the rapid test. Controls were selected by simple random sampling and matched using frequency matching to cases based on age (within a 5-year range) and test date (within 1 month).

We used Fisher exact test to compare the prevalence of risk factors between the cases and the total nonreactive population and to compare the prevalence of medical conditions between cases and controls. In addition, we calculated the specificity of the OraQuick ADVANCE HIV 1/2 test using data from the Projects TRUST and Umbrella databases. A full information maximum likelihood imputation was performed to estimate missing responses, and the case–control analysis was recomputed with the imputed responses.²⁶

We estimated there would be 10 to 20 false-positive cases identified. To achieve a power of 80% with a significance of .05, and an estimated odds ratio of 4.0 of the potential risk factor, we enrolled 4 controls per case. We matched with a ratio of 1:4, yielding 44 controls for the 11 cases ultimately identified.

Results

During 2008 to 2009, there were 15 094 patients tested for HIV through Project TRUST and Project Umbrella. One hundred and forty-six were HIV positive (0.97%) as determined by positive rapid test and confirmatory Western blot. There were 14 948 patients who tested negative for HIV: 14 937 had a negative rapid test and comprised the control pool from which the matched controls were identified; 11 had a reactive rapid test but negative Western blot (cases). These data revealed a test specificity of 99.9%.

Table 1 shows the characteristics of the case, controls, and total nonreactive groups. The plurality of cases was African American (36%) and the average age was 40.7 years. There were, however, no statistically significant differences among the groups with respect to age, gender, and ethnicity. Of the 11 cases, 4 (36%) were from one testing site (community health center) and 3 (27%) from another testing site (STI clinic); these sites represented 9% (1323) and 18% (2782) of the total nonreactive tests, respectively (P < .001).

Table 1.

Demographics of Cases (False Positives), Controls (Matched True Negatives), and Total Nonreactive Population

Patient Characteristics	Cases, N = 11	Controls, N = 44			Total Nonreactive Population, N = 14937
Patient Characteristics	Total N (%)	Total N (%)	P Value^a	OR^b	Total N (%)	P Value^c	OR^d
Age, y
<20	0 (0)	0 (0)	1		693 (4.6)	.58
21-30	5 (45.5)	19 (43.2)			5780 (38.7)
31-40	2 (18.2)	8 (18.2)			3534 (23.7)
41-50	1 (9.1)	4 (9.1)			2605 (17.4)
51-60	2 (18.2)	8 (18.2)			1271 (8.5)
>60	1 (9.1)	4 (9.1)			442 (3.0)
Missing	0 (0)	1 (2.2)			612 (4.1)
Gender
Male	8 (73)	22 (50)	.31		9274 (62)	.55
Female	3 (27)	22 (50)			5663 (38)
Ethnicity
White	2 (18.2)	12 (27.3)	.67		4526 (30.3)	.63
African American	4 (36.4)	12 (27.3)			4586 (30.7)
Latino	3 (27.2)	11 (25.0)			2270 (15.2)
Haitian	0 (0)	4 (9.2)			822 (5.5)
African^e	1 (9.1)	1 (2.3)			731 (4.9)
Asian	0 (0)	1 (2.3)			493 (3.3)
Cape verdean	1 (9.1)	1 (2.3)			717 (4.8)
Other	1 (9.1)	2 (4.5)			792 (5.3)
Site
Urgent care	1 (9.1)	7 (15.9)	.002		2107 (14.1)	<.0001
TB clinic	0 (0)	0 (0)			354 (2.4)
Emergency dept	0 (0)	0 (0)			86 (0.6)
Drop-in clinic	0 (0)	7 (15.9)			3406 (22.8)
Inpatient	2 (18.2)	12 (27.3)			2352 (15.8)
Outpatient clinic	0 (0)	5 (11.4)			1534 (10.3)
STD clinic	3 (27.2)	9 (20.5)			2760 (18.1)
Outreach clinic	0 (0)	4 (9.1)			1033 (6.9)
Comm. clinic	4 (36.4)	0 (0)			1305 (8.7)
Comm. clinic	1 (9.1)	0 (0)			0 (0)
IDU	3 (27)	3 (7.1)	.10	4.88 (0.83, 28.68)	1717 (11.6)	.13	2.87 (0.76, 10.82)
10+ partners^f	0 (0)	4 (12.5)	1.00	–	1001 (9.4)	1.00	–
Sex for money^g	1 (9.1)	0 (0)	.21	–	541 (3.5)	.33	2.74 (0.35, 21.44)
Sex under influence^h	4 (36.4)	12 (33.3)	.40	2.67 (0.51, 13.88)	3049 (27.9)	.74	1.47 (0.43, 5.02)
STI diagnosisⁱ	0 (0)	2 (5.6)	1.00	--	600 (5.5)	1.00	–
Sex with sex worker^j	1 (9.1)	2 (5.6)	.42	2.83 (0.22, 36.38)	354 (3.2)	.30	3.00 (0.38, 23.52)
Shared needles^k	1 (9.1)	3 (7.1)	1.00	1.30 (0.12, 13.87)	1257 (8.3)	.61	1.13 (0.14, 8.82)
Sex with HIV^l	2 (18.2)	1 (2.4)	.11	9.11 (0.74, 11.72)	504 (3.3)	.05	6.60 (1.42, 30.61)^m
Sex with IDUⁿ	1 (9.1)	5 (11.9)	1.00		1388 (9.1)	1.00	1.00 (0.13, 7.84)
MSM	3 (27.3)	3 (6.8)	.09	5.13 (0.87, 30.11)	1611 (10.8)	.11	3.11 (0.82, 11.72)

Abbreviations: IDU, injection drug users; MSM, men who have sex with men.

^a Statistical significance calculated using Fisher exact test comparing cases to controls.

^b Unadjusted odds ratio comparing cases to controls.

^c Statistical significance calculated using Fisher exact test comparing cases to total nonreactive population.

^d Unadjusted GEE odds ratio comparing cases to total nonreactive population.

^e All Africans not including Cape Verdeans.

^f 10+ partners implies having sex with more than 10 partners in the past 12 months.

^g Sex for money implies engaging in sexual intercourse for pay.

^h Sex under the influence implies having sex while intoxicated with alcohol or drugs.

ⁱ History of a sexually transmitted infection (STI) diagnosis.

^j Sex with sex worker implies sexual intercourse with a partner who engages in sexual intercourse for pay.

^k Shared needles implies injecting drugs using syringes used by other people.

^l Sex with HIV implies sexual intercourse with a partner who is infected with HIV.

^m P value <.05 level.

ⁿ Sex with IDU implies sexual intercourse with a partner who in an injection drug user.

With exception of sex with an HIV-infected partner, a risk factor present in 18% of the cases and 3% of the total population (P = .05), there was no significant difference in conventional HIV risk factors between the cases and the total population.

Table 2 reveals the results of the case–control analysis. Cases were more likely to have O-negative blood (P = .04). Two cases of the 3 typed had O-negative blood, and 1 control of the 15 typed had O-negative.

Table 2.

Medical Conditions of Cases (False Positives) and Controls (Matched True Negatives)

Patient Characteristics	Total Cases, N = 11 (%)^a	Total Controls, N = 44 (%)^a	P Value^b
Recent vaccine	2 (18.2)	8 (18.2)	1.00
Influenza	1 (9.1)	8 (18.2)	.67
HPV	0 (0)	2 (4.6)	1.00
HAV	1 (9.1)	1 (2.3)	.36
Blood type
O-negative	2 (66.7)	0 (0)	.04
O-positive	1 (33.3)	7 (42.9)
B-positive	0 (0)	1 (7.1)
B-negative	0 (0)	0 (0)
A-positive	0 (0)	5 (35.7)
A-negative	0 (0)	2 (14.3)
Hepatitis markers
HAV IgG Ab	3 (42.9)	6 (75.0)	1.00
HBs Ab	3 (50.0)	5 (55.6)	1.00
HBcore Ab	1 (14.3)	3 (23.1)	1.00
HbS Ag	2 (33.3)	1 (12.5)	.54
Hepatitis C^c	2 (18.1)	7 (15.9)	1.00
RPR^d	1 (16.7)	0 (0)	.22
PPD^e	1 (33.0)	3 (60.0)	1.00
Treated for tuberculosis	1 (9.1)	0 (0)	.20
Herpes zoster	1 (9.1)	0 (0)	.20
Currently pregnant	1 (9.1)	0 (0)	.20
Alcohol use^f
None/rare	5 (83.3)	9 (64.3)	.31
2-3 drinks/d	0 (0)	4 (28.6)
>4 drinks/d	1 (16.7)	1 (17.4)
Current smoker
Yes	4 (50.0)	18 (56.3)	1.00
No	4 (50.0)	14 (43.7)

Abbreviations: HAV IgG Ab, IgG antibodies to hepatitis A virus; HBs Ab, antibodies to hepatitis B virus S surface antigen; HBcore Ab, antibodies to hepatitis B virus S core antigen; HbS Ag, hepatitis B virus S surface antigen.

^a The total N changes for each data variable, because not all patients have data available on all variables.

^b Statistical significance calculated using Fisher exact test comparing cases to controls.

^c Hepatitis C virus diagnosed by both antibodies to hepatitis C virus or hepatitis C virus viral load.

^d Rapid plasma reagin (RPR) tests for nonspecific antibodies released by cells damaged by Treponema pallidum (syphilis).

^e Purified protein derivative (PPD) tuberculin nonspecies-specific injected intradermally used to diagnosis tuberculosis.

^f Alcoholic drink defined as 2 oz of hard liquor, 5 oz of wine, and 12 oz of beer.

There were no significant differences found between cases and controls with respect to vaccination history, immunity to HAV or HBV, chronic hepatitis C infection (HCV), history of STIs, purified protein derivative status, alcohol use, or smoking history. Data availability was dependent upon which elements had been collected and recorded during each patient’s clinical encounter. Therefore, not all data variables were available for all patients. When comparing the original case–control analysis and the analysis performed with imputed responses to account for missing data, we found no difference in the level of significance of the identified associated exposures.

Discussion

Because of the anticipated increase in false positives as a result of the proposed increased testing in a low-prevalence population under the 2003 CDC guidelines, we were interested in identifying patients at increased risk of obtaining a false-positive result on the OraQuick ADVANCE HIV 1/2 test.⁴ Comparing cases to the underlying population of patients correctly testing negative, having sex with an HIV-infected partner, and testing at certain sites were more frequently encountered in individuals who falsely tested positive. In the case–control analysis, O-negative blood type was more frequently observed in cases compared to controls.

The Minnesota DPH performed a similar retrospective analysis looking at demographic and behavioral factors as potential predictors of false positives in 2004.² Unlike our study, their study found an association with age but no association with sex with an HIV-infected patient or with testing site. Their study was performed in the setting of an outbreak of false positives with a false-positive rate of 4.6%. Other studies have also demonstrated a higher false-positive rate with oral fluid testing as compared with fingerstick-based tests.^1,3,4

In our fairly small population, we did not identify false-positive cases with medical histories significant for conditions previously linked to false-positive results, including autoimmune disease, recent vaccination, liver disease, renal disease, infectious diseases, multiparity, or malignancy. While we powered our study to find an effect size of 4 or greater, there is a chance the effect was smaller than 4 or there was confounding that we could not adjust for, given the small sample size of our study.

Our study identified the O-negative blood type, sex with an HIV-infected partner, and certain sites of care as risk factors for false-positive results. There is no documentation in the literature of an association between blood type and false positive HIV test results. Associations between blood type and false-positive autoimmune disorders exist, however. Blood group antibodies have been shown to cross-react with antibodies associated with Sjogren syndrome and pemphigus vulgaris, causing false-positive tests for these conditions.^27,28 If O-negative blood type is a novel risk factor for false-positive rapid HIV tests, blood group antibodies may cross-react with the HIV antibodies of the diagnostic test.

A study among blood donors found an association between false positivity and being in at-risk sexual relationships.¹⁷ Increased exposure to HIV virus during sexual intercourse may lead to the development of antibodies. Studies performed in HIV exposed, uninfected individuals have found HIV-1-specific immunoglobulin A (IgA) and IgG antibodies in vaginal fluid and semen.^29,30 Cohort studies of sexually exposed seronegative individuals also have found serum anti-HIV IgA but not IgG.³¹

Finally, several factors could link false-positive results to testing site: differences in test operators, test batches, and population.

While some of our findings may be novel, another possibility is type 2 error, false rejection of the null hypothesis. The multiple statistical tests required to analyze several potential risk factors increase the probability of this type of error.

The study’s limitations include the small number of cases and resultant low power to detect factors associated with false-positive results. In addition, our study was based on medical record review. As such, information bias could have been introduced if cases or controls were more likely to be tested for certain conditions; however, after imputation of missing data, we found the same level of significance among the identified associated exposures.

This study identified one novel and one relatively novel potential risk factor for false-positive results on the fingerstick-based OraQuick ADVANCE HIV 1/2 test: O-negative blood type and having sex with an HIV-infected person, as well as suggesting that different test sites—and the patients they serve—may lead to varying rates of false-positive results.

Further studies should continue to investigate these and other potential sources of false positives, given the published recommendations for broader HIV testing and thus a shift toward screening low-risk populations for HIV.⁴

Footnotes

Authors’ Notes

S. B. Rifkin and J. L. Greenwald designed the study. S. B. Rifkin and L. E. Owens carried out data abstraction. S. B. Rifkin analyzed the data. All authors contributed to the manuscript.

Acknowledgments

We would like to thank Dr Anthony J. Guarino for his contributions to the article, including advice on statistical methodology and edits to the manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Centers for Disease Control and Prevention (CDC). False-positive oral fluid rapid HIV tests—New York City, 2005-2008. MMWR Morb Mortal Wkly Rep. 2008;57(24):660–665.

Jafa

Patel

Mackellar

Investigation of false positive results with an oral fluid rapid HIV-1/2 antibody test. PLoS One. 2007;2(1):e185.

OraQuick ADVANCE Rapid HIV-1/2 Antibody Test customer letter and package insert [Internet]. Bethlehem, PA: OraSure Technologies. http://www.orasure.com/docs/pdfs/products/oraquick_advance/OraQuick-Advance-Package-Insert-English.pdf. Accessed April 2011.

Centers for Disease Control and Prevention. Advancing HIV prevention: new strategies for a changing epidemic—United States, 2003. MMWR Morb Mortal Wkly Rep. 2003;52(15):329–332.

Novick

Des Jarlais

Kreek

Specificity of antibody tests for human immunodeficiency virus in alcohol and parenteral drug abusers with chronic liver disease. Alcohol Clin Exp Res. 1988;12(5):687–690.

Barthel

Wallace

. False-positive human immunodeficiency virus testing in patients with lupus erythematosus. Semin Arthritis Rheum. 1993;23(1):1–7.

Celum

Coombs

Jones

Risk factors for repeatedly reactive HIV-1 EIA and indeterminate western blots. A population-based case-control study. Arch Intern Med. 1994;154(10):1129–1137.

Doran

Parra

. False-positive and indeterminate human immunodeficiency virus test results in pregnant women. Arch Fam Med. 2000;9(9):924–929.

Mason

Guo

Detection of retroviral antibodies in primary biliary cirrhosis and other idiopathic biliary disorders. Lancet. 1998;351(9116):1620–1624.

10.

Jindal

Solomon

Burrows

. False positive tests for HIV in a woman with lupus and renal failure. N Engl J Med. 1993;328(17):1281–1282.

11.

Challakere

Rapaport

. False-positive human immunodeficiency virus type I ELISA results in low-risk subjects. West J Med. 1993;159(2):214–215.

12.

de la Calle

Yebra

Portero

López

. False positivity of antibodies against human immunodeficiency virus in a case of Q fever [in Spanish]. Rev Clin Esp. 1998;198(6):401.

13.

Hsia

. False-positive ELISA for human immunodeficiency virus after influenza vaccination. J Infect Dis. 1993;167(4):989–990.

14.

Bronze

Warr

Spigel

Smith

Smalley

. False-positive enzyme immunoassay for human immunodeficiency virus due to acute cytomegalovirus infection. Clin Infect Dis. 1998;27(1):221–222.

15.

Silverstein

Aviles

Vehaskari

. False-positive human immunodeficiency virus antibody test in a dialysis patient. Pediatr Nephrol. 2004;19(5):547–549.

16.

Ghosh

Javeri

Mohanty

Parmar

Surati

Joshi

. False-positive serological tests in acute malaria. Br J Biomed Sci. 2001;58(1):20–23.

17.

Araujo

PRB

Albertoni

Arnoni

Rubella vaccination and transitory false-positive test results for human immunodeficiency virus Type 1 in blood donors. Transfusion. 2009;49(11):2516–2517.

18.

Hussain

Sinha

Katoch

Serum samples from patients with mycobacterial infections cross-react with HIV structural proteins Gp41, p55 and p18. Lepr Rev. 2007;78(2):137–147.

19.

Facente

Dowling

Vittinghoff

Sykes

Colfax

. False positive rate of rapid oral fluid HIV tests increases as kits near expiration date. PLoS One. 2009;4(12):e8217.

20.

Gasasira

Dorsey

Kamya

False-positive results of enzyme immunoassays for human immunodeficiency virus in patients with uncomplicated malaria. J Clin Microbiol. 2006;44(8):3021–3024.

21.

Lejon

Ngoyi

Ilunga

Low specificities of HIV diagnostic tests caused by Trypanosoma brucei gambiense sleeping sickness. J Clin Microbiol. 2010;48(8):2836–2839.

22.

Mao

Obi-Gwacham

Hill

Ryan

Hendricks

McKellar

. Acute Dengue fever causes false-positive reactivity in OraQuick rapid HIV-1/2 antibody test. J Acquir Immune Defic Syndr. 2010;55(5):641.

23.

Everett

Baisely

McNerney

Association of schistosomiasis with false-positive HIV test results in an African adolescent population. J Clin Microbiol. 2010;48(5):1570–1577.

24.

Chao

Sheffield

Wendel

Jr Ansari

McIntire

Roberts

SW.

Risk factors associated with false positive HIV test results in a low-risk urban obstetric population. J Pregnancy. 2012;2012:841979.

25.

Shida

Takahashi

Fujishima

False-positive human immunodeficiency virus antibody test and autoimmune hemolytic anemia in a patient with angioimmunoblastic T-cell lymphoma. Intern Med. 2011;50(20):2383–2387.

26.

Enders

. The performance of the full information maximum likelihood estimator in multiple regression models with missing data. Educ Psychol Meas. 2001;61:713–740.

27.

Lee

Silvestrini

Fulcher

. False-positive intercellular cement substance antibodies due to group A/B red cell antibodies: frequency and approach. Pathology. 2010;42(6):574–577.

28.

Goldblatt

Beroukas

Gillis

Antibodies to AB blood group antigens mimic anti-salivary duct autoantibodies in patients with limited sicca symptoms. J Rheumatol. 2000;27(10):2382–2388.

29.

Buchacz

Parekh

Padian

HIV-specific IgG in cervicovaginal secretions of exposed HIV-uninfected female sexual partners of HIV-infected men. AIDS Res Hum Retroviruses. 2001;17(18):1689–1693.

30.

Lo Caputo

Trabattoni

Vichi

Mucosal and systemic HIV-1-specific immunity in HIV-1-exposed but uninfected heterosexual men. AIDS. 2003;17(4):531–539.

31.

Mestecky

. Humoral immune responses to the human immunodeficiency virus type-1 (HIV-1) in the genital tract compared to other mucosal sites. J Reprod Immunol. 2007;73(1):86–97.