The objective of this study was to assess the rate and possible reasons for false-negative (FN) reflex human papillomavirus (HPV)-DNA tests.
The objective of this study was to assess the rate and possible reasons for false-negative (FN) reflex human papillomavirus (HPV)-DNA tests.
The authors reviewed all ThinPrep cervical specimens that were submitted for reflex HPV-DNA testing using the Digene Hybrid Capture II (HC2) method from January 2002 to January 2004. Follow-up biopsies were reviewed. The results were considered HPV-FN if the HPV-DNA test was negative and the biopsy was positive for grade ≥2 cervical intraepithelial neoplasia (CIN2+), and the results were considered true positive (HPV-TP) if the HPV-DNA test was positive and the biopsy showed CIN2+. HPV-FN cases were compared with HPV-TP cases regarding the grade and extent of CIN, the number of abnormal cells on the original ThinPrep slide, and the presence of amplifiable, viral DNA on biopsy.
In total, 1520 (66%) of 2309 patients who had diagnoses of atypical squamous cells of undetermined significance (ASCUS) were negative for HPV DNA and 789 patients of 2309 patients (34%) were positive for HPV DNA. Three hundred sixteen women (40%) who had a positive HPV-DNA test underwent a biopsy. Of those, 36 biopsies (11%) showed CIN2+ (HPV-TP), and 154 biopsies (66%) showed CIN1. Cervical tissue was available for review from 82 women who had negative HPV-DNA tests; of these, 6 tissue samples (7%) showed CIN2+ (HPV-FN), and 13 tissue samples (16%) showed CIN1. Therefore, in the total ASCUS population that was triaged with reflex HPV testing, there were at least 42 women who were diagnosed with CIN2+, for an estimated CIN2+ FN fraction of 14% (6 of 42 women). HPV-FN lesions were smaller (but the difference was not statistically significant) and shed significantly fewer abnormal cells than HPV-TP cases. Polymerase chain reaction testing for viral DNA in the biopsy was detected in 3 of 6 women who had HPV-FN results; none of those positive results demonstrated a viral type that was not included in the Digene probes.
Although the rate of FN high-grade lesions was significantly higher than that reported in the ASCUS/Low-grade Squamous Intraepithelial Lesion Triage trial, most missed lesions were small and shed few abnormal cells. It was assumed that those lesions were either in early stages or in regressing stages, which made their clinical significance uncertain. Cancer (Cancer Cytopathol) 2007. © 2007 American Cancer Society.
Quality-management programs in cytopathology practice long have been preoccupied with false-negative (FN) Papanicolaou (Pap) tests because of their potential health risk to patients and their litigation risk to the laboratory.1–3 Practice guidelines and federal regulations mandate procedures to identify FN results and to estimate their rates and their probable causes. Utilization of reflex human papillomavirus (HPV)-DNA testing introduces new types of errors, including FN/false-positive (FP) results for which, to date, there are no clear guidelines on reporting and no understanding of their significance or possible causes. Studies in these areas likely will help in error prevention and reduction, thereby alleviating the risk to patients and laboratories. The way to reduce the risk of litigation is not to raise expectations of 100% (or near 100%) sensitivity of the reflex HPV-DNA test; rather, it must be acknowledged that humans are fallible, and the discrepancy between routine practice (effectiveness of the test) and the rigor of controlled clinical trials (efficacy) must be recognized.4–6 For the current study, we examined reflex HPV-FN results that were obtained under routine clinical practice and attempted to estimate the rates of FN cases and to determine the possible reasons for FN HPV-DNA test results (HPV-FN) by comparing the cytologic/histologic characteristics of those missed cases with the true positive (TP) cases (HPV-TP).
This retrospective, cytohistologic correlational study was approved by the University of Alabama at Birmingham (UAB) Institutional Review Board. UAB Medical Center (UABMC) is Alabama's major tertiary care center, serving approximately 35,000 patients annually, and was 1 of the 4 Atypical Squamous Cells of Undetermined Significance (ASCUS)/Low-grade Squamous Intraepithelial Lesion Triage (ALTS) trial clinical centers. The UABMC laboratory receives approximately 23,000 Pap tests a year with an average ASCUS rate of 10%, a squamous intraepithelial lesion (SIL) rate of 5%, and an ASCUS:SIL ratio of 2:1. The mean annual FN fraction of screening is 5% and is calculated as the estimated FN/(estimated FN plus; true negative [TN]). Since August 2001, ThinPrep Pap tests (Cytyc, Boxborough, Mass) gradually have increased to reach a current rate of 90% of our total volume of Pap tests. The remaining Pap tests were conventional smears that were received mostly from the colposcopy clinic at UABMC. Slides were screened by 7 cytotechnologists (experience ranged from 5 years to 25 years) and were reviewed by 5 cytopathologists (experience ranged from 3 years to 13 years). Patients were triaged according to the 2002 American Society for Colposcopy and Cervical Pathology guidelines.7 Three surgical pathologists (experience ranged from 4 years to 15 years) read the cervical biopsies. Clinical and laboratory data were archived in the Cerner Classic laboratory information system (Kansas City, Mo).
The study included all cervical ThinPrep Pap tests that were performed at our institution from January 2002 to January 2004. Conventional cervical smears and all vaginal (conventional and ThinPrep) specimens were excluded from the study. Only results from biopsies that were obtained after the ThinPrep Pap test were reviewed and correlated with cytologic and virologic findings. If patients had multiple biopsies or Pap tests, then the highest diagnosis was used in the analysis. Both the cytologic and histologic slides of FN and TP cases were retrieved. ThinPrep Pap tests were reviewed to assess adequacy, to confirm the diagnosis of ASCUS, and to determine the total number and percentage of abnormal cells. Abnormal squamous cells were screened for systematically and were counted on the entire slide. After counting the number of cells along 2 perpendicular dimensions using a ×60 objective and determining the average number of cells per field (n), the total number of squamous cells present in the entire slide was determined using the following formula:
where N is the total number of cells, n is the average number of squamous cells per field, r ac is the radius of the ThinPrep circle, and r ao is the radius of the ocular that is equal to [field number]/[power of objective] = 22/60. The histologic slides were reviewed to confirm the diagnosis, to assess endocervical gland involvement, and to measure the linear extent of intraepithelial lesions using the scale on the microscopic stage (precision, 0.1 mm).
Reflex HPV testing for high-risk viral DNA (HPV types 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, and 68) was performed using the Digene HC2 method (Digene Corporation, Gaithersburg, Md). Cases were considered positive at the 1 pg/mL level. All 6 grade ≥2 cervical intraepithelial neoplasia (CIN2+) HPV-FN cases, 7 CIN2+ HPV-TP cases, 14 HPV-TN cases (ASCUS negative on reflex HPV-DNA test and negative on biopsy), and 9 CIN1 HPV-FN cases were retested for HPV DNA on biopsy material using polymerase chain reaction (PCR) (L1 GP5-positive/GP6-positive primers), as described previously.8 Sequencing of the amplimer was performed in cases of variant HPV types using an automated ABI Prism sequencer at the UAB DNA Sequence Core Facility.
The FN fraction (rate) = HPV-FN/(HPV-FN + HPV-TP), where HPV-FN = CIN2+ detected through reviewing cervical tissue obtained from reflex HPV DNA-negative cases for any reason. HPV-TP = CIN2+ cases detected through triage of reflex HPV DNA-positive cases. Because the number in each group was small, we used a nonparametric test (Mann-Whitney U test) to compare HPV-FN cases with HPV-TP cases.
Figure 1 is a flow chart of ThinPrep tests that were received during the study period. In total, 1520 of 2309 ASCUS cases (66%) were negative for HPV DNA, and 789 ASCUS cases (34%) were positive. Sixty-one percent of all ASCUS patients were aged ≥35 years. In total, 316 women (40%) who had ASCUS-HPV-positive results underwent a biopsy. Of these, 36 women (11%; HPV-TP results) showed CIN2+, and 154 women (66%) showed CIN1. Cervical tissue was available (hysterectomies, biopsies, loop biopsy, or endocervical curetting) for review in 82 women who had ASCUS-HPV-negative results; of these, 6 women (7%) showed CIN2+ (HPV-FN), and 14 women (17%) showed CIN1 (Fig. 1). In the ASCUS population that was triaged through reflex HPV testing, there were at least 42 CIN2+ results, representing 36 TP results and 6 FN results and producing an estimated CIN2+ FN fraction of 14% (6 of 42 results).
There was no difference in age between women who had HPV-TP results (mean age ± standard deviation [SD], 32 ± 13.5 years; age range, 17–68 years) and women who had HPV-FN results (mean age ± SD, 30 ± 13 years; age range, 17–53 years). In our study group, 29 patients had previous gynecologic history records available for review. Of these, only 3 patients had 3 consecutive, normal Pap tests, including 2 patients from the HPV-TP group and 1 patient from the HPV-FN group. Twenty-six patients cases (23 from the HPV-TP group and 3 from the HPV-FN group) had at least 1 abnormal Pap test (ASCUS and above). There was no significant difference between the HPV-TP group and the HPV-FN group with regard to the history of abnormal Pap tests.
The mean ± SD number of abnormal cells per specimen differed significantly between the 2 groups and was 54 ± 47 cells (range, 8–200 cells) for women with HPV-TP results versus 24 ± 14 cells (range, 8–39 cells) for women with HPV-FN results (P < .016; Mann-Whitney U test) (Fig. 2). The mean ± SD linear extension for CIN of HPV-TP was larger but did not differ statistically compared with that of HPV-FN (HPV-TP: 4.8 ± 4.3 mm; range, 0.9–22.3mm; HPV-FN: 3.1 ± 2.6 mm; range, 1.0–8.0 mm). There was no significant correlation between linear extension and the number of abnormal cells (Fig. 3). The total number of cells (mean ± SD) did not differ significantly between the 2 groups (HPV-TP: 36, 269 ± 19,785 cells; range, 5737–75,844 cells; HPV-FN: 27,099 ± 15, 655 cells; range, 9515–46,446 cells). The frequency of endocervical involvement—20 of 36 cases (55%) for HPV-TP versus 2 of 6 cases (33%) for HPV-FN—differed significantly.
There was no statistically significant correlation between the number of abnormal cells and linear extension. There was no significant difference in the number of abnormal cells in those women with or without cervical gland involvement.
Three of the women with HPV-FN results had positive PCR tests performed on biopsy specimens. B-actin controls were unsatisfactory in 1 woman, and 2 women were negative. All 9 randomly selected women who had HPV-TP CIN2+ results also had positive PCR tests, all 14 women who had HPV-FN CIN-1 results were negative, and all 8 women who had HPV-TN results were negative (Fig. 4). On sequencing the PCR products, all 3 women with HPV-FN results showed mixed viral DNA, and all types were included in the Digene HC2 probe.
Of all quality-control indices related to HPV infection testing, the FN fraction is particularly concerning because of potential medical risk to patients and litigation risk to laboratories from FN results.2 The 2002 American Society for Colposcopy and Cervical Pathology guidelines recommend that women with ASCUS who test negative on the reflex HPV-DNA test undergo a second Pap test at 12 months.7 During the transition from the old to the new guidelines, in our files, we identified 82 women with reflex HPV-DNA-negative results who underwent biopsies, probably because of high clinical risk. Six women showed high-grade lesions, resulting in an FN (for CIN2+) HPV-DNA test rate of 14%. The negative fraction probably would have increased if the rest of the women with HPV-negative results had undergone biopsies. Highlighting this finding is not in any way a call for yet another improvement in the sensitivity of HPV testing—such an increase undoubtedly would decrease the specificity and increase the number of women subjected to colposcopy. Instead, our results demonstrate the shortcomings of HPV testing as delivered outside the rigors of clinical trials (ie, in the real world).
Reported HPV-FN fractions generally varied from 0% to 17%.9–11 Factors that have been used to explain intrastudy variability have included the type of cervical device, review of histologic outcome by experts, clinical history of prior CIN, prevalence of CIN2+ in the study population, HPV testing type, test reproducibility, and age and size of the study population.10 In the current study, we used HC2 at a cut-off level of 1 pg/mL level, and 2 pathologists reviewed the histologic sections in additional to the original pathologist. Our study population, as noted above, had more perimenopausal and postmenopausal women than the population in the ALTS trial. Finally, the nature of the lesions may explain in part the FN results that we encountered during cytologic screening or reflex HPV-DNA testing. The myriad of factors that possibly may affect the performance of HPV during routine practice calls for establishing national performance benchmarks.
Numerous studies have demonstrated that lesions that were missed during routine cytologic screening were predominantly small.12–14 In a retrospective review of conventional Pap smears in which high-grade lesions were missed during initial screening, Hatem and Wilbur reported that the majority of these smears (12 of 16) had <100 abnormal cells, and only 1 smear had >100 abnormal cells.6 In a recent case-control study, O'Sullivan et al. observed that smears with ≤50 abnormal cells were 26 times more likely to be reported as negative for SIL than smears with >200 abnormal cells.15
Like the cytologic FN lesions, virologic FN lesions are smaller in size and shed significantly fewer cells than TP lesions, as observed in the current study. In the ALTS trial, high-grade lesions that were detected through reflex HPV testing generally were small and lacked the known features that have been associated with invasion.16 More noteworthy, during the ALTS exit visits, lesions that were HPV-negative were smaller than HPV-positive lesions.16 Short-term fluctuations in the detection of HPV DNA,17 viral clearance before resolution of the lesion,18 viral latency, or infection with viral types that are not included in the Digene probe may explain this discrepancy between histologic/virologic results. In the 3 cases that were positive for viral DNA in our study, none harbored a viral type that was not included in the Digene probes. Most likely, therefore, these lesions were missed because of sampling error.
This study is not without limitations, particularly given the small percentage (40%) of patients who underwent a biopsy after a positive HPV-DNA test and the selective nature of the patients who underwent a biopsy after a negative HPV-DNA test. Possible reasons for the small percentage of patients who had tissue confirmation after a positive test include that some patients were lost to follow-up for a short period, particularly those patients recruited late in the study, and/or that patients had normal colposcopic findings.
Regardless of these limitations, there seems to be a difference, although it is minor, between high-grade lesions after a positive HPV test versus lesions after an FN HPV test. The majority of the latter lesions shed few abnormal cells, and they are small, as observed in this study, or focal, as reported recently by Jastania et al.19 Presumably, these lesions represent either early or regressing stages of cervical disease, which makes their clinical significance uncertain.