Implications of False-Positives for Future Cancer Screenings

Abstract

BACKGROUND

False-positive cancer screenings may affect a patient’s willingness to obtain future screening.

METHODS

We conducted logistic regression analysis of 450,484 person-years of electronic medical records (2006–2015), in 92,405 individuals aged 50–75 years. Exposures were false-positive breast, prostate or colorectal cancer screening test results (repeat breast imaging or negative breast biopsy ≤3 m after screening mammography; repeat PSA test ≤3 m after PSA test result ≥4.0 ng/mL or negative prostate biopsy ≤3 m after any PSA result; negative colonoscopy [without biopsy/polypectomy] ≤6 m after positive fecal occult blood test [FOBT]). Outcomes were up-to-date status with breast or colorectal cancer screenings. Covariates included prior screening history, clinical information (e.g., family history, obesity, smoking), comorbidity and demographics.

RESULTS

Women were more likely to be up-to-date with breast cancer screening if they previously had false-positive mammography (adjusted odds ratio [AOR]=1.43, 95% CI=1.34–1.51 without breast biopsy; AOR=2.02, 95% CI=1.56–2.62 with breast biopsy; both P<0.001). The same women were more likely to be up-to-date with colorectal cancer screening (AOR=1.25–1.47 depending on breast biopsy, both P<0.001). Men who previously had false-positive PSA testing were more likely to be up-to-date with colorectal cancer screening (AOR=1.22, P=0.039 without prostate imaging/biopsy; AOR=1.60, P=0.028 with imaging/biopsy). Results were stronger for individuals with more false-positives (all P≤0.005). However, women with previous false-positive colorectal cancer FOBT screening were less likely to be up-to-date with breast cancer screening (AOR=0.73, P<0.001).

CONCLUSIONS

Patients who previously had a false-positive breast or prostate cancer screening test were more likely to engage in future screening.

False-positive cancer screening test results are common. Over 10 years, approximately 50–61% of women undergoing annual mammography and 10–12% of men undergoing regular PSA testing will experience a false-positive result.^1–4 Similarly, 23% of individuals undergoing regular fecal occult blood testing will experience a positive result with negative follow-up colonoscopy.⁵ Efforts to reduce false-positives are cited in support of biennial (vs. annual) mammography,⁶ while overdiagnosis—which might be considered a false-positive because patients are treated unnecessarily—is cited in recommendations against routine prostate cancer screening as well as breast and colorectal cancers for some populations.^6–9

False-positive cancer screening test results may affect individuals’ willingness to continue screening for cancer in the future; about 40% of women experiencing a false-positive mammogram labeled the experience as “very scary” or the “scariest time of my life.”¹⁰ However, implications are unclear. Individuals may believe they “almost had cancer” and become more interested in future screening. This hypothesis is supported by surveys suggesting that women consider false-positives an acceptable consequence of mammography,¹¹ with nearly all women reporting satisfaction with the initial screening test.¹⁰ Retrospective work suggests that women with false-positive mammography were more likely to return for subsequent screening.¹² Alternatively, individuals may recognize false-positive test results as a harm and become less interested in future screening. Some studies have found lower screening reattendance rates following false-positive mammography,^6,13,14 including recent work,¹⁵ but reasons are unclear.⁶ Consequences may vary by population; a meta-analysis showed 7% higher odds of future screening for US women after false-positive mammography, but 37% lower odds in Canadian women.¹⁶ Lower reattendance was also found in Europe following false-positive mammography,^17–19 particularly among women who underwent biopsy.¹⁸ Finally, false-positives may have no impact.^20,21 Little is known about implications of false-positive test results for colorectal and prostate cancers.

Historically, investigation of this issue has been limited by available data. Prospective studies are difficult because of need for long-term follow-up data. Medicare administrative claims are not ideal because many individuals experience a false-positive before age 65 years.^1–5 Commercial insurer data are limited by average enrollment of <3 years per patient.²² Administrative data also lack important clinical information, such as family history, obesity and tobacco, which may influence cancer risk and screening decisions.

In this study, we utilized 10 years of electronic medical records data to analyze the association between prior receipt of a false-positive cancer screening test result and future participation in routine cancer screenings, adjusted for a wide variety of patient risk factors and demographics.

METHODS

Data and Patients

We used electronic medical records (EMR) data from 2006–2015 for the Cleveland Clinic Healthcare System. The system has 1 large academic medical center, 13 regional hospitals, 21 family health centers and >75 outpatient locations. We considered encounters, diagnoses (on at least 2 encounters ≥30 d apart),²³ completed procedures, surgeries, prescriptions, preventive health maintenance, smoking status, family history, demographics, and address history (free text notes were not considered). Because we only had information on care received within the system, we identified patients who received continuing primary care, and therefore were more likely to have complete cancer screening data. Specifically, we required Ohio residence, ≥5 years of EMR data, and visits to an internal medicine or family medicine provider in ≥3 different years. For example, if a patient had primary care visits in each year of 2009–2012 and 2014, then the patient was included from 2009–2014 (but not earlier or later years). Results were robust to variation in these requirements. For each patient, we used the first 2 years’ data to establish cancer screening history (e.g., annual vs. biennial mammography and receipt of false-positive test results) and remaining years’ data for analyses. Patients were aged 50–75 years and had no known history of cancer (other than non-melanoma skin cancer) or bilateral mastectomy.

Cancer screening experiences

False-positive test results

Patients were considered to have undergone a false-positive cancer screening test if they received a result that might raise concern but did not result in cancer diagnosis. Specifically, patients with a positive fecal occult blood test (FOBT) and negative follow-up colonoscopy within 6 months, without biopsy or polypectomy, were considered to have experienced false-positive colorectal cancer screening; we did not consider false-positives for patients who underwent initial colonoscopy because repeat screening would not be expected during the study period. Women with repeat breast imaging (mammogram, ultrasound, MRI) or negative breast biopsy ≤3 months after screening mammography were considered to have experienced false-positive breast cancer screening. Finally, men without a history of benign prostate hyperplasia, lower urinary tract symptoms, or prostatitis who had a PSA test result ≥4.0 ng/mL and a repeat PSA test within 3 m, or any PSA test result with negative follow-up prostate biopsy within 3 months, were considered to have experienced false-positive prostate cancer screening. We reasoned that, even when prostate cancer screening was not recommended, men who had obtained it and experienced a false-positive might change their willingness to undergo guideline-recommended colorectal cancer screening. Most cancer screenings were ordered by request, but periodic organized initiatives (e.g., mailings) informed some patients of screening eligibility and invited them to schedule appointments. For mammography, medical assistants often facilitated screening before a physician entered the room. Diagnostic cancer testing (indicated by procedure names; e.g., “Occult Blood Diagnostic”) was not considered.

Up-to-date status

At the end of each year, we determined up-to-date status for each patient with US Preventive Services Task Force-recommended cancer screenings. (Patients without visits were considered not to have received screening in that year, but could be up-to-date based on previous years.) Patients were considered up-to-date with colorectal cancer screening following receipt of FOBT within 1 year, flexible sigmoidoscopy within 5 years (with or without mid-interval FOBT) or colonoscopy within 10 years. Women were considered up-to-date with breast cancer screening following receipt of mammography or breast MRI within 2 years (1 year in sensitivity analyses). Up-to-date status with prostate cancer screening was not included because guidelines sometimes recommended against routine screening;^4,24–27 however, receipt of prostate cancer screening in the past 2 years was addressed as supplementary analysis.

Covariates

For each patient-year, we identified cardiovascular risk factors (smoking status, body mass index, blood pressure, lipids, HbA1c); comorbidity (Charlson score=0, ≥1);²³ family history of colorectal or breast cancer (first-degree, second-degree, male breast cancer, unknown); receipt of influenza vaccination; demographics (age, race, ethnicity, marital status, health insurance type, census block group median income by race); primary care site and physician specialty (internal medicine, family medicine); primary care utilization (number of visits in previous 365 days); employee/dependent of the health system; and year. For colorectal cancer screening, we also included current anticoagulation prescriptions (yes/no) and diagnosis with familial adenomatous polyposis/hereditary nonpolyposis colon cancer, Crohn’s colitis and ulcerative colitis. For breast cancer screening, we also included current postmenopausal hormone replacement therapy prescriptions (yes/no) and each woman’s breast cancer screening frequency (based on the mode of historic data—annual, biennial, less than biennial). In sensitivity analyses, breast cancer screening frequency was redefined as the average of number of breast cancer screenings per year of historic data (≤0.5, >0.5) and the total number of breast cancer screenings in historic data (0, 1, 2, ≥3; max. 1/year). For these analyses, “historic” was defined as the first available year of EMR data for the patient since her 50th birthday (or earlier, if known receipt of breast cancer screening) through the current year under consideration.

Analyses

We employed logistic regression models to analyze the association between cancer screening experiences and up-to-date status with colorectal cancer screening (stratified by gender) or breast cancer screening (3 separate outcomes—one per cancer type/gender). For colorectal cancer screening, we included prior breast and prostate cancer false-positives, but excluded colorectal cancer false-positives because for most individuals, we only expected 1 colonoscopy during the study. For breast cancer screening, we included prior breast and colorectal cancer false-positives. We examined both unadjusted models (which only considered the independent variable of interest—prior receipt of a false-positive cancer screening) and models adjusted for covariates. An adjusted odds ratio (AOR) >1 would suggest that individuals with prior false-positives were more likely to undergo future cancer screenings. We expected to find AORs >1 across cancer types, indicating that individuals increased their intensity of screening practices following a false-positive. In alternate models, we considered the number of prior false-positives from breast and prostate cancer screenings (0, 1, ≥2) (multiple false-positives from colorectal cancer screening would not be expected) and fixed effects for each physician with ≥100 patients who underwent screenings each year. As a secondary outcome, we considered the association between up-to-date status with colorectal or breast cancer screening and receipt of other cancer screenings (regardless of screening test outcome).

To consider the possibility that EMR may have been incomplete in early years, we also conducted analyses that eliminated 2006–2008 data. Finally, in 2 separate models, we tested the association between prior receipt of a false-positive cancer screening and diagnosis with hyperlipidemia or diabetes, outcomes that should be unrelated to cancer screening status (expected AOR=1) but were considered because both cancer screening and lipids/diabetes diagnoses (which require patients to obtain a blood test) may proxy for healthcare system engagement.

Standard errors were clustered by individual. Significance tests were 2-sided at the 5% level. This study was approved by the Cleveland Clinic Institutional Review Board.

RESULTS

We identified 92,405 patients aged 50–75 years across 450,484 person-years. Tables 1–2 show summary statistics. Among at-risk individuals, 62% and 85% were up-to-date with colorectal and breast cancer screenings, respectively. About 29% of at-risk individuals had a prior breast cancer false-positive, 3% had a prior prostate cancer false-positive, and 2% had a prior colorectal cancer false-positive.

Table 1.

Summary Statistics, 2006–2015

	All Mean (SD)*	Female Mean (SD)*	Male Mean (SD)*
N person-years†	450,484	262,242	188,242
N individuals	92,405	52,871	39,534
Below values in percent (%)
Colorectal cancer screening
Up-to-date with colorectal cancer screening	61.5 (48.7)	61.5 (48.7)	61.6 (48.6)
Colonoscopy within 10 y	53.1 (49.9)	53.2 (49.9)	52.9 (49.9)
Flexible sigmoidoscopy within 5 y	0.5 (7.0)	0.5 (7.1)	0.5 (6.8)
FOBT within 1 y	8.0 (27.1)	7.8 (26.7)	8.2 (27.5)
Ever had false-positive colorectal cancer screening
Positive FOBT (negative follow-up colonoscopy within 6 m, without biopsy/polypectomy)‡	2.1 (14.2)	1.8 (13.2)	2.4 (15.4)
Breast cancer screening
Up-to-date with breast cancer screening (past 2 y)	-	84.6 (36.1)	-
Ever had false-positive breast cancer screening
Repeat breast imaging within 91 d (no biopsy)	-	28.8 (45.3)	-
Breast biopsy	-	1.5 (12.3)	-
Prostate cancer screening
Received prostate cancer screening in past 2 y	-	-	74.7 (43.5)
Ever had false-positive prostate cancer screening
PSA ≥4.0 ng/mL (repeated within 91 d, no imaging/biopsy)	-	-	2.8 (16.5)
Prostate imaging/biopsy	-	-	0.5 (7.1)

^*All P<0.001 except up-to-date with colorectal cancer screening (P=0.49).

^†Each individual had an additional 2 years’ EMR data, used to inform cancer screening history.

^‡We did not consider false-positives for patients who underwent initial colonoscopy because repeat screening would not be expected during the study period.

Table 2.

Additional Summary Statistics

	Up-to-date with colorectal cancer screening	P	Up-to-date with breast cancer screening	P
Females [N=262,242]
Ever had false-positive breast cancer screening
No	62.1%	<0.001	88.5%	<0.001
Repeat breast imaging within 91 d (no biopsy)	68.8%		95.0%
Breast biopsy	71.0%		96.2%
Ever had false-positive colorectal cancer screening
No	†		90.1%	0.75
Positive FOBT (negative follow-up colonoscopy within 6 m, without biopsy/polypectomy)*	†		90.3%
Males [N=188,242]
Ever had false-positive prostate cancer screening
No	62.7%	<0.001	-
PSA ≥4.0 ng/mL (repeated within 91 d, no imaging/biopsy)	70.4%
Prostate imaging/biopsy	75.9%

^*We did not consider false-positives for patients who underwent initial colonoscopy because repeat screening would not be expected during the study period.

^†Repeat screening after a false-positive would not be expected during the study period.

In both unadjusted and adjusted models, a prior false-positive breast or prostate cancer test was strongly associated with being up-to-date with cancer screenings (Table 3). Women were more likely to be up-to-date with breast cancer screening if they previously had a false-positive mammogram (AOR=1.43, 95% CI=1.34–1.51 without breast biopsy; AOR=2.02, 95% CI=1.56–2.62 with negative biopsy; both P<0.001) (column 4). The same women were also more likely to be up-to-date with colorectal cancer screening (AOR=1.25, 95% CI=1.18–1.32 without breast biopsy; AOR=1.47, 95% CI=1.23–1.77 with negative biopsy; both P<0.001) (column 2). Similarly, men who previously had a false-positive PSA test were more likely to be up-to-date with colorectal cancer screening (AOR=1.22, 95% CI=1.01–1.48, P=0.039 without prostate imaging/biopsy; AOR=1.60, 95% CI=1.05–2.42, P=0.028 with negative imaging/biopsy) (column 6). However, women were less likely to be up-to-date with breast cancer screening if they previously had a false-positive colorectal cancer FOBT screening (AOR=0.73, 95% CI=0.61–0.81, P<0.001) (column 4).

Table 3.

Association between past false-positives on cancer screening tests and future receipt of cancer screenings^*

	Females								Males
	Up-to-date with colorectal cancer screening				Up-to-date with breast cancer screening				Up-to-date with colorectal cancer screening
Column	(1)		(2)		(3)		(4)		(5)		(6)
	Unadjusted		Adjusted		Unadjusted		Adjusted		Unadjusted		Adjusted
	OR (95% CI)	P	AOR (95% CI)	P	OR (95% CI)	P	AOR (95% CI)	P	OR (95% CI)	P	AOR (95% CI)	P
Ever had false-positive breast cancer screening
Repeat breast imaging within 91 d (no biopsy)	1.35 (1.30–1.39)	<0.001	1.25 (1.18–1.32)	<0.001	2.42 (2.29–2.56)	<0.001	1.43 (1.34–1.51)	<0.001
Breast biopsy	1.50 (1.32–1.70)	<0.001	1.47 (1.23–1.77)	<0.001	3.25 (2.52–4.19)	<0.001	2.02 (1.56–2.62)	<0.001
Ever had false-positive colorectal cancer screening†
Positive FOBT with negative follow-up colonoscopy within 6 m, without biopsy/polypectomy†					0.97 (0.81–1.15)	0.69	0.73 (0.61–0.81)	<0.001
Ever had false-positive prostate cancer screening
PSA ≥4.0 ng/mL (repeated within 91 d, no imaging/biopsy)									1.42 (1.27–1.58)	<0.001	1.22 (1.01–1.48)	0.039
Prostate imaging/ biopsy									1.87 (1.43–2.45)	<0.001	1.60 (1.05–2.42)	0.028
Receipt of prior cancer screening
Up-to-date with breast cancer screening (past 2 y)			1.78 (1.68–1.90)	<0.001
Up-to-date with colorectal cancer screening‡							1.81 (1.69–1.93)	<0.001
Had prostate cancer screening (past 2 y)											1.34 (1.27–1.42)	<0.001
Psuedo-R2	0.03		0.51		0.29		0.41		0.02		0.53
N person-years	262,242		262,242		262,242		262,242		188,242		188,242
N individuals	52,871		52,871		52,871		52,871		39,534		39,534

^*Adjusted models controlled for covariates as described in manuscript.

^†We did not consider false-positives for patients who underwent initial colonoscopy because repeat screening would not be expected during the study period.

^‡Receipt of FOBT within 1 year, flexible sigmoidoscopy within 5 years or colonoscopy within 10 years.

A greater number of prior breast or prostate cancer false-positives was associated with a trend toward being up-to-date with screening (Table 4). Women who had 1 and ≥2 false-positive breast cancer screenings had 16% and 50% higher odds, respectively, of being up-to-date with colorectal cancer screening than women with no false-positives (AOR=1.16, 95% CI=1.10–1.23 and AOR=1.50, 95% CI=1.35–1.67, respectively, P for trend<0.001) (column 1). The same women had 34% and 163% higher odds, respectively, of being up-to-date with breast cancer screening, as compared to women with no false-positives (P for trend<0.001) (column 2). Similarly, men with 1 and ≥2 false-positive prostate cancer screenings had 18% and 40% higher odds, respectively, of being up-to-date with colorectal cancer screening (P for trend=0.005) (column 3). In supplemental analysis, men who previously had a colorectal cancer FOBT false-positive were more likely to have received prostate cancer screening in the past 2 y (P<0.001).

Table 4.

Number of false-positives^*

	Females				Males
	Up-to-date with colorectal cancer screening		Up-to-date with breast cancer screening		Up-to-date with colorectal cancer screening
Column	(1)		(2)		(3)
	AOR (95% CI)	P for trend	AOR (95% CI)	P for trend	AOR (95% CI)	P for trend
Number of false-positive breast cancer screenings
0	Ref.		Ref.
1	1.16 (1.10–1.23)	<0.001	1.34 (1.26–1.43)	<0.001
≥2	1.50 (1.35–1.67)		2.63 (2.23–3.10)
Number of false-positive prostate cancer screenings
0					Ref.
1					1.18 (1.05–1.32)	0.005
≥2					1.40 (1.11–1.71)
Psuedo-R2	0.52		0.40		0.59
Number of person-years	262,242		262,242		188,242
Number of individuals	52,871		52,871		39,534

^*Nnumber of false-positive FOBTs was not assessed, because average risk patients with 1 false-positive and a negative follow-up colonoscopy would not be expected to re-screen during the sample period. Adjusted for covariates as described in manuscript.

Individuals who received one cancer screening were more likely to obtain other screenings (Table 3). Women who were up-to-date with breast cancer screening were more likely to be up-to-date with colorectal cancer screening (AOR=1.78, 95% CI=1.68–1.90, P<0.001) (column 2), and similarly, women who were up-to-date with colorectal cancer screening were more likely to be up-to-date with breast cancer screening (AOR=1.81, 95% CI=1.69–1.93, P<0.001) (column 4). Men who had received prostate cancer screening in the prior 2 years were more likely to be up-to-date with colorectal cancer screening (AOR=1.34, 95% CI=1.27–1.42, P<0.001) (column 6).

Results were robust to definition of breast cancer screening history, controls for each physician and exclusion of 2006–2008 data. As expected, adjusted odds ratios for alternate outcomes of diagnosis with hyperlipidemia or diabetes were not significantly different from 1.

DISCUSSION

In this study examining electronic medical records for >90,000 individuals followed over approximately 7 years, past experiences with cancer screening were positively associated with receipt of future cancer screenings. In general, those who were up-to-date with one cancer screening were more likely to obtain other screenings, an intuitive finding noted by others.^28–30 More interestingly, individuals who had experienced a prior false-positive breast or prostate cancer screening test result were much more likely to participate in future screenings than individuals who had not experienced a false-positive. This is consistent with prior survey and psychosocial research suggesting that women who experience false-positive mammograms feel relief at not having cancer,^10,11,16 but contrasts with recent retrospective work finding lower mammography rates after false-positive breast cancer screenings.¹⁵ Additionally, in our analysis, not all false-positives had the same effect; false-positive mammograms requiring biopsy had the strongest impact, whereas women with a prior false-positive colorectal cancer FOBT screening had lower odds of being up-to-date with breast cancer screening.

It is desirable for individuals to obtain guideline-recommended screening, so if false-positives increase the probability of adherence, they might be construed as having beneficial aspects. This interpretation would depend on reasons for increased adherence. There are at least 3 possibilities. False-positives may simply be salient reminders to screen for cancer,³¹ without detriment to the individual’s quality of life. Second, false-positives may prompt individuals to increase their perceived likelihood of developing cancer and consequently, raise screening interest.³² Because studies suggest that most people overestimate their cancer risk³³ and screening benefits,³⁴ from this perspective, false-positives may be harmful. Finally, if improved adherence is due to increased anxiety or vulnerability about having cancer, then false-positives may impose long-lasting negative consequences.³⁵ This viewpoint follows 5 systematic reviews,^6,16,36–38 including a 2016 study accompanying the recent US Preventive Services Task Force breast cancer screening recommendation,⁶ that found mostly temporary psychological responses in women with false-positive mammograms. Our findings, if consistent with this theory, would suggest persistent anxiety. Yet, nearly all systematic review findings were limited by small sample size, narrowly selected participants, lack of clinical information and limited follow-up.⁶ Although our study suggests a strong association between prior receipt of a breast or prostate cancer false-positive and future participation in cancer screenings, we cannot discern the reason. Additional research is needed to understand the impact of false-positive testing and potential ways to ameliorate it.

Similarly, research is needed to understand why false-positive colorectal cancer FOBT screenings may lead to decreased breast cancer screening adherence. One possible explanation is that women who chose FOBT (instead of colonoscopy) were initially more hesitant to undergo screening, and a false-positive reinforced that belief. If this explanation is correct, then results may vary by gender; an exploratory analysis found that men with false-positive colorectal cancer FOBT screenings had 15% higher odds of receiving prostate cancer screening.

Limitations

Our results may be confounded by past screening history; individuals with more cancer screenings are more likely to both experience false-positives and engage in future screening. However, we included breast cancer screening history as a covariate, and results were robust to various definitions of that term. Second, we had no information on healthcare received outside of a single system. However, we limited our analysis to patients with ongoing primary care, where cancer screening often takes place. Third, EMR data may have been incomplete for some variables (e.g., family history) or in early years, although sensitivity analyses excluding 2006–2008 yielded similar results. We had no information on education, attitudes toward screening, screening barriers, or out-of-pocket costs. Fourth, only 1.3% of patients were insured by Medicaid, limiting generalizability to low-income populations. Fifth, previous breast biopsy is a risk factor for future breast cancer diagnosis,³⁹ suggesting a valid reason for higher future screening rates in some women. A more common situation, repeat breast imaging without biopsy, may also be associated with future breast cancer,⁴⁰ although this association is recently described and unlikely to explain our results. Sixth, most cancer screenings were ordered during routine clinical care, so results may not extend to organized cancer screening programs. Individuals with false-positives may have been reminded about the need for cancer screenings during follow-up testing. Finally, our observational design allowed us to identify associations, not causality.

Conclusion

Prior experience with cancer screenings may influence a patient’s willingness to continue future screening. Patients who previously experienced a false-positive breast or prostate cancer screening were more likely to engage in future screening for all cancers, while women with a false-positive colorectal cancer FOBT screening were less likely to engage in future breast cancer screening. Future research should seek to confirm our results, explore reasons for the possible long-lasting influence of false-positives, consider whether providers should be alerted to monitor cancer screening after a false-positive (e.g., mammography after a false-positive FOBT), and address ways for healthcare providers to minimize negative impact on patients.