Trends in Colorectal Cancer Screening from the National Health Interview Survey: Analysis of the Impact of Different Modalities on Overall Screening Rates

Colorectal cancer is the third most diagnosed cancer and the second leading cause of cancer-related death in the United States (1). The American Cancer Society (ACS) estimates that in 2023, there will be 153,020 new cases of colorectal cancer diagnosed in the United States and 52,550 people will die from the disease (1). However, survival rates are higher if colorectal cancer is detected at an earlier stage (2–4). The 5-year relative survival rate for patients diagnosed with localized colorectal cancer is 91% while the rate for those diagnosed with advanced-stage colorectal cancer is 14% (2). The most recent authoritative guidelines for colorectal cancer screening recommend that individuals at average risk for colorectal cancer begin screening at age 45, as the incidence of both colorectal cancer and premalignant lesions increases sharply after this age (5–7). The U.S. Preventive Services Task Force (USPSTF) guidelines recommend various colorectal cancer screening modalities without any preference to a specific modality including colonoscopy (every 10 years), high-sensitivity guaiac fecal occult blood test (FOBT) or fecal immunochemical test (FIT) every year, CT colonography every 5 years, flexible sigmoidoscopy every 5 years, and multitarget stool DNA (mt-sDNA) every 3 years (5). Despite the USPSTF giving colorectal cancer screening a “Grade A” recommendation in average-risk adults ages 50 to 75 years, colorectal cancer screening rates among this population were 71.8% in 2021 (8), a considerably lower figure than the national target of 80% established by the National Colorectal Cancer Roundtable to improve colorectal cancer outcomes.

Recognizing the importance of increasing overall colorectal cancer screening rates nationally, the USPSTF and other major guideline review groups, including the ACS and National Comprehensive Cancer Network, recommend multiple screening strategies with equal positioning. These include having a FIT, FOBT, an mt-sDNA test, colonoscopy, sigmoidoscopy, or CT colonography at regular recommended intervals.

While data on overall colorectal cancer screening trends are available (9), the impact of each screening modality on the overall screening rates is not. It is important to understand overall trends and the contribution of each screening modality to the overall screening rates to identify changes in patient and provider preferences over the years. These data may be used to inform strategies to achieve the national colorectal cancer screening goal of 80%.

The study objectives were 2-fold. The first objective was to examine trends in colorectal cancer screening rates from 2005 to 2021 among adults ages 50–75 years. The second objective was to quantify the contribution of each colorectal cancer screening modality to overall colorectal cancer screening rates between 2018 and 2021 among individuals who were adherent to any colorectal cancer screening modality (adherent population).

The National Health Interview Survey (NHIS; www.cdc.gov/nchs/nhis/index.htm) is a nationwide cross-sectional survey of the civilian population ages 18 years and older conducted by the National Center for Health Statistics. The survey is conducted annually, and data are collected via personal interviews from a nationally representative sample of households and noninstitutionalized settings (e.g., boarding houses) across the country. It is a principal source of information regarding the health and well-being of the civilian, noninstitutionalized U.S. population, and data are collected on a broad range of health topics. The final sample adult response rates were 69.0%, 60.8%, 61.2%, 55.2%, 53.1%, 59.1%, and 50.9% in 2005, 2010, 2013, 2015, 2018, 2019, 2021, respectively (available at: https://www.cdc.gov/nchs/nhis/data-questionnaires-documentation.htm).

This was a retrospective cross-sectional analysis of the NHIS database for the years in which the Cancer Control Supplement was administered (2005, 2010, 2013, 2015, 2018, 2019, and 2021). The supplement specifically focuses on data relating to cancer prevention and early detection and includes cancer screening-related questions, particularly around colorectal cancer screening behaviors in those years.

The study population was adults ages 50–75 years inclusive. Eligible participants included those who had responded to the colorectal cancer screening–related questions regarding whether they had a screening test and when using at least one of the screening modalities: colonoscopy, FOBT/FIT, mt-sDNA, sigmoidoscopy, or CT colonography.

The latest USPSTF colorectal cancer recommendation (2008, 2016, 2021), corresponding to each time period was used to determine the proportion of individuals up-to-date with colorectal cancer screening (study outcome) as described by Shapiro and colleagues (9). For NHIS data between 2005 and 2015, “up-to-date” was defined as the use of an FOBT or FIT within the past year, colonoscopy within the past 10 years, or the combination of sigmoidoscopy within the past 5 years and FOBT or FIT within the previous 3 years. Those reporting that their most recent colorectal cancer screening test occurred within a timeline consistent with recommendations were considered up to date. Starting in 2018, colorectal cancer screening test use was defined as the use of an FOBT or FIT within the past year, colonoscopy within the past 10 years, CT colonography, or sigmoidoscopy within the previous 5 years, or mt-sDNA test within the past 3 years. Using these definitions, the proportions of respondents who reported having a screening test during a study year were calculated to assess colorectal cancer screening trends from 2005 to 2021. Considering the NHIS's complex sample design, weighted estimates were calculated using the data weighting procedures described in the NHIS documentation (available at: https://nhis.ipums.org/nhis/userNotes_weights.shtml). This allowed the estimated rates or averages to be generalizable to the broader population. Unweighted and weighted point estimates were calculated and reported together with either SEs or 95% confidence intervals (CI), where applicable. Using combined data from all NHIS years, age and sex-standardized point estimates were calculated across all years (2005–2021) to assess trends over time.

Note that the question related to the mt-sDNA use was added to the NHIS survey in 2018. To estimate the extent to which each colorectal cancer screening modality contributed to overall screening rates for 2018, 2019, and 2021, the proportion of each modality attributed to the overall screening rate for each year was calculated. Trends in attribution rates for each modality were subsequently examined among adherent population.

To estimate the proportion of the change in colorectal cancer screening rates accounted for by each modality from 2018 to 2021, a pseudo-time-series cross-sectional (pseudo-TSCS) analysis was conducted utilizing a combination of random sampling, stratification, and unit proportion distribution. First, observations for the samples for 2018, 2019, and 2021 (years in which mt-sDNA screening rates were available) were grouped according to sex (dichotomous), and age group (three categories), yielding six groups. An iterative process was subsequently adopted to determine an efficient and appropriate unit size of approximately 40–50 respondents per unit. This was undertaken using STATA's random number allocation function and each of the six groups was subdivided into units that averaged approximately 50 respondents per unit. Variable values (i.e., screening rates) for those units were defined as the mean proportion of screening for each colorectal cancer screening modality for the overall group so that the values for each unit within each class (differentiated by sex and age group) were comparable. Because of the insufficient number of respondents in each stratified subgroup per timepoint, we excluded CT colonography from this analysis.

The units were then randomly matched to comparable units within the same group for the other samples (to enable longitudinal analysis). Using these units, variables, and samples, a random-effects generalized least squares (GLS) analysis was conducted, including a lagged dependent variable to analyze the longitudinal increase, rather than the cross-sectional constitution of the overall screening rate. These regression estimates were then used to calculate the relative marginal contribution, that is, the contribution to the overall screening change from 2018 to 2021 for each additional percent of screening with that modality and overall relative contribution per test modality (marginal relative contribution * estimated average change from 2018 to 2021).

All statistical analyses were performed using a standard software package (Stata, version. 17.0; StataCorp). For extracting the NHIS data files, we adapted the STATA codes provided for individual study years at https://www.cdc.gov/nchs/nhis/data-questionnaires-documentation.htm.

The source data used in this study are publicly available at https://www.cdc.gov/nchs/nhis/data-questionnaires-documentation.htm. The full STATA code for extracting the relevant NHIS data files is available on request from the authors.

A total of 85,571 records (unweighted) of individuals ages 50–75 years who met the inclusion criteria were identified from the NHIS data from 2005 to 2021 and included in the analysis.

Among overall population, the estimated colorectal cancer screening increased from 47.7% in 2005 to 69.9% in 2021. The largest increase was between 2005 and 2010 (47.7%–60.7%), after which rates plateaued until 2015. Between 2015 and 2021, the rates further increased from 63.5% to 69.9%. From 2018 to 2021, mt-sDNA tests had the largest increase from 2.5% to 6.6%. During this time, the use of colonoscopy increased from 61.2% to 64.0%, while FOBT/FIT utilization increased from 5.3% to 5.5%. The use of CT colonography (1.0%–1.7%) and sigmoidoscopy (1.1%–1.3%) also increased from 2018 to 2021 in this age group (Fig. 1).

Among adherent population, colonoscopy attributed the most (>80.0%) to the overall screening rates across all years, with the percent attribution increasing from 2005 to peak in 2015 at 88.5%, after which then subsequently declined. Although the attribution of colonoscopy and FIT/FOBT to the overall colorectal cancer screening rate decreased after 2015, the overall screening increased steadily over the years primarily due to an increase in the mt-sDNA attribution rate. The % attribution of mt-sDNA to the overall screening rate increased from 3.5% in 2018 to 8.3% in 2021 (Fig. 2).

Results of the pseudo-TSCS analysis including a GLS random effects model to estimate the longitudinal impact of each screening modality on the overall increase in screening rates among adherent population indicated that from 2018 and onward, mt-sDNA contributed the most to the increase in overall colorectal cancer screening rates (77.3%; 95% CI = 64.7–89.9) followed by colonoscopy (11.6%; 95% CI = 10.8–12.4), FOBT/FIT 2.3%; 95% CI = 1.6–4.2), and sigmoidoscopy (1.3%; 95% CI = 0.9–1.9; all P < 0.05; Fig. 3).

Results of this retrospective cross-sectional analysis of NHIS data showed that the estimated overall colorectal cancer screening rate increased substantially from 47.7% in 2005 to 69.9% in 2021 in adults ages 50–75 years. The largest increase was between 2005 and 2010 (47.7%–60.7%) at which time rates subsequently plateaued until 2015. There was a further increase from 63.5% in 2015 to 69.9% in 2021, largely driven by the use of newly added mt-sDNA test, which more than doubled from 2018 to 2021, while the use of other modalities such as colonoscopy, FOBT/FIT, sigmoidoscopy, and CT colonography increased steadily over the same time. The use of colonoscopy for colorectal cancer screening increased from 61.2% in 2018 to 64.9% in 2021, while FOBT/FIT utilization increased from 5.3% in 2018 to 5.5% in 2021. Colonoscopy remained the dominant modality for colorectal cancer screening across all years.

We found that the mt-sDNA utilization increased, colonoscopy and FOBT/FIT utilization was stable, and sigmoidoscopy declined during the study years in the overall population. Because NHIS data are cross-sectional, it is difficult to assess whether mt-sDNA resulted in new adherence or a change in modalities for those who were already adherent. A retrospective review of patients in a large primary care network showed that among newly adherent individuals, the proportion screened by colonoscopy decreased from 89% to 80%, while uptake of mt-sDNA increased from 0.9% to 6.8%, respectively (10).

These results are consistent with the previously reported studies that examined colorectal cancer screening rates using the NHIS survey (9, 11, 12). For example, Shapiro and colleagues examined trends in colorectal cancer screening rates from 2010 to 2018 using NHIS data among adults ages 50–75 years (9). This study, using a similar methodology to the current study, reported that an estimated 66.9% of U.S. adults had a colorectal cancer screening test within recommended intervals in 2018. Also consistent with our findings were their findings that colonoscopy was the most commonly used test in this age group (61.1%) and that there was a statistically significant increase in the use of stool-based tests from 2015 to 2018 (7%–10.9%), partly due to an increase in mt-sDNA. The colorectal cancer screening rates in 2019 and 2021 are similar to those reported by Liu and colleagues (11) Using NHIS data, Liu and colleagues estimated the proportion of adults up to date with colorectal cancer screening increased from 66.7% in 2019 to 70.9% in 2021 with a larger shift toward noninvasive tests. The 2021 screening rates in our study (69.9%) are slightly lower than those reported by the NCI (71.8%; ref. 8), perhaps due to differences in the screening rate definitions and underlying study population (denominator).

While previous studies reported trends in colorectal cancer screening rates and increased utilization of stool-based tests, to our knowledge, this is the first study quantifying the impact of each modality on the overall colorectal cancer screening rates among adherent population using advanced statistical models. The 2018 NHIS is the first large nationally representative survey that collected colorectal cancer screening information on mt-sDNA testing. These questions were also included in the 2019 and 2021 surveys, hence data from 2018 onward were included in analyses to understand the impact of each modality on overall screening. Note that screening rates for all modalities including sigmoidoscopy increased during 2018–2021. Using a pseudo-TSCS analysis, we were able to estimate the longitudinal impact of each screening modality on the overall screening rate. Major advantages to this approach include being able to obtain relatively robust estimates of changes in similar (sub)populations and, in this case, the effects of the introduction of new screening tests on changes in overall screening rates. Quantifying the effects of existing screening tests on the change in overall screening rates is somewhat more complicated but possible with additional assumptions. Overall, while not as reliable as a true longitudinal study, this approach provided a sound alternative when only cross-sectional data were available. These analyses revealed that mt-sDNA contributed 77.3% to the increase in overall colorectal cancer screening rates from 2018 to 2021 while colonoscopy contributed 11.6%. FOBT/FIT and sigmoidoscopy contributed 2.3% and 1.3%, respectively. Given the socioeconomic disparities in colorectal cancer screening (9, 10), future studies are encouraged to understand how different colorectal cancer screening modalities may impact screening rates among different demographic subpopulations to inform targeted outreach.

These results are crucial as they provide guidance on how each modality is contributing to the overall colorectal cancer screening rates. To improve the overall colorectal cancer screening rates and adherence, there needs to be collective efforts to promote targeted screening initiatives including modalities based on patient and provider preferences. Potential barriers associated with low colorectal cancer screening uptake include inadequate access to care, lack of recommendation from health care providers, lack of health insurance coverage for colorectal cancer screening tests, limited knowledge about the screening requirements, fear of invasive testing, and logistical barriers such as transportation, language, and scheduling appointments. Noninvasive, stool-based tests, such as FIT and mt-sDNA, that can be performed at home and offer a low-cost screening option with improved convenience and accessibility. The utilization rates of stool-based tests increased sharply in recent years, particularly during the COVID-19 pandemic (11, 13, 14). It is important to note that the positive results on stool-based tests require a follow-up colonoscopy to complete the screening paradigm, hence adherence to the follow-up colonoscopy after a positive stool-based test is also imperative.

Strengths of this study include analysis of a large dataset representative of the U.S. adult population from 2005 to 2021. The large sample size allowed us to estimate recent colorectal cancer screening rates and trends both overall and by test modality within the population. The uniqueness of our study is quantifying the longitudinal impact of each modality on overall screening rates. Note that the NHIS data are cross-sectional, hence longitudinal patterns in screening behavior cannot be tracked at the patient level using traditional time-series analyses. However, this was accomplished using pseudo-TSCS analyses. For the pseudo-TSCS statistical models, we needed to treat the units as individuals. While the models should not have suffered from common heteroscedasticity or autocorrelation issues (more than any other dataset, at least), the units could not be fully treated as independent measurements and GLS models were used to compensate for this limitation in the analyses. In addition, we did not have repeated measurements and there was a risk of overfitting our models and multicollinearity issues if all screening tests were included, especially in combination with a lagged dependent variable. Overall, we were largely dependent on the representativeness of units for our population per year, for which stratification and randomization were required as well as a sufficient sample size which was particularly important. Finally, compared to other tests, limited data are available on mt-sDNA utilization. Hence, the impact of each test modality on the increase in overall colorectal cancer screening tests was calculated from 2018 to 2021 only and additional data are needed to confirm these trends.

There are several limitations inherent to the cross-sectional survey data to note. First, the NHIS sample adult response rates varied across the years and were the lowest for the year 2021 (50.9%). Although we used weighted estimates for all years, some nonresponse biases may still exist. We excluded individuals with a personal history of colorectal cancer from the trend analyses; however, we were unable to identify other high-risk factors, such as inflammatory bowel disease and personal/family history of colorectal cancer and/or advanced adenomas, which would preferentially lead to colonoscopy for colorectal cancer screening. Furthermore, due to the cross-sectional nature of the NHIS data, we are unable to assess the prior use of colorectal cancer screening modalities at individual level. NHIS data are self-reported and hence reliant on the accuracy of respondents’ recall and truthfulness regarding the screening test and timing. While the majority of the individuals reported one screening modality (∼92%) during a 1-year timeframe, up to 10% reported receiving multiple screening modalities in a given study year potentially overestimating the screening rates particularly colonoscopy as NHIS data do not distinguish between screening versus follow-up colonoscopy. Nevertheless, studies have generally found moderate to good agreement between self-reported colorectal cancer test use and information from medical records (15, 16).

Although colorectal cancer screening rates have increased since 2005, they are well below the crucial national target of 80% necessary to achieve optimum colorectal cancer outcomes. In recent years, home-based noninvasive screening such as mt-sDNA and FOBT/FIT have contributed significantly to the overall colorectal cancer screening rate, even as colonoscopy remains the primary colorectal cancer screening method. Our results suggest that the growing adoption of mt-sDNA is correlated with the increase in overall screening in this average-risk population. Encouraging greater use of at-home stool-based tests through public health campaigns and other initiatives may further improve screening rates. A collaborative approach between health care providers and screening eligible individuals is needed to empower patients with a choice for colorectal cancer screening to achieve the national screening goal.

D.W. Ebner reports other support from Exact Sciences during the conduct of the study. L.J. Finney Rutten reports other support from Exact Sciences during the conduct of the study; other support from Exact Sciences outside the submitted work. L.-A. Miller-Wilson reports other support from Exact Sciences during the conduct of the study. V. Vahdat reports personal fees and other support from Exact Sciences outside the submitted work. A.B. Ozbay reports other support from Exact Sciences Inc. during the conduct of the study; other support from Exact Sciences Inc. outside the submitted work. P.J. Limburg reports other support from Exact Sciences during the conduct of the study; other support from Exact Sciences and Exact Sciences outside the submitted work. No disclosures were reported by the other author.

D.W. Ebner: Conceptualization, supervision, validation, methodology, writing–original draft, writing–review and editing. L.J. Finney Rutten: Conceptualization, resources. L.-A. Miller-Wilson: Conceptualization, resources, supervision, methodology, writing–review and editing. N. Markwat: Data curation, software, formal analysis, visualization, writing–original draft, writing–review and editing. V. Vahdat: Formal analysis, validation, visualization, methodology, writing–review and editing. A.B. Ozbay: Conceptualization, supervision, funding acquisition, writing–review and editing. P.J. Limburg: Resources, writing–review and editing.

Financial support for this study was provided by a contract with Exact Sciences Corporation. All authors contributed to the conceptual framework of the study. Exact Sciences designed the study and determined the data sources in conjunction and agreement with all the authors. The data of this analysis were available to all authors. All analyses were conducted by Exact Sciences. Exact Sciences contributed to the interpretation of the data; the preparation, review, and approval of the article; and the decision to submit the article in conjunction with all co-authors. The funding agreement ensured the authors’ independence in designing the study, interpreting the data, writing, and publishing the report.