Outcome Quality After Colorectal Cancer Resection in Certified Colorectal Cancer Centers

Abstract

Background

In this observational study, patient-reported outcomes and short-term clinical outcome parameters in patients with colorectal cancer were studied 12 months after the start of treatment. Outcomes were also compared across German Certified Colorectal Cancer Centers.

Methods

Data were collected from 4239 patients with colorectal cancer who had undergone elective tumor resection in one of 102 colorectal cancer centers and had responded to a quality-of-life questionnaire before treatment (EORTC QLQ-C30 and -CR29). 3142 (74.1%) of these patients completed a post-treatment questionnaire 12 months later. Correlation analyses were calculated and case-mix adjusted comparisons across centers were made for selected patient-reported outcomes, anastomotic insufficiency, and 30-day mortality.

Results

At 12 months, mild improvements were seen in mean quality-of-life scores (66 vs. 62 points), constipation (16 vs. 19), and abdominal pain (15 vs. 17). Worsening was seen in physical function (75 vs. 82) and pain (22 vs. 19). Better patient-reported outcomes at 12 months were associated with better scores before treatment. Better results in at least three of the five scores were associated with male sex, higher educational level, higher age, and private health insurance. Major worsening of fecal incontinence was seen among patients with rectal cancer without a stoma. The largest differences across centers were found with respect to physical function. Anastomotic insufficiency was found in 4.3% of colon cancer patients and 8.2% of rectal cancer patients. 1.9% of patients died within 30 days after their resection.

Conclusion

Clinicians can use these findings to identify patients at higher risk for poorer patient-reported outcomes. The differences among cancer centers that were found imply that measures for quality improvement would be desirable.

With around 60 000 new cases each year, colorectal cancer is the third most commonly occurring malignancy among men in Germany and the second-ranking form of cancer in women (1). Apart from survival, the essential goal of the German clinical practice guideline is minimization of symptoms and functional impairments ([2], e.g., the recommendations in section 7.5 ff.). Patient-reported outcomes (PRO) are eminently suitable for the documentation of symptoms and functional impairments (3, 4). PRO measure “patients’ perceptions of their health status, clinical outcomes, mobility and quality of life” (5, page 25). An example of a PRO is fecal incontinence, which is an item in the questionnaire EORTC QLQ-CR29—also used in this study—with the single question “During the past week: Have you had leakage of stools from your back passage/stoma bag?” and the possible responses “Not at all,” “A little,” “Quite a bit,” and “Very much” (7). Randomized studies have shown that PRO are beneficial for treatment planning or monitoring of patients with cancer: better functional outcomes and longer survival were found after intensified PRO monitoring (8– 10), and amelioration of impaired quality of life resulted from a tailored interdisciplinary PRO intervention (4, 11).

Despite being intended when PRO were first developed and recommended by many professional groups (e.g., 13, 14), to date PRO have not been used extensively for assessment of outcome quality or comparison of treating facilities. The symptoms and functional impairments that occur during treatment are so significant, however, that PRO should be viewed as an important patient-centered complement to quality assurance and quality development procedures (16). There are already a number of promising initiatives in other countries: In the USA, for instance, standardized documentation of PRO in physiotherapy offices enables cross-office comparisons (17). In the UK, the National Health Service (NHS) has the National Prostate Cancer Audit (18). Examples in Germany are the QS-Reha project (19) and the PCO study in prostate cancer (20). Overall, there is still too little emphasis on PRO in German quality research (13).

This article presents results from the EDIUM Study (“Ergebnisqualität bei Darmkrebs: Identifikation von Unterschieden und Maßnahmen zur flächendeckenden Qualitätsentwicklung”—“Outcome in Colorectal Cancer: Identification of Differences and Measures for Nationwide Quality Development”), in which PRO and short-term clinical outcome quality parameters from over 100 Certified Colorectal Cancer Centers are compared. The study was funded by the Innovation Committee of the Federal Joint Committee (G-BA) from July 2018 to December 2021 (grant number VSF1_2017–169). To begin with, associations between patient and center characteristics and selected dimensions of EORTC QLQ-C30 and EORTC QLQ-CR29 (7, 21) are described. This is followed by case-mix-adjusted comparisons among centers based on these analyses. Furthermore, findings regarding the short-term clinical parameters anastomotic insufficiency and 30-day mortality are presented.

Method

Patient cohort and data acquisition

Each year, almost half of the patients with newly diagnosed colorectal cancer in Germany are treated in centers certified by the German Cancer Society (22). From these centers we drew a stratified random sample with a move-up procedure. At the 106 centers that initially took part, all primary colorectal cancer patients scheduled for elective tumor resection or nonsurgical palliative care in 2019 were invited to join the study before the intervention; in 6 centers, recruitment started as a pilot phase in October 2018. The analysis presented here is limited to the patients who underwent elective tumor resection, with or without (neo)adjuvant therapy. Emergency patients were excluded because they could not complete the pre-treatment questionnaire. The patients who received nonsurgical palliative care were analyzed separately and the findings were communicated to the funding body (the G-BA) in a publicly available results report. Before the commencement of treatment (tumor resection or neoadjuvant therapy), the participants completed the baseline questionnaire (T0; EORTC QLQ and sociodemographic data) either on paper or online, according to preference. However, not all centers offered the web-based option. Completion of the post-treatment questionnaire (T1; EORTC-QLQ) followed 12 months after tumor resection. Furthermore, as stipulated in the catalogue of requirements for certification the centers documented characteristics of the clinical findings and treatment, which were then linked with survey data by means of the software OncoBox at each center. The individual center data sets were sent to the certification institute OnkoZert for quality assurance and to the principal investigator for analysis. The study protocol (DRKS00008724) stipulates the EORTC dimensions at 12 months as primary endpoints and anastomotic insufficiency (measured separately for cancer of the colon and cancer of the rectum) and 30-day mortality as secondary endpoints. For reasons of space, this article reports on only a selection of EORTC dimensions together with anastomotic insufficiency and 30-day mortality. In line with the protocol, the remaining endpoints are presented in the publicly available results report.

Variables and statistical analyses

First, five dimensions relevant to colon cancer and rectal cancer (“global health status/QoL,” “physical function,” “pain,” “constipation,” “abdominal pain”) from both EORTC QLQ-C30 and EORTC QLQ-CR29 were investigated. Based on a survey of the participating centers, the scores were then selected according to the relevance of the dimensions for treatment planning (23). Scoring proceeded in line with the manuals, with scores from 0 to 100 possible. Cut-off points for clinical relevance are available for the three C30 dimensions “physical function” (cut-off: 83), “pain” (cut-off: 25), and “constipation” (cut-off: 50) (24). On functional scales (“global health status/QoL,” “physical function”) higher scores indicate better function, but on symptom scales (“pain,” “constipation,” “abdominal pain”) higher scores mean more severe symptoms. Because of the importance of incontinence particularly in rectal cancer (23), changes in “fecal incontinence” from T0 to T1 were also taken into consideration. In contrast to the procedure for the other five scores, owing to the peculiarities of the score this was accomplished purely by means of contingency tables according to site (colon/rectum) and presence or absence of stoma (eMethods). The secondary endpoints anastomotic insufficiency (in patients with an anastomosis) and 30-day mortality after resection were documented as stipulated in the requirements for certification. The covariates used for the association analyses and case-mix adjustment are listed in the eMethods and described, together with others, in Table 1 and Table 2.

The primary correlation analyses for the five EORTC scores specified above ensued, in line with the procedure for case-mix adjustment, as a series of multivariable linear regressions, presented in the form of forest plots (eTable 1). Case-mix adjustment was performed according to the NHS procedure (25). For each score, only those centers were included which contributed post-treatment questionnaires from at least ten patients. Graphically the adjusted values were provisionally set in relation to minimally important differences (MID) (26– 28). An MID is the smallest change in a score that is perceived as important by the patient. Details of the statistical procedure for the primary and secondary endpoints and of further analyses (mixed and tobit models) can be found in the eMethods.

Results

Sample

Between October 2018 and December 2019, 4239 patients from 102 centers completed the baseline questionnaire prior to tumor resection (table 1). Just short of 7% of the participants completed the questionnaire online, while the rest used the paper version. The median response rate of the centers in the study as a whole was 51%. The association analyses were based on the 3142 patients (74.1%) who took part in the post-treatment survey. The drop-out rate was 25.9%. Around a third of the drop-outs were accounted for by documented deaths, while the reasons for the remaining cases are unknown (Table 1, eFigure 1).

Table 1. Characteristics of the sample.

	Colon (n = 2859)	Rectum (n = 1380)	Total (N = 4239)
Age
Mean (SD)	70.8 (11.6)	67.7 (11.2)	69.8 (11.5)
Median (IQR)	73 (63–79)	68 (60–77)	71 (62–79)
Age (grouped)
≤ 39	35 (1.2%)	9 (0.7%)	44 (1.0%)
40–49	94 (3.3%)	59 (4.3%)	153 (3.6%)
50–59	357 (12.5%)	275 (19.9%)	632 (14.9%)
60–69	686 (24.0%)	413 (29.9%)	1099 (25.9%)
70–79	983 (34.4%)	385 (27.9%)	1368 (32.3%)
≥ 80	704 (24.6%)	239 (17.3%)	943 (22.2%)
Sex
Male	1580 (55.3%)	891 (64.6%)	2471 (58.3%)
Female	1279 (44.7%)	489 (35.4%)	1768 (41.7%)
ASA score
ASA 1	113 (4.0%)	57 (4.1%)	170 (4.0%)
ASA 2	893 (31.2%)	439 (31.8%)	1332 (31.4%)
ASA 3	707 (24.7%)	331 (24.0%)	1038 (24.5%)
ASA 4	43 (1.5%)	10 (0.7%)	53 (1.3%)
ASA 5	0 (0.0%)	0 (0.0%)	0 (0.0%)
Missing	1103 (38.6%)	543 (39.3%)	1646 (38.8%)
Nationality
German	2729 (95.5%)	1321 (95.7%)	4050 (95.5%)
Other	112 (3.9%)	43 (3.1%)	155 (3.7%)
Missing	18 (0.6%)	16 (1.2%)	34 (0.8%)
Health insurance
Statutory	2449 (85.7%)	1171 (84.9%)	3620 (85.4%)
Private	367 (12.8%)	183 (13.3%)	550 (13.0%)
Other/not insured	20 (0.7%)	10 (0.7%)	30 (0.7%)
Missing	23 (0.8%)	16 (1.2%)	39 (0.9%)
Education level
General secondary school	1312 (45.9%)	596 (43.2%)	1908 (45.0%)
Intermediate secondary school	560 (19.6%)	306 (22.2%)	866 (20.4%)
Polytechnic secondary school	152 (5.3%)	99 (7.2%)	251 (5.9%)
Qualification for university of applied sciences	249 (8.7%)	114 (8.3%)	363 (8.6%)
Qualification for university	443 (15.5%)	202 (14.6%)	645 (15.2%)
Other	67 (2.3%)	27 (2.0%)	94 (2.2%)
No qualification	25 (0.9%)	16 (1.2%)	41 (1.0%)
Missing	51 (1.8%)	20 (1.4%)	71 (1.7%)
UICC stage
UICC I	812 (28.4%)	498 (36.1%)	1310 (30.9%)
UICC II	961 (33.6%)	332 (24.1%)	1293 (30.5%)
UICC III	773 (27.0%)	395 (28.6%)	1168 (27.6%)
UICC IV	313 (10.9%)	155 (11.2%)	468 (11.0%)
Stoma (initially created)
Yes	192 (6.7%)	1010 (73.2%)	1202 (28.4%)
No	2667 (93.3%)	370 (26.8%)	3037 (71.6%)
Stoma (still present at 12 months)
No	2025 (70.8%)	632 (45.8%)	2657 (62.7%)
Yes	102 (3.6%)	325 (23.6%)	427 (10.1%)
Missing	732 (25.6%)	423 (30.7%)	1155 (27.2%)
Deceased
Not deceased/unknown	2607 (91.2%)	1265 (91.7%)	3872 (91.3%)
Deceased	252 (8.8%)	115 (8.3%)	367 (8.7%)
Deceased (30 days)
Survived > 30 days	2800 (97.9%)	1359 (98.5%)	4159 (98.1%)
Survived ≤ 30 days	59 (2.1%)	21 (1.5%)	80 (1.9%)

Missing values are given only for variables where missing values occurred.

ASA, American Society of Anesthesiologists; IQR, interquartile range; SD, standard deviation;

UICC, Union Internationale Contre le Cancer

Patient-reported outcomes

As shown in Table 2, the mean global health status/QoL was slightly better after 12 months than at the time of the baseline survey (score 66 versus 62 points), and both constipation (16 versus 19) and abdominal pain (15 versus 17) were on average less severe. The mean scores worsened for physical function (75 versus 82) and pain (22 versus 19). The arithmetic means after treatment for the three dimensions for which cut-offs were available were 75 (physical function), 22 (pain), and 16 (constipation). Therefore, a mean impairment of physical function was identified that could have relevance for treatment planning (24; 2).

eTable 2 shows that fecal incontinence became much worse in patients with cancer of the colon who had a stoma (mean change from 8 to 26 points, n = 63) and in patients with rectal carcinoma who did not have a stoma (mean change from 14 to 27 points, n = 522).

The multivariable linear regression analyses (Figure 1, eTable 1, eFigure 2) explain 11–45% of the variance (R²) for the dependent variables.

Figure 1.

Results of multivariable linear regression analyses: estimator (95% confidence intervals [95% CI])

ASA, American Society of Anesthesiologists; EL, Education level; UICC, Union Internationale Contre le Cancer

The pre-treatment scores were positively associated with the PRO in all models. Better PRO at 12 months were associated with better general health status (global health status/QoL, pain), cancer of the colon (constipation, global health status/QoL), higher level of education (pain, physical function, global health status/QoL), private health insurance (constipation, pain, physical function, global health status/QoL), male sex (abdominal pain, physical function, constipation), and a lower Union Internationale Contre le Cancer (UICC) stage for physical function and global health status/QoL but a higher UICC stage for abdominal pain and constipation. The age group over 79 years had worse PRO than the reference group of 70- to 79-year-olds for two scores, while the 50–59 age group had worse PRO for four scores. Presence of a stoma 12 months after treatment was associated with restriction of physical function and global health status/QoL but with a better score for constipation. With regard to distribution of residuals, testing of the model assumptions showed deviations from normal distribution, especially for constipation and abdominal pain. Tobit models showed larger estimators for the baseline scores but otherwise yielded similar results (eTable 3). Owing to singularity, multilevel models could be calculated only for global health status/QoL: the proportion of variance was small (ICC < 1%) at center level, and center characteristics failed to explain the variance in the sole converging multilevel model (eTable 4).

Comparison of the centers after case-mix adjustment revealed differences in scores (Figure 2 [illustrating pain and physical function], eFigure 3; observed and adjusted data available on request). The greatest interquartile range (69–79, larger than an MID) was found for the physical function score. For the other scores the interquartile ranges (4, 7, 5, and 7 points) were smaller than an MID (7, 10, 10, 9). Some centers had outcomes that were clearly below or above average, especially for physical function. Descriptively, there were no clear patterns with regard to associations between the centers’ adjusted scores and the following center characteristics: surgical case numbers, hospital funding type, teaching status, community size (eTable 5– 8).

Anastomotic insufficiency and 30-day mortality

Of the patients who had an anastomosis, 4.3% (colon, 121/2783) and 8.2% (rectum, 87/1061) experienced anastomotic insufficiency; 1.9% (80/4239) died within 30 days of resection. The multivariable regression analyses of the short-term clinical endpoints anastomotic insufficiency and 30-day mortality yielded R² < 0.03. The findings are illustrated in eFigure 4 and eFigure 5. The case-mix-adjusted results for the secondary endpoints are not shown because of the small degree of variance explained.

Discussion

This study examined the associations between patient characteristics and both PRO and short-term clinical outcome quality parameters 12 months after elective resection of colorectal cancer, comparing the results across Certified Colorectal Cancer Centers. The association analyses provided evidence of previously little investigated correlations of post-treatment PRO. These can be used by clinicians to identify patients with a higher risk of poor PRO a year after surgery. The sometimes strong associations with regard to the socioeconomic characteristics are particularly striking. For example, global health status/QoL, controlled for clinical characteristics and baseline status, is 5 points higher for persons with private than for those with statutory health insurance and 4 points higher for persons with a qualification for university entrance than for those whose education went no further than general secondary school (year 9 lower secondary school certificate). Evidence for associations between outcome quality and socioeconomic status has already been shown for Germany (29). With regard to the American Society of Anesthesiologists (ASA) score, as can be expected PRO were predominantly worse with severe disease. It is striking that four scores were worse in 50- to 59-year-olds than in the 70–79 age group.

Comparison of the centers showed that the interquartile ranges mostly differed to the extent of about an MID, although the distribution was wide. Case-mix adjustment and graphic presentation of the data enable the identification of some centers with distinctly worse outcomes than would be expected from the case mix. The largest differences were in the score for physical function (figure 2), where the interquartile range was 10 points, larger than an MID. This score was based on five items, so it may differentiate better than the other scores, which were calculated from either one or two items. The items are not specific to colorectal cancer and are of central relevance to the patients: An example is provided by the item “Do you have any trouble taking a long walk?,” with the possible responses “Not at all,” “A little,” “Quite a bit,” “Very much”. The center characteristics examined in the additional analyses contributed only slightly or not at all to explanation of the variance.

Figure 2.

Case-mix adjustment—point estimator of the centers’ adjusted post-treatment patient-reported outcome (PRO) scores. Only centers with at least 10 post-treatment questionnaires are included. Each dot represents an anonymized center.

a) Function score (the higher the score, the better the outcome): physical function (75, 69–79)

b) Symptom score (the lower the score, the better the outcome): pain (23, 19– 26)

The vertical lines (“antennae”) extending upward and downward represent provisional minimally important differences (MID) as one third of the standard deviation at patient level: physical function 7, pain 10

With regard to the change from before surgery to 12 months thereafter, results from the Netherlands point in a similar direction (30, 31). There too, only small differences were found between baseline and 12 months, although some of the differences for physical function were pronounced. Altogether, there is little in the way of comparable data from countries other than Germany.

The frequency of the short-term clinical endpoints anastomotic insufficiency and 30-day mortality in our sample is almost identical to that in the certified centers overall (22). Apart from sex and tumor site (for anastomotic insufficiency) and age and ASA score (for 30-day mortality), there were no factors associated with the short-term clinical endpoints, and overall the amount of variance explained by the models is low. Some risk factors known from the literature, e.g., body mass index, were not measured in this study. The procedure described here therefore seems unsuitable for case-mix-adjusted comparison of the short-term clinical endpoints; additional adjustor variables, for instance, would be needed for this purpose.

To our knowledge, our study is the first to compare PRO across centers in patients with colorectal cancer. Differences among treating facilities have been found for other diseases, e.g., orthopedic diagnoses (15), prostate cancer (28), and other surgical interventions (32). In some of these cases the differences were much greater. The differences across centers may be less pronounced in certified centers than they would be in a sample of other hospitals because of the relative homogeneity of the former. Certified centers are obliged to implement a whole range of quality requirements and have on average a better oncological outcome quality than uncertified institutions (33– 35). They thus represent a selection of the hospitals offering treatment for colorectal cancer. However, this is also true for prostate cancer centers, for which the differences in the PRO are larger (28). Although resection for colorectal cancer does not entail relatively frequently occurring functional impairments like the incontinence and erectile dysfunction that may follow treatment for prostate cancer, differences can by all means be anticipated owing to the challenging nature of interventions in the rectum. With this in mind, it would be beneficial to differentiate among different sites and between presence and absence of stoma in future research.

EORTC QLQ-C30 and EORTC QLQ-CR29 are recommended, for example, in the ICHOM standard set for colorectal cancer (36). However, no recommendations exist for prioritization among the more than 30 scores for patients with colorectal cancer or on whether stratified analyses are necessary. Further research efforts are required. Work is also needed to establish the significance of different surgical procedures with regard to PRO and also rehabilitation and aftercare, which can only partly be influenced by the Certified Colorectal Cancer Centers.

When interpreting the findings, the strengths and limitations of the study have to be taken into account: Unlike many other measures of care quality, questionnaire data necessitate recruitment efforts by the centers involved and active cooperation on the part of the patients. The median proportion of eligible patients who completed the baseline questionnaire was 51%, a satisfactory figure. Around one quarter of the patients did not take part in the follow-up survey. Death was documented for one third of this group. The remaining patients who did not respond at T1 had been more severely ill on average at T0 (data available on request).

The basic clinical documentation used for the purposes of this study was applied uniformly across all centers and is also obligatory for the certification process. The documentation is therefore close to complete, with simultaneous external validation by the auditors during certification. Nevertheless, a high proportion of missing data is found for the ASA classification (eMethods). Although the participating centers had high case numbers compared with German centers as a whole, low numbers in a few centers (median primary number of patients undergoing elective surgery: 76), together with the selective inclusion of patients, may distort the results. The centers with sufficient data for case-mix adjustment did not differ in hospital funding type, teaching status, and community size from those with insufficient data, but had higher primary case numbers (data available on request).

Finally, it can be stated that measurement of PRO for comparison of outcome quality may be especially useful to identify centers with results far above or below average. This gives those responsible for treatment an additional instrument to help them understand what patients need and what must be done to improve care provision further. The center-oriented data presented here may serve to stimulate steps towards quality improvement, e.g., the development of measures for the benefit of patients at high risk of reduced function; consideration in certification audits or in internal quality circles; and identification of centers with particularly good results. In this way we can all learn from one another (37).

Table 2. EORTC scores for patients who took part in the post-treatment survey (N = 3142).

		T0	T1
Global health status/ QoL	n missing	41	18
	Mean (SD)	62 (23)	66 (21)
	Median (IQR)	67 (50–83)	67 (50–83)
Physical function	n missing	7	9
	Mean (SD)	82 (22)	75 (24)
	Median (IQR)	93 (73–100)	80 (60–93)
Pain	n missing	3	10
	Mean (SD)	19 (28)	22 (28)
	Median (IQR)	0 (0–33)	17 (0–33)
Constipation	n missing	30	31
	Mean (SD)	19 (31)	16 (26)
	Median (IQR)	0 (0–33)	0 (0–33)
Abdominal pain	n missing	23	19
	Mean (SD)	17 (27)	15 (24)
	Median (IQR)	0 (0–33)	0 (0–33)

IQR, Interquartile range; SD, standard deviation; T0, before commencement of treatment;

T1, 12 months after tumor resection

eTable 2. Stoma-specific score “fecal incontinence” at T0 and T1 in colon cancer and rectal cancer for the strata: stoma present after 12 months and stoma absent after 12 months (all study patients with scores at T0 and T1).

	Colon				Rectum				Total
	Stoma yes		Stoma no		Stoma yes		Stoma no		Stoma yes		Stoma no
	T0	T1	T0	T1	T0	T1	T0	T1	T0	T1	T0	T1
Fecal incontinence	(n = 63)	(n = 63)	(n = 1 609)	(n = 1 609)	(n = 247)	(n = 247)	(n = 522)	(n = 522)	(n = 310)	(n = 310)	(n = 2 131)	(n = 2 131)
Mean (SD)	7.94 ± 19.60	26.45 ± 30.03	5.61 ± 17.26	10.07 ± 21.69	24.69 ± 32.33	26.85 ± 30.11	13.79 ± 25.15	26.82 ± 31.93	21.29 ± 30.90	26.77 ± 30.04	7.62 ± 19.80	14.17 ± 25.62
Median (IQR)	0 [0; 0]	33.33 [0; 50]	0 [0; 0]	0 [0; 0]	0 [0; 33.33]	33.33 [0; 33.33]	0 [0; 33.33]	33.33 [0; 33.33]	0 [0; 33.33]	33.33 [0; 33.33]	0 [0; 0]	0 [0;33.33]

IQR, Interquartile range; SD, standard deviation