Body mass index and colorectal cancer recurrence and mortality: A nationwide cohort study in Denmark

Abstract

Obesity may exacerbate colorectal cancer (CRC) prognosis, yet previous studies have shown inconsistent findings due to methodological limitations. In this nationwide cohort study, we investigated the association between body mass index (BMI) on CRC recurrence and mortality, while examining the influence of demographic, clinical, and socioeconomic characteristics. From the Danish Colorectal Cancer Group database and population‐based registries, we assembled a cohort of stage I–III CRC patients who underwent curative intent surgery between 2002 and 2020. We identified 5‐year CRC recurrence and mortality using validated algorithms and registries. We estimated cumulative incidence functions (CIFs) and applied Cox regression modeling BMI via splines and categorically. We examined synergistic effects via stratification and interaction analyses. Median follow‐up was 4.4 years for recurrence and 4.5 years for mortality; 5877 patients experienced CRC recurrence (CIF = 18.4%) and 6568 died (CIF = 21.8%). BMI was not associated with the recurrence of colon or rectal cancer. Underweight (BMI <18.5 kg/m²) and class II–III obesity (BMI >35 kg/m²) were associated with increased mortality, especially for rectal cancer. For rectal cancer, class II–III obesity was associated with excess mortality, particularly in younger patients (30–50 years) and males. For colon and rectal cancer, male sex, young age and comorbidities interacted to increase mortality in patients with underweight, whereas short education and smoking did so for colon cancer, and alcohol consumption for rectal cancer. Type 2 diabetes had minor influence on the associations. BMI‐related mortality in CRC may be modified by demographic, clinical and socioeconomic characteristics. Targeted interventions considering the high‐risk subgroups may help improve CRC outcomes.

What's New?

Obesity may worsen colorectal cancer prognosis, but the existing evidence remains inconclusive. This large nationwide cohort study comprehensively assessed how BMI interacts with demographic, clinical, and socioeconomic factors to influence colorectal cancer prognosis. Using high‐quality data, validated algorithms, and complete follow‐up, the authors found that BMI had little impact on colorectal cancer recurrence. However, BMI was associated with mortality, especially in younger patients and in males with obesity and in patients with underweight and comorbidities. These findings identify high‐risk subgroups and offer evidence to inform targeted interventions and improve colorectal cancer survivorship care.

Abbreviations

AP

attributable proportion

BMI

body mass index

CCI

Charlson Comorbidity Index

CI

confidence interval

CIF

cumulative incidence function

CRC

colorectal cancer

CUP Global

Global Cancer Update Programme

DCCG

Danish Colorectal Cancer Group

IQR

interquartile range

UICC

Union for International Cancer Control

1. INTRODUCTION

Colorectal cancer (CRC) patients treated with curative intent surgery have a 5‐year recurrence rate of ~20% and a survival rate of 73%–91%, depending on stage at diagnosis. ¹ , ² , ³ Meta‐analyses indicate that underweight and obesity increase the risk of CRC recurrence, cancer‐specific, and all‐cause mortality. ⁴ , ⁵ , ⁶ , ⁷ A potential central mechanism is the role of visceral fat in promoting tumor growth via pro‐inflammatory cytokines. In contrast, visceral fat may also act as an energy reserve during cancer cachexia, ⁸ , ⁹ potentially contributing to the obesity paradox (i.e., a protective effect of overweight and obesity). ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ¹⁰ Obesity‐related comorbidities and conditions associated with underweight, such as malnutrition, frailty, or chronic diseases may also contribute to excess mortality in CRC patients. ¹¹

A 2024 report from the World Cancer Research Fund Global Cancer Update Programme (CUP Global) stated that evidence on the relationship between body mass index (BMI) and CRC prognosis remains limited and inconclusive. It outlined methodological limitations, including heterogeneity in study designs, study populations, follow‐up, and healthcare settings, limiting comparability and generalizability. ⁴

Socioeconomic position may also influence CRC outcomes. Low socioeconomic position is associated with obesity, ¹² reduced participation in cancer screening and rehabilitation, ¹³ , ¹⁴ and with elevated risk of emergency surgery and mortality. ¹⁵ , ¹⁶ , ¹⁷ Furthermore, low socioeconomic position correlates with lifestyle factors like smoking and alcohol consumption, which may influence CRC prognosis. ¹⁸ , ¹⁹ , ²⁰ The CUP Global report suggested that lifestyle factors and patient medical histories may confound or act synergistically to modify the association of BMI with CRC prognosis. ⁴ Disentangling interactions between socioeconomic position, lifestyle, comorbidities, and BMI could help understand their impact on CRC prognosis.

We therefore leveraged high‐quality data from Danish population‐based registries to investigate the association between BMI and CRC recurrence and all‐cause mortality in CRC patients operated with curative intent. Distinctively, this also enabled evaluation of the synergistic effects of lifestyle factors, type 2 diabetes, comorbidities, and socioeconomic position on these associations.

2. MATERIALS AND METHODS

2.1. Data sources

At birth and immigration, Danish residents are assigned a unique civil registration number, enabling individual‐level linkage across registries and databases. Diagnosis and treatment of CRC are performed according to national multidisciplinary guidelines prepared by the Danish Colorectal Cancer Group (DCCG). ²¹ We combined information from the DCCG database, the Civil Registration System, the Cause of Death Registry, the Danish National Patient Registry, the Danish National Prescription Registry, the Danish Cancer Registry, the Danish National Pathology Registry, and administrative data from Statistics Denmark. See Supplemental materials & methods S1, for more information on the data sources.

2.2. Study population

We included all patients registered in DCCG with a diagnosis of Union for International Cancer Control (UICC) stage I–III CRC during 2002–2020, aged ≥30 years, treated with curative intent surgery, without a previous cancer, and with data on BMI recorded in DCCG. Patients also had to be alive and cancer‐free 6 months after surgery (i.e., index date). We excluded patients who emigrated within that period (Figure S1).

2.3. Body mass index

From the DCCG, we ascertained height and weight recorded at the time of surgery and calculated BMI in kg/m². We analyzed BMI as a continuous variable and in categories according to the World Health Organization's categorization: underweight (BMI <18.5 kg/m²), normal weight (BMI 18.5–24.9 kg/m²), overweight (BMI 25–29.9 kg/m²), class I obesity (BMI 30–34.9 kg/m², hereafter “obesity”) and class II–III obesity (BMI >35 kg/m²).

2.4. CRC recurrence and mortality

We defined CRC recurrence via a registry‐based algorithm incorporating data from the Danish Cancer Registry, the Danish National Pathology Register, and the Danish National Patient Registry (Figure S2). ²² We obtained information on death from any causes from the Civil Registration System. For supplemental analyses, we incorporated data on cause of death from the Cause of Death Registry. ²³

2.5. Covariates

From the DCCG, we collected information on pathological UICC stage, tumor localization, American Society of Anesthesiologists score, surgery year, surgical approach, surgical priority, smoking, and alcohol consumption. Using the Danish National Patient Registry, we collected and summarized information on comorbidities via the Charlson Comorbidity Index (CCI) using ICD‐10 codes (Table S1). ²⁴ We identified type 2 diabetes diagnosed any time before the index date, via ICD‐10 codes and prescriptions redeemed for glucose lowering drugs from the Danish National Prescription Registry. From Statistics Denmark, we collected information on education and cohabitation status (Table S2).

2.6. Statistical analyses

We followed patients from 180 days after the date of primary CRC surgery to recurrence/death, emigration, 5 years post‐diagnosis, or 31 December 2021. When examining recurrence, we included new primary cancers (except non‐melanoma skin cancer) and deaths as competing events. For colon and rectal cancer separately, we calculated cumulative incidence functions (CIFs) of recurrence and all‐cause mortality using the Aalen–Johansen estimator.

For the non‐linear associations, we modeled BMI as cubic splines in Cox regression models with knots at the 5th, 35th, 65th, and 95th percentiles. Based on the nadir in the mortality models, we used BMI = 25 kg/m² as reference (i.e., normal weight). When utilizing BMI categories, we calculated age‐, sex‐, and calendar‐period standardized incidence ratios for recurrence and standardized mortality ratios comparing the number of recurrences/deaths observed in BMI groups with that expected in the normal weight group. We calculated associated 95% confidence intervals (CIs) using the approximate bootstrap method. ²⁵

In complete case analyses, Cox regression models were adjusted for sex, age (cubic spline), calendar year (cubic spline), alcohol, smoking, cohabitation status, and education.

To assess synergistic effects of obesity‐related factors, we evaluated the extent of statistical interaction and effect measure modification by age, sex, tumor localization, CCI, UICC stage, type 2 diabetes, smoking status, alcohol consumption, education, and cohabitation status using stratification. We evaluated biological interaction on the additive scale by calculating interaction contrasts and the attributable proportion (AP) of the rates explained by the interaction of the combinations of the modifiers, by dividing the interaction contrast with the rates in the relevant strata. ²⁶ We separated analyses for any comorbidity (CCI ≥1) and for higher burden of comorbidity (CCI >3).

Post‐hoc, we added supplemental analyses on cancer‐specific mortality, recognizing the unknown validity of cause of death registrations. ²³ Here, we included deaths from other causes as competing events. We conducted several sensitivity analyses. To ensure a more homogeneous population, we restricted the study cohort to patients undergoing elective surgery, as emergency surgery often involves different clinical pathways and prognosis. ¹⁵ , ²⁷ Moreover, we restricted to patients who underwent surgery in 2010–2020 to account for advancements in treatment and survival outcomes. ²⁸

To assess selection bias, we included patients with missing data on BMI (n = 6441, Table S3), smoking (5%), alcohol (6%), education (4%) and cohabitation status (0.2%) and used the fully conditional specification method to impute missing covariates. ²⁹ In the imputation, we included the Aalen Johansen estimate of the outcome analyzed and the variables listed in Table 1 (except for surgical priority). We generated 10 imputed datasets and repeated the Cox regression models within each dataset and combined the spline curves and CIs using Rubin's Rule. ³⁰

TABLE 1.

Descriptive characteristics of stage I–III CRC patients who underwent curative intent surgery during 2002–2020 in Denmark by BMI group.

		Body mass index (kg/m2)
	Total cohort	<18.5 underweight	18.5–25 normal weight	25–30 overweight	30–35 obesity	35+ obesity class II–III
Total	33,926 (100.0)	952 (2.8)	14,548 (42.9)	12,455 (36.7)	4470 (13.2)	1501 (4.4)
Age, median (IQR)	70 (62; 77)	73 (64; 81)	71 (63; 78)	70 (63; 76)	69 (62; 74)	67 (59; 72)
Females, n (%)	15,439 (45.5)	751 (78.9)	7615 (52.3)	4595 (36.9)	1758 (39.3)	720 (48.0)
Education, n (%)
Short	9348 (27.6)	294 (30.9)	3752 (25.8)	3438 (27.6)	1371 (30.7)	493 (32.8)
Medium	16,350 (48.2)	412 (43.3)	6791 (46.7)	6137 (49.3)	2243 (50.2)	767 (51.1)
Long	6942 (20.5)	174 (18.3)	3352 (23.0)	2489 (20.0)	721 (16.1)	206 (13.7)
Missing	1286 (3.8)	72 (7.6)	653 (4.5)	391 (3.1)	135 (3.0)	35 (2.3)
Cohabitation, n (%)
Cohabiting/married	22,632 (66.7)	473 (49.7)	9433 (64.8)	8766 (70.4)	3034 (67.8)	942 (62.7)
Living alone	11,268 (33.2)	479 (50.3)	5115 (35.2)	3689 (29.6)	1436 (32.2)	559 (37.2)
Missing a	26 (0.1)
Type 2 diabetes, n (%)	4569 (13.5)	47 (4.9)	1236 (8.5)	1787 (14.3)	1016 (22.7)	483 (32.2)
CCI score, n (%)
None	23,109 (68.1)	590 (62.0)	10,290 (70.7)	8532 (68.5)	2847 (63.7)	850 (56.6)
1–2	8986 (26.5)	312 (32.8)	3607 (24.8)	3270 (26.3)	1302 (29.1)	495 (33.0)
3+	1831 (5.4)	50 (5.3)	651 (4.5)	653 (5.2)	321 (7.2)	156 (10.4)
Smoking, n (%)
Never smoked	12,713 (37.5)	280 (29.4)	5426 (37.3)	4664 (37.4)	1731 (38.7)	612 (40.8)
Previous smoker	13,209 (38.9)	238 (25.0)	5131 (35.3)	5279 (42.4)	1924 (43.0)	637 (42.4)
Current smoker	6266 (18.5)	391 (41.1)	3198 (22.0)	1880 (15.1)	615 (13.8)	182 (12.1)
Missing	1738 (5.1)	43 (4.5)	793 (5.5)	632 (5.1)	200 (4.5)	70 (4.7)
Alcohol, n (%)
Abstainers	7238 (21.3)	290 (30.5)	3034 (20.9)	2516 (20.2)	983 (22.0)	415 (27.6)
1–14 units per week	19,783 (58.3)	490 (51.5)	8640 (59.4)	7267 (58.3)	2591 (58.0)	795 (53.0)
>14 units per week	4778 (14.1)	91 (9.6)	1872 (12.9)	1943 (15.6)	672 (15.0)	200 (13.3)
Missing	2127 (6.3)	81 (8.5)	1002 (6.9)	729 (5.9)	224 (5.0)	91 (6.1)
ASA score, n (%)
Score 1	8450 (24.9)	179 (18.8)	4336 (29.8)	3086 (24.8)	734 (16.4)	115 (7.7)
Score 2	18,794 (55.4)	495 (52.0)	7695 (52.9)	7049 (56.6)	2707 (60.6)	848 (56.5)
Score 3+	6333 (18.7)	269 (28.3)	2345 (16.1)	2203 (17.7)	986 (22.1)	530 (35.3)
Missing	349 (1.0)	9 (0.9)	172 (1.2)	117 (0.9)	43 (1.0)	8 (0.5)
UICC stage, n (%)
Stage I	8349 (24.6)	173 (18.2)	3292 (22.6)	3273 (26.3)	1224 (27.4)	387 (25.8)
Stage II	13,953 (41.1)	470 (49.4)	6336 (43.6)	4891 (39.3)	1692 (37.9)	564 (37.6)
Stage III	11,624 (34.3)	309 (32.5)	4920 (33.8)	4291 (34.5)	1554 (34.8)	550 (36.6)
Tumor localization, n (%)
Colon	22,540 (66.4)	701 (73.6)	9668 (66.5)	8040 (64.6)	3054 (68.3)	1077 (71.8)
Rectum	11,386 (33.6)	251 (26.4)	4880 (33.5)	4415 (35.4)	1416 (31.7)	424 (28.2)
Year of surgery, n (%)
2002–2006	6076 (17.9)	211 (22.2)	2836 (19.5)	2212 (17.8)	632 (14.1)	185 (12.3)
2007–2013	11,759 (34.7)	348 (36.6)	5233 (36.0)	4281 (34.4)	1460 (32.7)	437 (29.1)
2014–2020	16,091 (47.4)	393 (41.3)	6479 (44.5)	5962 (47.9)	2378 (53.2)	879 (58.6)
Surgical priority, n (%)
Elective	32,148 (94.8)	856 (90.0)	13,642 (93.8)	11,910 (95.6)	4294 (96.1)	1457 (97.1)
Emergency	1765 (5.2)	96 (10.0)	906 (6.2)	546 (4.4)	176 (3.9)	44 (2.9)
Missing a	13 (0.0)

Abbreviations: ASA, American Society of Anesthesiologists; CCI, Charlson Comorbidity Index; UICC, Union for International Cancer Control.

^a Missing values were allocated to remaining categories based on their respective proportions within the BMI group to avoid cell counts <5, in accordance with Danish data protection rules.

All analyses were performed in SAS version 9.4 (SAS Institute Inc., Cary, NC). The forest plots were performed in RStudio version 2022.12.0 + 353.

3. RESULTS

Our cohort of 33,926 patients (Figure S3) had a median (interquartile range, IQR) follow‐up of 4.4 (1.7–4.6) years for recurrence, and 4.5 (2.7–4.6) years for mortality. At diagnosis, 3% of patients had underweight, 43% normal weight, 37% overweight, 13% obesity, and 4% class II–III obesity (Table 1). Patients with underweight were older (median age 73, IQR 64–81), and more often female (79%), smokers (41%), and living alone (50%). Patients with class II–III obesity were younger (median age 67, IQR 59–72), and more often had short education (33%), type 2 diabetes (32%), and a CCI score >3 (10%).

Overall, 5877 patients experienced CRC recurrence (CIF = 18.4%); 3612 patients (CIF = 17.1%) developed colon cancer recurrence, and 2265 (CIF = 21.1%) developed rectal cancer recurrence. For colon cancer recurrence by BMI groups, CIFs were fairly similar, ranging from 16.5% for overweight to 18.7% for class II–III obesity. For rectal cancer recurrence, CIFs were consistently higher across all BMI groups, though slightly increasing from 20.3% for underweight to 22.0% for class II–III obesity (Figure S4).

We did not detect a clear association of BMI with cancer recurrence (Figure 1, Table S4). Despite a tendency toward increasing risk of colon cancer recurrence in individuals aged 30–50 years (Figure 1B) and in males with class II–III obesity (Figure 1C), the null association was consistent across the remaining strata (Figures S5, S6).

FIGURE 1.

Cox regression models of the association between continuous BMI (modeled with a spline) and cancer recurrence among colon and rectal cancer patients. Panels (A–C) correspond to colon cancer, and panels (D–F) to rectal cancer. Results are shown for all patients (A, D), by sex (B, E), and by age group (C, F). Models were adjusted for sex, age, year of diagnosis, smoking, alcohol consumption, cohabitation status, and education. Age and year of diagnosis were modeled using restricted cubic splines. The graphs were plotted for BMI values between the 5th and 95th percentiles.

Overall, 6568 patients died (CIF = 21.8%): 4424 colon cancer patients (CIF = 22.1%) and 2144 rectal cancer patients (CIF = 21.1%). In both colon cancer and rectal cancer, this was highest for underweight (41.3% and 37.9%), and otherwise similar across BMI groups (Figure S7).

The nonlinear (Figure 2) and categorical models (Table S5) suggested U‐shaped associations—strongest for rectal cancer—with class II–III obesity being particularly associated with mortality in the youngest age group and in males. For rectal cancer, there was a tendency for increased risk for smoking and alcohol consumption, but estimates were imprecise. No differences were seen in analyses stratified by type 2 diabetes, comorbidities, or stage (Figures S8, S9). Results for cancer‐specific mortality were similar (Figure S10, Table S6).

FIGURE 2.

Cox regression models of the association between continuous BMI (modeled with a spline) and mortality among colon and rectal cancer patients. Panels (A–C) correspond to colon cancer, and panels (D–F) to rectal cancer. Results are shown for all patients (A, D), by sex (B, E), and by age group (C, F). Models were adjusted for sex, age, year of diagnosis, smoking, alcohol consumption, cohabitation status, and education. Age and year of diagnosis were modeled using restricted cubic splines. The graphs were plotted for BMI values between the 5th and 95th percentiles.

In colon cancer, several factors interacted to increase mortality among those with underweight (Table S7), particularly comorbidities (AP = 48% for CCI >3), while younger age (30–50 vs. 70+) interacted to decrease mortality (AP = –219%), meaning that older age was associated with increased mortality risk. In contrast, younger age interacted to increase mortality among those with class II–III obesity (AP = 149%, Table S8).

For rectal cancer, mortality in patients with underweight increased in males (AP = 25%), and among those with high alcohol consumption (AP = 21%) and CCI >3 (AP = 15%), whereas younger age decreased mortality (AP = –51%, Table S7). For class II–III obesity, male sex (AP = 34%) and younger age (AP = 112%) increased mortality (Table S8).

Findings were similar in all sensitivity analyses (Figures S11, S12).

4. DISCUSSION

We did not detect clinically relevant associations between BMI and CRC recurrence. In contrast, we observed higher mortality in patients with underweight and class II–III obesity. For class II–III obesity, the association was strongest in males, younger patients and those with rectal cancer. For both cancer sites, comorbidities and male sex contributed to increased mortality in patients with underweight.

Our findings for recurrence contradict with the three‐ to four‐fold increased risk in patients with class II–III obesity in the CUP Global report. Several factors may explain this discrepancy. Unlike the heterogeneous studies included in CUP Global, our study leverages a nationwide, population‐level cohort of stage I–III patients treated with curative intent surgery. We had long‐term, complete follow‐up and data on demographic and clinical factors. By stratifying our analyses on these key factors, we were able to identify high‐risk subgroups and yield a more precise evaluation of the role of BMI in CRC prognosis. Regarding mortality, other studies showed a J‐shaped association, with highest mortality in class II–III obesity. ⁴ , ³¹ The contrasting highest mortality among patients with underweight observed in our study may reflect advanced age, stage, comorbidities or frailty, which are more frequently associated with underweight. Although metastatic cancers were excluded, lower BMI might reflect pre‐existing illness—that is, pre‐diagnostic weight loss attributable to CRC. We therefore quantified the contribution of comorbidities to mortality in patients with underweight. Our findings support that underlying health conditions, particularly comorbidities, likely influence the observed association between underweight and mortality. ³²

Our observed sex differences seem biologically plausible. Adiposity impacts health differently in males and females due to physiological, hormonal, and anatomical differences. Males tend to accumulate visceral fat, increasing the risk of metabolic disturbances like insulin resistance, hypercholesterolemia, and higher rates of type 2 diabetes and cardiovascular disease. ³³

The higher mortality among younger patients with class II–III obesity gives cause for concern. Obesity may alter hormonal and immunological responses, accelerating aging. In older patients, age‐related immune decline may mask the impact of obesity. Furthermore, younger CRC patients are more likely to have hereditary cancer syndromes associated with poorer prognosis. ³⁴ Although we excluded patients aged below 30 years, the poorer prognosis in those aged 30–50 years may still partly reflect the influence of such syndromes. Moreover, a national CRC screening program offered to all residents aged 50–74 years was introduced in Denmark in 2014. ³⁵ Younger patients are not included, which may lead to diagnostic delays and more advanced disease at diagnosis (see Table S9). ³⁶ However, as stage did not modify the association between BMI at diagnosis and the outcomes, this is unlikely to explain our findings.

Our findings do not align with the obesity paradox, ³⁷ perhaps due to limited collider bias. Collider bias occurs when BMI influences CRC diagnosis, leading to an overrepresentation of patients with obesity compared with the general population. We did not observe such overrepresentation in our cohort, so this bias was likely minimized.

Several issues should be considered when interpreting our findings. Restricting our cohort to those with stage I–III disease who were alive and cancer‐free 180 days post‐surgery may have introduced survivor bias. We applied this restriction to avoid immortal time bias when implementing the recurrence algorithm and to allow for completion of primary treatment. ³⁸ Moreover, this restriction reduced the part of the association potentially driven by the effect of advanced cancer on BMI discussed above. This part was further reduced from the exclusion of patients with missing information on BMI, which was primarily those who underwent emergency surgery or were diagnosed in the earlier years, when CRC prognosis was generally poorer. Our imputed analyses and sensitivity analyses restricting to elective surgery confirmed, however, our findings. Anthropometric measures like waist and hip circumference could offer better information on adiposity, but they offer limited improvements in cancer risk prediction beyond BMI. ³⁹ However, pre‐diagnostic BMI might have offered additional insights on potential mechanisms—particularly given the possibility that some patients experienced weight loss before surgery, especially those with rectal cancer undergoing neoadjuvant chemo‐radiotherapy. If so, such an effect would likely lead to a modest underestimation of associations with high BMI and overestimation of associations with low BMI. Nevertheless, improved knowledge on the impact of BMI measured at the time of surgery remains a highly clinically relevant metric for clinical decision‐making regarding treatment and follow‐up care.

We lacked detailed information on adjuvant treatments for CRC. However, all patients in our cohort underwent surgery with curative intent, which remains the most important prognostic factor in CRC. ⁴⁰ The management of stage II and stage III CRC has remained relatively stable in Denmark, in line with international guidelines. ⁴⁰ , ⁴¹ , ⁴² For stage III patients, adjuvant chemotherapy was introduced in Denmark in 1997 (5‐FU), with oxaliplatin added from 2005. Since 2009, selected high‐risk stage II patients have also received such regimens. For rectal cancer, neoadjuvant radiotherapy is registered in the DCCG database, but fewer than 10% of our cohort received it. These patients are unlikely to undergo additional adjuvant systemic therapy.

Our study included pre‐treatment BMI data, which is not currently used to guide systemic therapy. ²¹ As adjuvant therapies would not have impacted BMI at CRC diagnosis, they do not meet the criteria for confounding. Therefore, we believe the lack of data on adjuvant therapies is unlikely to have biased our observed association. Nonetheless, it limited our ability to further risk stratify based on specific treatment regimens.

We examined cancer‐specific mortality to explore the possibility that the relationship between obesity and all‐cause mortality was driven by non‐cancer‐related deaths—this did not seem to be the case. However, cause of death data may be misclassified, as in Denmark, fewer than 10% of deaths are autopsied. ²³ This limits our ability to draw firm conclusions about obesity's differential impact on cancer‐related versus non‐cancer‐related deaths. Nonetheless, our inclusion of recurrence as a distinct endpoint provides important insight into the potential influence of BMI on CRC itself.

Our findings are generalizable to countries with similar healthcare settings. Countries lacking comprehensive welfare programs or with privatized healthcare may detect a higher impact of comorbidities and socioeconomic position on CRC outcomes. Even in Denmark, potential socioeconomic differences in both underweight and obesity are likely to become more pronounced in the future, with substantial differences in access to obesity treatment, for example, semaglutide use for weight loss. ⁴³ This could potentially exacerbate current socioeconomic differences in cancer prognosis and act synergistically on the association for BMI. The population of Denmark is mainly of European origin. In other populations, stronger associations may emerge, as adiposity and its consequences may differ by race and ethnicity. ⁴⁴ Our focus on patients undergoing curative intent surgery limits generalization of our findings to patients curatively treated with organ preservation strategies, which are increasingly used in rectal cancer.

In summary, our findings did not support an impact of BMI on the risk of CRC recurrence but showed a U‐shaped relationship between BMI and mortality, particularly for rectal cancer, in males, and younger patients with class II–III obesity. Among patients with underweight, older age and comorbidities emerged as critical factors exacerbating mortality risk. Tailored clinical strategies addressing the multifactorial influences of BMI in CRC survivorship, may improve patient outcomes and reduce mortality in persons living with or after CRC.

AUTHOR CONTRIBUTIONS

Cathrine F. Hjorth: Conceptualization; investigation; writing – original draft; methodology; visualization; writing – review and editing; project administration; formal analysis; validation; data curation. Maja H. Olsen: Conceptualization; writing – review and editing; investigation. Sinna P. Ulrichsen: Formal analysis; data curation; visualization; writing – review and editing; methodology; validation. Jesper Nors: Conceptualization; validation; methodology; writing – review and editing; investigation. Jannik Wheler: Writing – review and editing; methodology; investigation. Signe Borgquist: Conceptualization; writing – review and editing. Lene H. Iversen: Conceptualization; writing – review and editing. Susanne O. Dalton: Conceptualization; writing – review and editing; investigation; methodology. Deirdre Cronin‐Fenton: Conceptualization; investigation; funding acquisition; writing – original draft; writing – review and editing; project administration; supervision; methodology.

CONFLICT OF INTEREST STATEMENT

The authors declare no potential conflicts of interest.

ETHICS STATEMENT

The study was approved by the Danish Data Protection Agency (2022‐0367531, #2624) and the Danish Colorectal Cancer Group. It adheres to the General Data Protection Regulations. The use of registry‐based data for scientific studies in Denmark requires no consent from the participants.

Supporting information

Data S1. Supporting information.

DATA AVAILABILITY STATEMENT

The data is stored at secure servers at Statistics Denmark. Requests can be made directly to Statistics Denmark, the Danish Health Data Authority and The Danish Healthcare Quality Institute (DHQI) administrating the Danish registries. Further information is available from the corresponding author upon request.