To our knowledge, this study is the first to investigate the risk factors and reasons for refusing surgery in patients with liver metastases, providing insights beyond those typically derived from registry studies.

Patients who refused surgery were more likely to have a lower BMI and a longer duration since their primary tumor diagnosis. Lower BMI has been associated with poorer overall survival in CRLM, with the proposed mechanism being cancer cachexia [10]. However, this mechanism is not fully understood, and low BMI may instead reflect frailty, sarcopenia, or cachexia. Studies from the National Cancer Database and the Surveillance, Epidemiology, and End Results Database across cancer types have shown that refusal was more common among older, unmarried, non-white ethnic patients, those with higher comorbidity burden, or insufficient insurance coverage [6,7,8, 11,12,13,14,15,16,17,18,19,20,21]. Denmark’s universal healthcare system eliminates insurance status as a risk factor for surgery refusal, which is often significant in other healthcare settings [6, 7, 11,12,13,14, 16, 19, 20]. The refusal rate in this study is equivalent to colorectal cancer surgery refusal rates, reported at 0.6–3.5% [6, 17].

Unlike most registry-based studies, our research utilized detailed patient records, enabling a more nuanced understanding of risk factors. However, this reliance on records limited our sample size and statistical power.

Improved understanding of surgical refusal may facilitate shared decision-making, as awareness of patient concerns enables physicians to address them more effectively. In our study, a substantial proportion of refusals appeared related to treatment fatigue, with the primary reason being declining further surgery. This is further supported by the number of patients choosing no further treatment or opting for palliative chemotherapy. The reasons for refusal were comparable to a breast cancer study, finding that alternative therapy, while easy to blame for patients refusing standard healthcare, was not the only nor main reason [8]. However, CRLM patients were more likely to cite cumulative treatment burden and treatment fatigue compared to breast cancer patients, where psychiatric comorbidity was the main factor. This difference likely reflects the advanced disease stage and extensive treatment history of patients with CRLM.

Interestingly, disease intensity, as indicated by the number of MDT conferences, prior treatments, and surgeries, was not associated with increased refusal rates. Instead, longer disease duration influenced refusal, indicated by more months since primary tumor diagnosis. However, prior experiences with complicated surgeries, with several patients citing aversion to large or open surgery and hoping for a smaller intervention. These findings suggest that advancements in minimally invasive surgical techniques could address the needs of patients currently refusing surgery.

Accepting surgery resulted in a median survival time of 4.76 years, comparable to that observed in a similar Norwegian cohort (median survival 4.78 years) [22]. In our analysis, refusal of surgery was associated with poorer survival outcomes among patients with CRLM, corresponding to a median survival difference of 2.3 years. The survival curves diverged after approximately two years, which may be clinically relevant when counselling patients about the potential benefits of surgical treatment. However, due to the limited number of events, subgroup analyses assessing the differential impact of post-refusal treatment strategies, including delayed surgery, were not feasible.

Multivariable cox regression analysis showed that refusal of surgery almost doubled the risk of mortality. This finding remained significant when adjusting for CCI and ASA. While residual confounding, such as increased frailty not captured by CCI or ASA, cannot be excluded, comparable tumor characteristics between groups suggest that refusal itself may impact survival. These findings may contribute to more informed discussions with patients regarding expected prognosis and the potential consequences of refusing surgery.

This impact of refusing surgery was also found across cancer types. In CRC, refusing surgery was associated with an increase in mortality risk (HR: 2.32–5.98) [11, 12, 14]. In other cancer types, including hepatocellular carcinoma, breast cancer, head and neck and ovarian cancer, the same pattern was seen (HR: 2.42–2.88) [15, 16, 18, 20].

The retrospective design of this study enabled detailed data collection. However, the principal limitation is the small sample size, which reduces statistical power across all analyses. It also limited our ability to include additional prognostic variables in the Cox regression model—such as tumor size, CEA level, age, presence of extrahepatic metastases, and KRAS status—as including multiple variables would have resulted in model overfitting. The small sample size, due to the rarity of surgery refusal and the number of cases available over the 10-year study period, may also have restricted the detection of some risk factors. Although not statistically significant, baseline characteristics suggest that the likelihood of refusal increases with factors such as older age, comorbidity burden, psychiatric disorders, and the loss of a partner.

There was a discrepancy between the referral performance status (PS) recorded in the dataset used for patient matching and the preoperative PS documented in the patient records. Baseline characteristics showed that a higher proportion of patients in the refusal group had poorer PS compared with those who accepted surgery, possibly reflecting selection bias. However, PS at the preoperative assessment was not available in the database used for matching controls. Although data extraction from medical records enabled a high level of detail and improved data validity, missing or inaccurately recorded information remains a potential limitation.

Furthermore, this study was done retrospectively, reducing the quality of the qualitative outcomes, and making it impossible to further investigate the reasons other than what is recorded in medical journals, as most patients who refused surgery were dead. However, this study cannot be done prospectively, as the occurrence of refusal is so rare. There is a risk of residual confounding, as the cohort is small, the poorer survival might not be due to refusal, but rather the differences between the groups that we cannot adjust for.

As the incidence is small, individual physicians will not encounter many patients refusing treatment, so investigating these factors will help better physicians’ knowledge. Quantifying the impact of refusal on survival has not been done before in CRLM. Understanding reasons for surgery refusal and its impact on survival will help the physician provide individual guidance in the preoperative decision-making. It also challenges the misconception that patients who refuse conventional medicine, are prone to select alternative therapy.

While individual numbers for each cancer type are small, the aggregate number of patients refusing treatment across cancer types highlights the importance of understanding these factors. Further exploration includes better and more in-depth registration of reasons behind refusal in databases, not only in liver surgery but across cancer diseases. This will enable us to adjust treatment according to patient preferences, for example avoiding open surgery if possible or increasing neoadjuvant therapy to reduce the surgical burden. Investigations should strive to make comparisons across cancer types, to gain an understanding of which factors are transferable and make a larger impact in this rare area.