Introduction

The global cancer burden is expected to grow exponentially by 50% over the next two decades [1]. The resulting needs arising from the booming population of cancer patients and survivors demand greater attention. Patients and survivors of cancer who are treated with conventional therapies often experience treatment-related toxicities and chronic symptoms [2]. The use of traditional, complementary, and integrative medicines (TCIM) is becoming common among patients with cancer to improve cancer outcomes, manage the complications of therapies and cancer-related symptoms, and address their holistic needs [3]. A recent systematic review reported that approximately half of cancer patients worldwide used TCIM in 2018, demonstrating a significant increase in use from 25% prior to 1990 [3]. Of the TCIM modalities, dietary supplements are the most commonly used by patients with cancer in the US (36–81%) [4, 5] and the UK (41%)[6].

Dietary supplements refer to products taken orally that contain “dietary ingredients” (such as vitamins, minerals, herbs, amino acids) that are intended to supplement the diet and are marketed for use in dosage form [7, 8]. Evidence for the association between the use of dietary supplements and cancer-related outcomes, such as overall survival and mortality, is conflicting and inconclusive [9,10,11,12,13]. There are also concerns about the generalizability of the findings of previous studies and reviews, as many have focused on vitamins and mineral supplements and have paid less attention to the effects of herbal and other supplements [9, 13,14,15]. Moreover, most studies have been conducted in patients with breast cancer only, which may not be generalizable to other cancer types [11,12,13, 16, 17]. Furthermore, previous studies usually focused on the effects of a single supplement [9, 18]. While these studies provide more precise estimates for the efficacy/effectiveness of individual supplements, they may not address real-world scenarios and practice settings, where dietary supplement users tend to take combinations of supplements or multi-ingredient products [5, 11, 19].

With the overarching aim of mimicking real-life clinical scenarios, where patients may take multiple types/classes of supplements, this study examined the association between the overall use of dietary supplements by patients diagnosed with cancer and the risk of mortality (all-cause and cause-specific mortality) in a large population-based cohort in the UK. In addition, the study explored the associations between the use of different classes of supplements (vitamins, minerals, and non-vitamin non-mineral [NVNM] dietary supplements) and the risk of mortality.

Methods

This study was registered with the UK Biobank (ref.: 74158), and is reported according to the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) guidelines [20].

Study population

The UK Biobank is a large population-based cohort of approximately 500,000 participants in the UK. It facilitates investigation of a wide range of complex diseases of middle and old age [21]. The UK Biobank study first recruited participants aged 40–69 across the UK from 2006 to 2010 and has since conducted repeated assessments. The methods used have been reported in detail elsewhere [21].

In this study, participants were excluded if (1) they had not been diagnosed with malignant cancer at any time from prior to baseline recruitment to subsequent follow-up visits or (2) they were not recruited in subsequent visits after their cancer diagnosis to provide responses to questions related to the use of dietary supplements.

Cancer ascertainment

UK Biobank is linked to national cancer registries (Health and Social Care Information Centre and the National Health Service Central Register) [22]. Cancer diagnoses were coded according to the International Classification of Diseases, 9th and 10th revisions (ICD-9 and ICD-10). We included only malignant neoplasms (ICD-9: 140–208; ICD-10: C00–C97), except non-melanoma skin cancer (ICD-9: 173; ICD-10: C44), that had been diagnosed on or before July 31, 2019 (the censoring date of the cancer registry when the data set was acquired).

Ascertainment of the use of supplements

The participants were asked “Do you regularly take any of the following?” and they were provided with a list of supplements in a touchscreen questionnaire. If the participants forgot to record any answers or took supplements that were not listed, they were able to provide the missing details during a verbal interview. The type of supplements that were considered in this study included vitamins; minerals; and NVNM dietary supplements, such as herbs and amino acids. The ingredients that were classified as supplements are listed in Supplementary Table 1.

Supplement users were defined as those who selected any of the supplements mentioned in the touchscreen questionnaire or listed any supplements at the interview stage during any of the assessments (baseline or follow-up) after the date of their cancer diagnosis. Participants who did not provide answers to the questions (did not select any of the supplements or indicated “None of the above”) in the touchscreen questionnaire and did not state the use of any supplements during their verbal interviews were excluded.

Ascertainment of mortality outcomes

The UK Biobank obtained comprehensive mortality data (date and cause of death) from the Information Centre (England and Wales) and the National Health Service Central Register Scotland [21]. ICD-10 codes were used in the death records to identify the causes of death. In the current study, all-cause and specific mortality due to cancer (C00–C97), cardiovascular diseases (CVD, I00–I99), respiratory diseases (J09–J98), digestive diseases (K20–K93), and nervous system diseases (G00–G99) were analyzed. The participants were followed up from the date of recruitment (2006–2010) until the date of death or the end of the follow-up period (March 23, 2021 or earlier if they were lost to follow-up), whichever occurred first.

Assessment of confounders

Potential confounders commonly associated with the use of dietary supplements were selected a priori based on data from the literature [3, 6, 23,24,25,26]. These included sociodemographic information (sex, age, Townsend Deprivation Index score, and educational level) and lifestyle behaviors (smoking status, alcohol consumption, physical activity, fruit and vegetable intake, and body mass index [BMI]), which were collected from the baseline touchscreen questionnaire. Clinical confounders included cancer types, age at cancer diagnosis and comorbidities [24, 26]. The Charlson Comorbidity Index (CCI) was used to quantify the comorbidity burden of the participants prior to their cancer diagnosis [27]. The details of the confounders are presented in Supplementary Table 1.

Statistical analyses

Descriptive statistics were used to summarize the pattern of supplement use among cancer patients in the cohort and the baseline characteristics of the participants. Multiple imputation with chained equations was used to deal with missing values and reduce inferential bias [28]. All of the factors included in the multivariate model were included in the imputation model. As the proportion of missing data was small (< 5%) for all of the covariates, five imputed data sets were deemed sufficient [29]. Logistic regression analysis was performed to verify the significant associations between the pre-identified confounders listed in the previous section (Supplementary Table 2).

Associations between the use of dietary supplements and the risk of all-cause and cause-specific mortality were analyzed using Cox proportional hazards models. Three models were run, including a crude model (Model 1); a model adjusted for age and sex (Model 2); and a model fully adjusted for factors identified a priori, including age, sex, socioeconomic factors, lifestyle factors, cancer types, age at cancer diagnosis, and CCI score (Model 3). Index date was defined as the date of primary cancer diagnosis. A competing risk analysis was performed to measure the associations between supplement use and cancer mortality, while considering death due to causes other than cancer as competing risks. Subgroup analyses were performed for different classes of supplements (vitamins, minerals, and NVNM dietary supplements) and their combinations. Stratified analyses were performed to assess the potential modifying effects of factors identified earlier, using the pre-determined groupings presented in Supplementary Table 1.

Three sensitivity analyses were performed. First, all of the analyses described above were repeated using the non-imputed data set. Second, the association between supplement use and the risk of cancer mortality were analyzed without considering competing risks. Third, the analyses were repeated after excluding all patients diagnosed with cancer before 2006 (the initial year of recruitment for the UK Biobank cohort), to reduce the possibility of survival bias. All of the statistical analyses were performed using R, version 4.0.3. A p value < 0.05 was considered statistically significant.

Results

Baseline characteristics

Of the 502,412 total participants in the cohort, participants were excluded if (1) they had not been diagnosed with malignant cancer at any time from prior to baseline recruitment to subsequent follow-up visits (n = 446,233) or (2) they were not recruited in subsequent visits after their cancer diagnosis and, therefore, had not provided answers to questions related to the use of dietary supplements (n = 25,940). Finally, 30,239 participants were included in the analysis (Supplementary Fig. 1).

The mean age of the participants upon cohort entry was 60.0 (SD = 7.0) years, and 61.9% of them were female (Table 1). The mean age at first cancer diagnosis was 52.8 (SD = 10.3 years). More than half (n = 17,464, 57.8%) of the participants reported the use of supplements after their cancer diagnosis. Most supplement users reported the use of dietary supplements (non-vitamin, non-mineral and non-herbal) (n = 12,460, 41.2%), followed by vitamins (n = 12,043, 39.8%) (Fig. 1). Fish oil (n = 10,234, 33.8%) and multivitamins (n = 6959, 23.0%) were the most commonly used individual supplements. Around 70% of the supplement users (n = 11,991/17,464) took more than one supplements, and more than half (n = 10,222/17464, 58.5%) of them used multiple categories of supplements. The top three combinations are multivitamins–fish oil (n = 4324, 14.3%), glucosamine–fish oil (n = 3999, 13.2%) and vitamin D–calcium (n = 2991, 9.9%).

Table 1 Baseline characteristics of eligible participants (N = 30,239)
Full size table
Fig. 1
figure 1

Pattern of supplement use among cancer patients in the UK Biobank cohort (n = 30,239)

Full size image

Compared with non-users, those who reported supplement use were more likely to be female, have a higher socioeconomic status, and be more educated (Supplementary Table 2). Supplement users generally had a healthier lifestyle, as they were less likely to be obese or current smokers, were more physically active, and had higher fruit and vegetable intake. They also tended to have higher comorbidity burden, were older at cancer diagnosis, and diagnosed with cancers that have poorer average prognosis.

Associations of supplement use with all-cause mortality and cause-specific mortality

At a median follow-up of 11.9 years (IQR = 10.2–13.6 years), 5577 all-cause deaths were registered. Cause-specific mortality included 4208 cancer-related deaths, 516 CVD-related deaths, 213 respiratory disease-related deaths, 145 digestive disease-related deaths, and 126 nervous system disease-related deaths. In all three models (crude, age and sex-adjusted, and multivariate), significant inverse associations were found between supplement use and the risk of all-cause and cancer-related mortality (Table 2). The adjusted hazard ratios (aHRs) associated with supplement use were 0.95 (95% confidence interval [CI] = 0.90–0.99, p = 0.04) for all-cause mortality and 0.89 (95% CI = 0.83–0.95, p < 0.001) for cancer-related mortality. No significant associations were observed between supplement use and the risk of cause-specific mortality due to CVD, respiratory, digestive, or nervous system diseases.

Table 2 Association of supplement use with risk of overall and cause-specific mortality
Full size table

Subgroup analyses were performed to analyze the associations of different classes of supplements and their combinations with all-cause and cause-specific mortality (Table 3 and Supplementary Table 3). For all-cause mortality, Only NVNM dietary supplements were significantly associated with a lower risk of all-cause mortality (aHR = 0.88, 95% CI = 0.83–0.93, p < 0.001) in the multivariate model. The use of both vitamins (aHR = 0.93, 95% CI = 0.87–0.99, p = 0.04) and NVNM dietary supplements (aHR = 0.88, 95% CI = 0.82–0.94, p < 0.001), was associated with a small decrease in the risk of cancer-related mortality. The combined use of vitamins with NVNM supplements and all three types of supplements were associated with a lower risk of all-cause and cancer-related mortality (aHR = 0.85–0.91, all p < 0.05). Notably, mineral supplements were found to be associated with an increased risk of mortality due to respiratory (aHR = 1.53, 95% CI = 1.12–2.11, p = 0.008) and digestive diseases (aHR = 1.56, 95% CI = 1.06–2.31, p = 0.03), and similar findings were observed for the combined use of vitamins and minerals (aHR = 1.57–1.64, all p < 0.03).

Table 3 Subgroup analyses of association of supplement use with risk of overall and cause-specific mortality
Full size table

Stratified analyses were conducted and significant interactions were found between supplement use and sex on the risk of all-cause mortality and cancer-related mortality (p = 0.02) (Supplementary Fig. 2). The associations between supplement use and a lower risk of mortality were observed among female patients, but not male patients. No significant interaction effects were observed between supplement use and other sociodemographic, lifestyle, or clinical factors on the risks of all-cause or cancer-related mortality, indicating that the observed associations were not significantly modified by other specific risk factors.

Sensitivity analyses

The results of our sensitivity analyses of the associations between the overall use of dietary supplements and all-cause and cause-specific mortality were mostly consistent with the results of the main analyses (Supplementary Table 4). Significant inverse associations between supplement use and all-cause and cancer-related mortality were still observed when we analyzed the non-imputed data set. However, the associations between supplement use and all-cause mortality were not observed after excluding participants who were diagnosed with cancer before 2006.

For the subgroup sensitivity analyses that only included patients diagnosed with cancer after 2006, the use of mineral supplements was associated with a lower risk of all-cause mortality (aHR = 0.90, 95% CI = 0.85–1.00, p < 0.05) and cancer-related mortality (aHR = 0.85, 95% CI = 0.75–0.96, p = 0.01) but the associations were borderline significant (Supplementary Tables 5 and 6). Similarly, the associations between NVNM supplement use and mortality were insignificant in this sensitivity analysis. All the combinations were associated with a lower risk of all-cause and cancer-related mortality (p < 0.05). No significant associations were observed between supplement use and other types of cause-specific mortality.

Discussion

This is one of the largest prospective cohort studies to investigate the association between supplement use and the risk of mortality in cancer patients. The large sample and the availability of well-defined mortality data from UK Biobank enabled analyses that considered a wide range of potential confounders and different categories of supplements. Previous studies have mainly focused on the effect of individual supplements; however, in our study, we found that more than half of the participants consumed more than one classes of supplements. Hence, this warranted our approach to analyze supplements as a whole and by categories rather than by individual supplements. It provided a broader picture of how supplements use as a growing phenomenon among patients with cancer can impact their mortality outcomes. We found that the use of supplements was associated with a modest reduction in the risk of all-cause mortality and cancer-related mortality, after adjusting for clinically relevant confounders. Of the three types of supplements analyzed, the use of NVNM dietary supplements was associated with a lower risk of all-cause and cancer-related mortality, while vitamin use was associated with a marginally lower risk of cancer-related mortality. Our findings should be interpreted cautiously considering the small effect sizes. Moreover, the associations were small and could be due to other unmeasured confounders not able to be addressed in this study besides supplement usage. Nevertheless, these results derived from real-world data have significant implications for the development of future research priorities for evaluating the role of supplements in cancer care. This will subsequently lead to refinement of guidelines and recommendations for the evidence-informed use of supplements by cancer patients.

Our finding that vitamin and NVNM dietary supplement use reduced mortality was consistent with previous results. For example, a study of four cohorts of breast cancer survivors revealed the potential protective effects of vitamin C and E and multivitamins against the risk of death [17], while recent reviews and meta-analysis suggested the potential role of vitamin D in increasing cancer survival rates [9, 10, 30, 31]. Our findings suggested a consistent association between NVNM dietary supplement use and reduced mortality. However, evidence for the benefits of NVNM dietary supplements is relatively limited compared with that of vitamins. A meta-analysis showed that omega-3 supplementation increased overall survival in pancreatic cancer patients [32], while preliminary evidence showed some survival benefits in patients taking fish oil supplements [33, 34]. Overall, our study supported the finding of these studies that vitamins and NVNM dietary supplements may confer survival benefits to cancer patients in general. Future studies should continue to explore the potential benefits of these supplements, especially NVNM supplements, in cancer patients and survivors.

There are multiple potential mechanisms, whereby these supplements have a protective effect against cancer-related mortality. First, while inflammation is involved in the pathophysiology of cancer [35, 36], elevated concentrations of inflammatory markers, such as C-reaction protein, have been shown to be associated with an increase in all-cause mortality and cancer-related mortality [37, 38]. Many vitamins and NVNM supplements, such as vitamin D [39], coenzyme Q10 [40], fish oil [34, 41], and glucosamine [41, 42], have anti-inflammatory effects. Another potential mechanism is the alteration of metabolism. For example, omega-3 may help correct cancer cachexia, which is a common syndrome that is inversely associated with cancer survival [32, 43]. Further clinical or epidemiological studies are needed to confirm other potential mechanisms, such as antioxidative effects of various supplements via modulating intracellular signaling pathways of cancer cells and the microenvironment [44, 45], as well as to explore in depth the potential benefits of supplements for clinical outcomes that affect mortality and the quality of life of patients diagnosed with cancer, and whether specific groups of patients may benefit more from dietary supplementation.

Nevertheless, our findings regarding mineral supplement use and mortality remained inconclusive. Our main analysis suggested that minerals may lead to higher rates of mortality among cancer patients, whereas the sensitivity analysis suggested otherwise. This discrepancy may be attributable to differences in treatment efficacy, as patients diagnosed with cancer before 2006 may have received less effective cancer treatment, leading to an increase in susceptibility to late effects and a higher risk of mortality. As treatment data were not available for this cohort, we were unable to explore its confounding effects. Therefore, our results should be interpreted cautiously in view of changes in cancer treatment strategies over time and the consequent potential for survival bias. Future data enhancements (such as the availability of information about cancer treatment) in the UK Biobank may enable more comprehensive analyses of the effect of dietary supplementation on mortality [46].

Our findings obtained from real-world data may help to optimize current policies and recommendations for the use of TCIM in cancer patients, particularly in low- and middle-income countries, where TCIM modalities play an important role in primary health care [47]. Our finding that vitamins and NVNM dietary supplements may have potential protective effects may guide the future recommendation of such low-cost and easily accessible supplements to improve cancer outcomes globally, after validation of its effectiveness and safety in randomized trials on specific supplements. International collaboration and effectively implemented multinational randomized trials and large-scale observational studies are needed to generate high-quality evidence on the effectiveness and safety of supplements. If these supplementations are found to be effective and safety, their usage in routine cancer care would benefit patients worldwide.

Despite using a large sample and a prospective observational design with well-characterized mortality outcomes, this study has some potential limitations. First, the use of dietary supplements after the cancer diagnosis was entirely self-reported, potentially leading to recall inaccuracies during the verbal interviews and introducing immortal time bias [48]. There may be time periods between patients being diagnosed with cancer and the questionnaire assessments or interviews that could be misclassified as users or non-users. Therefore, our results may have to be interpreted cautiously. Second, there were a range of supplements included in the study which have different pharmacological effects. The small associations found in the study may, therefore, be due to other confounders not addressed instead of supplements alone. Besides cancer treatment data, some information, including cancer staging, indications for supplement use, and the use of other TCIM, is also not available. The stage of cancer may affect the cancer prognosis and mortality rate, while may also influence the decision on supplement use. Future studies can include these questions to obtain a more complete picture of supplement or TCIM use among cancer patients. Furthermore, the UK Biobank study is known to have a low participation rate and a selection bias toward healthy volunteers with relatively low mortality rates and healthy lifestyles [49]. All of the participants provided information on supplement use at baseline but to avoid potential misclassification, we only included patients who also reported their supplement use status after their cancer diagnoses in this analysis. This has led to the exclusion of a considerable number of subjects from our analysis as only half of participants who completed baseline assessment were invited or recruited in the repeated assessment visit or subsequent studies by the UK Biobank. However, many studies have shown that this cohort may still provide valid inferences of risk factors and exposure-disease associations that are generalizable [49, 50]. To further increase the generalizability of the findings, they should be further validated in other large cohorts, with data on overall supplement use rather than just data for a few individual supplements.

Conclusions

More than half of the participants in the UK Biobank cohort reported using supplements after their cancer diagnosis, reflecting the popularity of supplement use among the adult cancer population. The use of supplements was associated with a small reduction in the risk of all-cause mortality and cancer-related mortality, after adjusting for clinically relevant factors. However, these associations should be interpreted cautiously due to the variations among supplements and the small effect size, and therefore, the effects may be not due to supplements alone but other unmeasured confounders. These findings warrant future research, including multinational randomized trials and large-scale observational studies, to investigate the potential of specific supplements, particularly NVNM supplements, in improving clinical outcomes that affect mortality and other key outcomes in patients with cancer, thereby guiding the integration of dietary supplements into routine cancer care in the future.