Volume 125, Issue 13 p. 2309-2317
Original Article
Free Access

Preventive drugs in the last year of life of older adults with cancer: Is there room for deprescribing?

Lucas Morin MS

Corresponding Author

Lucas Morin MS

Aging Research Center, Karolinska Institutet, Stockholm, Sweden

Corresponding author: Lucas Morin, MS, Aging Research Center, Karolinska Institutet, Tomtebodavagen 18 A, 171 77 Stockholm, Sweden; [email protected]

Search for more papers by this author
Adam Todd MPharm, PhD

Adam Todd MPharm, PhD

School of Pharmacy, Faculty of Medical Sciences, Newcastle University, Newcastle, United Kingdom

Search for more papers by this author
Stephen Barclay MA, FRCGP, MD

Stephen Barclay MA, FRCGP, MD

Department of Public Health and Primary Care, University of Cambridge, Cambridge, United Kingdom

Search for more papers by this author
Jonas W. Wastesson PhD

Jonas W. Wastesson PhD

Aging Research Center, Karolinska Institutet, Stockholm, Sweden

Search for more papers by this author
Johan Fastbom MD, PhD

Johan Fastbom MD, PhD

Aging Research Center, Karolinska Institutet, Stockholm, Sweden

Search for more papers by this author
Kristina Johnell MPharm, PhD

Kristina Johnell MPharm, PhD

Aging Research Center, Karolinska Institutet, Stockholm, Sweden

Search for more papers by this author
First published: 25 March 2019
Citations: 28
Clinical data and individual data from the Swedish Prescribed Drug Register cannot be made publicly available. However, additional information can be made available upon reasonable request to the authors.



The continuation of preventive drugs among older patients with advanced cancer has come under scrutiny because these drugs are unlikely to achieve their clinical benefit during the patients’ remaining lifespan.


A nationwide cohort study of older adults (those aged ≥65 years) with solid tumors who died between 2007 and 2013 was performed in Sweden, using routinely collected data with record linkage. The authors calculated the monthly use and cost of preventive drugs throughout the last year before the patients’ death.


Among 151,201 older persons who died with cancer (mean age, 81.3 years [standard deviation, 8.1 years]), the average number of drugs increased from 6.9 to 10.1 over the course of the last year before death. Preventive drugs frequently were continued until the final month of life, including antihypertensives, platelet aggregation inhibitors, anticoagulants, statins, and oral antidiabetics. Median drug costs amounted to $1482 (interquartile range [IQR], $700-$2896]) per person, including $213 (IQR, $77-$490) for preventive therapies. Compared with older adults who died with lung cancer (median drug cost, $205; IQR, $61-$523), costs for preventive drugs were higher among older adults who died with pancreatic cancer (adjusted median difference, $13; 95% confidence interval, $5-$22) or gynecological cancers (adjusted median difference, $27; 95% confidence interval, $18-$36). There was no decrease noted with regard to the cost of preventive drugs throughout the last year of life.


Preventive drugs commonly are prescribed during the last year of life among older adults with cancer, and often are continued until the final weeks before death. Adequate deprescribing strategies are warranted to reduce the burden of drugs with limited clinical benefit near the end of life.


In high-income countries, individuals aged ≥70 years now account for nearly two-thirds of cancer-related deaths.1 Chronic multimorbidity thus has become the norm rather than the exception in oncology,2 and is associated with poorer chances of survival and with a higher burden of functional impairments and physical symptoms.3 Multimorbidity also comes with a higher burden of long-term pharmacological treatments. In the United States and in Europe, approximately 40% of individuals aged ≥65 years use ≥5 drugs concomitantly.4, 5 This polypharmacy is particularly problematic among older individuals with advanced cancer6 because the potential to develop serious drug-drug interactions is amplified by the use of anticancer agents and complementary medicines.7, 8 Moreover, the probability of experiencing adverse drug reactions increases because the main pharmacokinetic parameters are affected not only by age but also by the physiological impact of cancer (eg, modified drug absorption due to gastrointestinal symptoms or to impairments in the gut wall function, a decrease in the volume of distribution caused by weight loss, or renal impairment due to the nephrotoxicity of chemotherapy).9, 10

Beyond pharmacology, polypharmacy within the context of advanced cancer also raises important questions from a clinical and ethical viewpoint. As cancer progresses and the prognosis worsens, the net benefit of each additional medicine gradually decreases while the risk of harm increases. This “law of diminishing returns” makes the continuation or initiation of long-term treatments particularly questionable for older patients with advanced cancer. Preventive drugs are prescribed either to avert or delay the onset of a disease among individuals who are considered to be at high risk of developing that disease in the future (primary prevention), or to avoid the recurrence of a condition that the patient experienced in the past (secondary prevention). These drugs typically need several years before the physiological and biological changes that they produce translate into measurable and clinically meaningful health outcomes. Thus, the time until benefit of preventive agents often is much longer than the remaining lifespan of an older adult with a serious illness.11 Recent randomized controlled trials have demonstrated that lipid-lowering medications can safely be deprescribed among older adults with a limited life expectancy, and that the discontinuation of antihypertensives among individuals without cardiovascular disease is safe in the short term.12, 13 Other long-term treatments such as bisphosphonates retain their effect 3 to 5 years after their withdrawal.14 Nevertheless, a few observational studies have reported that preventive medications are prescribed during the last year of life among patients with life-limiting disease, and have cast doubt on the benefit of these treatments.15 To the best of our knowledge, there has been limited investigation to date of the continuation and discontinuation of medications throughout the last months of life and with little information available regarding the costs of these medications and about potential variation across cancer types. Therefore, the objective of the current study was to evaluate the prescribing of preventive drugs throughout the final year of life among older adults who died with cancer across Sweden, and to estimate the direct costs of preventive drugs.

Materials and Methods

Study Design and Data

The current study was a retrospective cohort study based on routinely collected data in Sweden, a country with a universal health care system. Data from the national Cause of Death Register were linked through deterministic matching to the total population register, the National Patient Register, the Swedish Prescribed Drug Register, the social services register, and the Swedish Register of Education. The Regional Ethical Review Board in Stockholm approved the study.

Study Population

The current study included older adults aged ≥65 years who died in Sweden between 2007 and 2013 because these were the most recent available data. Decedents were considered as eligible for inclusion if a diagnosis of a solid cancer (International Statistical Classification of Diseases and Related Health Problems, 10th revision [ICD-10] codes C00-C76 and C80) was reported either in a hospital discharge report during the last 2 years of life or as an underlying or contributing cause of death. We decided a priori to exclude decedents with missing cause of death, those with missing drug prescription history throughout the last 6 months of life, and those who remained hospitalized continuously during the last 3 months before death. Older adults with concomitant hematological malignancies (ICD-10 codes C81-C95) also were excluded to select a homogenous population of individuals diagnosed only with solid cancers. Indeed, previous studies have shown that individuals with hematological malignancies experience a rapid functional decline at the end of life, which makes survival prediction particularly challenging. The potential for cure until late in the course of the disease trajectory differentiates these older patients from those dying of a solid cancer.16, 17


The use and cost of preventive drugs during the last 12 months of life were the main study outcomes. Preventive drugs with a questionable benefit near the end of life have been identified in a recent systematic review of the literature,15 and include drugs for diabetes, vitamins, mineral supplements, antithrombotic agents, antihypertensives, statins, bisphosphonates, and medications for chronic anemia. The list of corresponding Anatomical Therapeutic Chemical classification codes is available in Supporting Table 1.

We computed monthly exposure to specific drug classes based on data from the Swedish Prescribed Drug Register, which contains detailed information regarding all prescription drugs delivered in community pharmacies in Sweden since 2005 (including those drugs dispensed to nursing home residents, with the exception of a few facilities with their own drug storeroom). Methods for constructing periods of drug exposure have been presented in detail elsewhere,5, 18 and are illustrated in Supporting Figure 1A. Continuation of preventive drugs was calculated as the percentage of older adults who still were receiving preventive drugs during the last month before death among those exposed 1 year before, whereas initiation was calculated as the percentage of older adults who had therapy initiated with preventive drugs during the last year of life. Drug costs were estimated through a 2-step approach, as described in Supporting Figure 1B. We first divided the total cost of each purchase by the number of days covered to obtain the average daily cost. Second, we multiplied this average daily cost by the expected number of days of exposure during a given month, which allowed for the distribution of drug costs according to the assumed length of exposure. This approach provides a more realistic estimate of the costs instead of artificially concentrating all expenditures at the purchase date. Drug costs were standardized using the harmonized index of consumer prices with 2013 as the reference year to correct for inflation over time and then were converted from the Swedish currency (krona; SEK) into US dollars (USD) based on Sweden’s Central Bank annual average exchange rate from January 1 to December 31, 2013 to facilitate international comparisons (1 SEK = 0.1535 USD).

Assessment of Individual Characteristics

Sex and date of birth were extracted from the total population register and were cross-validated with data reported on the study participants’ death certificates. Solid malignancies were categorized into 14 distinct locations. Details regarding the corresponding ICD-10 codes are presented in Supporting Table 2. The overall burden of chronic multimorbidity was measured with a recently validated tool that captures a set of 60 distinct chronic diseases based on different data sources (contributing causes of deaths, inpatient and outpatient diagnoses reported during the last 3 years of life, and specific drugs unequivocally linked to chronic conditions).19 Living arrangement at the time of death was defined as “community” or “nursing home,” whereas the place of death was reported as either “hospital” or “usual place of living.” The decedent’s level of education was categorized as “primary,” “secondary,” and “tertiary” education in accordance with the International Standard Classification of Education.

Statistical Analysis

Multivariable quantile regressions were used to model drug costs across different cancer types while controlling for sex, age, number of chronic diseases, living arrangement, and level of education. Although linear regression allows for modeling the mean of an outcome, quantile regression is used to model quantiles of the outcome when the distribution of the outcome is highly skewed.20 Beta coefficients obtained from quantile regression models can be interpreted as the adjusted median difference (AMD) in costs compared with the reference group, and are reported together with their 95% confidence intervals (95% CIs). We compared the results with estimates drawn from generalized linear models with log link function and gamma distribution to ensure that the average median effects reported in the current study are concordant (in both direction and magnitude) with average mean effects.21 Variations in the cost of preventive drugs then were represented graphically in a series of contour graphs plotting the average cost by patient age at the time of death and number of comorbidities. Two sets of sensitivity analyses were performed to mitigate the risk of bias due to the potentially unpredictable time of death of older adults with cancer, which would explain why preventive drugs were continued until the very end of life. We first excluded patients whose underlying cause of death suggested an acute and sudden fatal event (see Supporting Table 3); we then stratified the main analyses according to the time between the cancer diagnosis and death, separating decedents who were diagnosed >12 months before death from those who were diagnosed during the last 6 months of life. Individuals with missing data regarding the time between diagnosis and death (7863 patients; 5.2%) were excluded from this sensitivity analysis. Statistical analyses were performed using JMP (version 13.0; SAS Institute Inc, Cary, North Carolina) and Stata (version 14.1; StataCorp LLC, College Station, Texas) statistical software. This study adheres to the RECORD (REporting of studies Conducted using Observational Routinely-collected health Data) guidelines (see Supporting materials).22


Characteristics of the Study Population

Among a total of 165,821 older adults who died with cancer in Sweden between 2007 and 2013, 151,201 (91.2%) met the eligibility criteria for the current study (Fig. 1). The mean age of the patients at the time of death was 81.3 years (standard deviation, 8.1 years) and approximately 45% of decedents were women, 18% lived in nursing home facilities, and 47% died in hospitals. As shown in Table 1, the most common cancer types were of the male genital organs (17%), respiratory organs (12%), and colorectum (11%). A large majority of patients had been diagnosed with cancer >12 months (60%) or between 6 to 12 months (12%) before death. Hypertension, ischemic heart disease, heart failure, atrial fibrillation, and type 2 diabetes were the most commonly diagnosed comorbidities. Older adults who died without cancer reported as their cause of death on their death certificate (29,984 patients; 19.8%) were on average older, lived more often in nursing homes, and had a greater number of chronic comorbidities than those who died of cancer (see Supporting Table 4).

Details are in the caption following the image
Study population flowchart.
Table 1. Characteristics of Older Adults Who Died With Solid Cancers in Sweden, 2007 Through 2013 (N = 151,201)
Sex, no. (%)
Men 83,429 (55.2)
Women 67,772 (44.8)
Age at time of death, y
Mean (SD) 81.3 (8.1)
65-74 35,690 (23.6)
75-84 56,950 (37.7)
85-94 52,474 (34.7)
≥95 6087 (4.0)
Level of education, no. (%)
Primary 71,661 (48.9)
Secondary 57,937 (39.5)
Tertiary 17,030 (11.6)
Living arrangement, no. (%)
Community 123,702 (81.8)
Nursing home 27,499 (18.2)
Place of death, no. (%)
Usual place of living 80,439 (53.2)
Hospital facility 70,762 (46.8)
Primary malignancy, no. (%)
Respiratory organs 18,435 (12.2)
Esophagus and stomach 5014 (3.3)
Colorectum 16,102 (10.6)
Liver and intrahepatic bile duct 3711 (2.5)
Pancreas 7808 (5.2)
Other digestive organs 3643 (2.4)
Breast 9920 (6.6)
Urinary tract 10,231 (6.8)
Male genital organs 25,642 (17.0)
Female genital organs 6868 (4.5)
Melanoma of skin 2651 (1.8)
Brain and meninges 2266 (1.5)
Unknown primary site 4030 (2.7)
Other primary malignancy 16,502 (10.9)
Multiple solid tumors 18,378 (12.2)
Time between diagnosis and death, no. (%)
>12 mo 86,032 (60.0)
6-12 mo 16,440 (11.5)
<6 mo 40,866 (28.5)
No. of chronic comorbidities (%)
Mean (SD) 4.5 (2.8)
0 6216 (4.1%)
1 14,242 (9.4%)
2 19,570 (12.9%)
3 22,039 (14.6%)
4 21,529 (14.2%)
≥5 67,605 (44.7%)
Main chronic comorbidities, no. (%)
Hypertension 66,553 (44.0%)
Ischemic heart disease 50,896 (33.7%)
Heart failure 42,049 (27.8%)
Atrial fibrillation 36,584 (24.2%)
Diabetes 31,279 (20.7%)
Cerebrovascular disease 28,730 (19.0%)
Cataract and other lens diseases 24,388 (16.1%)
COPD, emphysema, chronic bronchitis 22,465 (14.9%)
Dementia 17,784 (11.8%)
  • Abbreviations: COPD, chronic obstructive pulmonary disease; SD, standard deviation.
  • Missing values were for education (4573 patients; 3%) and time from diagnosis to death (7863 patients; 5.2%).

Use of Preventive Drugs

Throughout the last year of life, the mean number of prescribed drugs increased from 6.9 to 10.1 (mean difference, 2.1; 95% CI, 2.0-2.2) and the percentage of individuals using ≥10 drugs rose from 26% to 52%. Preventive drugs frequently were prescribed near the end of life (Table 2). Antihypertensives were prescribed to 60.1% of the decedents during their last month of life, including β-blockers (38.2%), angiotensin-converting enzyme inhibitors (18.5%), and calcium channel blockers (15.9%). Antithrombotic agents, antianemics, lipid-lowering drugs, mineral supplements, and drugs for diabetes also were commonly prescribed. We observed little change in the use of preventive drugs over the course of the last year before death. The percentage of older adults who continued therapy until the final month of life ranged from 56.6% for bisphosphonates to 65% for statins and vitamins, and up to ≥80% for insulin, β-blockers, and vitamin B12 or folic acid. Overall, 28.2% of decedents initiated antithrombotic agents (including 13.4% who initiated platelet aggregation inhibitors) during their last year of life, approximately 23.2% initiated medication for high blood pressure (including 13.3% who initiated β-blockers), and 4.9% initiated treatment with statins. Differences in the use of preventive drugs across cancer types are reported in Supporting Table 5. In sensitivity analyses, the results remained very similar after excluding individuals who died of acute and possibly unpredictable causes of death (see Supporting Table 6), or while comparing patients who had been diagnosed with cancer >12 months before death with individuals who were diagnosed at a time closer to death (see Supporting Table 7).

Table 2. Use of Preventive Drugs During the Last Year of Life Among Older Adults (Those Aged ≥65 Years) With Solid Cancers in Sweden, 2007 Through 2013
Prevalence (N = 151,201) Continuationb Until the Final Month of Life Initiationc During the Last Year of Life
12th Month Before Death Last Month Before Death Absolute Change
Percent Percent Percentage points (95% CI)a Percent (95% CI) Percent (95% CI)
Drugs used in diabetes 14.0% 14.9% +0.9 (0.6 to 1.2) 87.3 (86.8 to 87.7) 3.6 (3.5 to 3.7)
Insulin and analogues 7.6% 10.0% +2.4 (2.2 to 2.6) 89.3 (88.8 to 89.9) 4.0 (3.9 to 4.1)
Blood glucose-lowering drugs 8.7% 7.1% -1.6 (-1.8 to -1.4) 68.2 (67.4 to 69.0) 1.8 (1.7 to 1.9)
Vitamins 8.2% 9.2% +1.0 (0.8 to 1.2) 64.9 (64.1 to 65.7) 6.7 (6.6 to 6.8)
Mineral supplements 14.7% 19.2% +4.5 (4.2 to 4.8) 68.4 (67.7 to 69.9) 14.2 (14.0 to 14.4)
Calcium 10.5% 11.1% +0.6 (0.4 to 0.8) 65.7 (64.9 to 66.4) 6.5 (6.4 to 6.7)
Potassium 4.6% 7.8% +3.2 (3.0 to 3.4) 64.5 (63.3 to 65.6) 6.8 (6.6 to 6.9)
Antithrombotic agents 46.6% 48.1% +1.5 (1.1 to 1.9) 79.2 (78.9 to 79.5) 28.2 (27.9 to 28.5)
Vitamin K antagonists 7.7% 5.6% -2.1 (-2.3 to -1.9) 47.6 (46.7 to 48.5) 3.8 (3.7 to 3.9)
Heparin group 2.7% 10.0% +7.3 (7.1 to 7.5) 49.3 (47.8 to 51.9) 14.9 (14.6 to 15.9)
Platelet aggregation inhibitors 37.7% 36.2% -1.5 (-1.8 to -1.2) 77.4 (77.1 to 77.8) 13.4 (13.2 to 13.6)
Drugs used in the treatment of hypertension 60.4% 60.1% -0.3 (-0.6 to 0.0) 86.4 (86.2 to 86.7) 23.2 (22.9 to 23.6)
Low-ceiling diuretics 6.3% 5.2% -1.1 (-1.3 to -0.9) 61.2 (60.2 to 62.1) 1.9 (1.8 to 1.9)
Potassium-sparing agents 7.3% 11.2% +3.9 (3.7 to 4.1) 69.0 (68.1 to 69.9) 7.6 (7.5 to 7.8)
β-blocking agents 37.5% 38.2% +0.7 (0.4 to 1.0) 82.9 (82.6 to 83.3) 13.3 (13.1 to 13.6)
Calcium channel blockersd 18.9% 15.9% -3.0 (-3.3 to -2.7) 68.8 (68.2 to 69.3) 4.9 (4.7 to 5.7)
ACE inhibitors 20.3% 18.5% -1.8 (-2.1 to -1.5) 71.8 (71.3 to 72.3) 6.6 (6.4 to 6.7)
Angiotensin II antagonists 11.7% 9.9% -1.8 (-2.0 to -1.6) 71.3 (70.6 to 71.9) 2.4 (2.3 to 2.4)
Lipid modifying agents 21.5% 16.8% -4.7 (-5.0 to -4.4) 65.0 (64.4 to 65.5) 5.4 (5.3 to 5.5)
HMG CoA reductase inhibitors 21.0% 16.3% -4.7 (-5.0 to -4.4) 64.9 (64.4 to 65.4) 4.9 (4.7 to 5.6)
Bisphosphonates 4.2% 3.9% -0.3 (-0.4 to -0.2) 56.6 (55.3 to 57.8) 2.8 (2.7 to 2.9)
Antianemic preparations 25.7% 30.4% +4.7 (4.4 to 5.0) 79.7 (79.3 to 82.1) 17.6 (17.4 to 17.8)
Iron preparations 7.4% 11.0% +3.6 (3.4 to 3.8) 55.8 (54.9 to 56.8) 11.1 (11.0 to 11.3)
Vitamin B12 and folic acid 21.0% 23.2% +2.2 (1.9 to 2.5) 82.4 (82.0 to 82.8) 8.9 (8.7 to 9.1)
  • Abbreviations: 95% CI, 95% confidence interval; ACE, angiotensin-converting enzyme; HMG CoA, β-hydroxy β-methylglutaryl-CoA.
  • a Difference in percentages.
  • b Percentage of older adults who received drugs during the last month before death among those exposed 12 months before death.
  • c Percentage of older adults who received drugs during the last year of life among those not exposed 12 months before death.
  • d Excluding selective calcium channel blockers with direct cardiac effects (Anatomical Therapeutic Chemical [ATC] code C08D).

Drug Costs During the Last Year of Life

The median drug cost during the last year of life was $1482 (interquartile range [IQR], $700-$2896]) per person, ranging from $961 among decedents with cancers of an unknown primary site to $1811 among women with breast cancer and up to $3073 among men with cancers of the male genital organs (Table 3). After adjusting for multiple confounders, significantly higher costs were found for patients with breast cancer, gynecological cancers, cancers of the male genital organs, and multiple solid tumors compared with individuals who died with lung cancer. The median monthly drug costs increased from $80 to $153 over the course of the last year of life, although there was significant variation according to the type of cancer (see Supporting Table 8).

Table 3. Drug Costs During the Final Year of Life by Cancer Type
No. of Decedents Total Costs for Prescription Drugs Per Capita, US$a Costs for Preventive Drugs Per Capita, US$b Percentage of Total Drug Costs Dedicated to Preventive Agents, %
Median (IQR) β (95% CI)c Median (IQR) β (95% CI)c Total Last Year of Life 12th Month Last Month
Respiratory organs 18,435 1371 (662-2619) Reference 205 (61-523) Reference 23.6% 24.6% 21.8%
Esophagus and stomach 5014 1145 (552-2267) -122 (-178 to -65) 199 (68-479) 6 (-4 to 16) 22.9% 28.1% 15.2%
Colorectum 16,102 1074 (538-2107) -161 (-199 to -122) 209 (72-479) 11 (4 to 18) 26.5% 28.3% 21.1%
Liver and intrahepatic bile duct 3711 1079 (505-2117) -224 (-288 to -161) 222 (82-514) 19 (7 to 31) 23.7% 23.9% 23.6%
Pancreas 7808 1263 (627-2353) -47 (-94 to 1) 213 (69-520) 13 (5 to 22) 23.0% 24.8% 20.6%
Other digestive organs 3643 1041 (500-2110) -162 (-227 to -98) 191 (65-426) -7 (-19 to 5) 12.8% 12.5% 13.0%
Breast 9920 1811 (851-3410) 528 (482 to 575) 218 (81-528) 19 (11 to 28) 26.3% 26.5% 24.1%
Urinary tract 10,231 1221 (626-2274) -113 (-158 to -69) 232 (93-508) 11 (3 to 19) 25.1% 26.1% 21.1%
Male genital organs 25,642 3073 (1593-4559) 1826 (1790 to 1863) 209 (80-450) 13 (6 to 19) 13.3% 13.2% 12.7%
Female genital organs 6868 1350 (675-2568) 39 (-12 to 91) 239 (86-573) 27 (18 to 36) 26.5% 26.2% 23.3%
Melanoma of skin 2651 1015 (520-1944) -165 (-239 to -91) 200 (68-458) 12 (-2 to 25) 25.6% 27.1% 22.8%
Brain and meninges 2266 1216 (640-2190) -149 (-227 to -70) 205 (63-572) 3 (-11 to 17) 27.7% 28.6% 24.1%
Unknown primary site 4030 961 (475-1816) -224 (-286 to -162) 203 (82-431) 12 (0 to 23) 22.6% 22.2% 23.5%
Other primary malignancy 16,502 1185 (627-2234) -81 (-120 to -42) 221 (93-444) 4 (-3 to 12) 19.8% 20.4% 19.1%
Multiple solid tumors 18,378 1746 (796-3409) 342 (305 to 379) 219 (74-545) 13 (7 to 20) 18.9% 18.8% 16.9%
Total cohort 151,201 1482 (700-2986) 213 (77-490) 20.2% 20.5% 18.5%
  • Abbreviations: 95% CI, 95% confidence interval; IQR, interquartile range; US$, US dollars.
  • a Expenditures for all prescription drugs dispensed in community pharmacies (Anatomical Therapeutic Chemical [ATC] codes A to S).
  • b Expenditures for the prescription drugs mentioned in Table 2 (ATC codes available in Supporting Table 2).
  • c The quantile regression model was adjusted for sex, age at the time of death, number of chronic diseases, living arrangement, and educational level (4573 patients had missing values). β coefficients can be interpreted as the adjusted median difference in costs compared with decedents with cancer of the respiratory organs.

The median cost for preventive drugs during the last year of life amounted to $213 (IQR, $77-$490) in the total study population and varied across cancer types. Compared with older adults who died with lung cancer (median cost, $205; IQR, $61-$523), those who died with pancreatic cancer (AMD, $19; 95% CI, $7-$31), breast cancer (AMD, $19; 95% CI, $11-$28), and gynecological cancers (AMD, $27; 95% CI, $18-$36) had the highest costs per person. Throughout the last year of life, the percentage of total drug costs corresponding to preventive drugs was 20.2%; this percentage decreased from 20.5% during the 12 months before death to 18.5% during the last month before death. However, despite this relative reduction, we found an absolute increase in the costs due to preventive drugs (see Supporting Table 9). Overall, costs were found to be highest among older adults aged <80 years and among those who had ≥5 chronic comorbidities, although the data from the current study indicated that women with breast cancer had significantly higher costs for preventive drugs, even with a low burden of chronic multimorbidity (see Supporting Fig. 2). In sensitivity analyses, we found only marginal differences according to the time between diagnosis and death (see Supporting Table 10).


The current large, nationwide study demonstrated 3 main findings. First, a substantial number of older adults who die with solid cancers continue to receive preventive drugs until their final month of life. Second, preventive drugs account for approximately 20% of the total costs of prescribed drugs, and this percentage decreases only slightly as death approaches. Third, there are important differences between cancer types with regard to the use and costs of preventive drugs, which can be explained only in part by age and chronic multimorbidity.

The current study builds on previous work exploring the use of preventive drugs in terminally ill patients.23, 24 In Australia, Currow et al demonstrated that patients were prescribed on average 2.6 drugs for managing comorbid conditions at the time of referral to palliative care.25 Many patients who receive preventive cardiovascular drugs continue to do so until the very end of life.26, 27 For example, the prescribing of antihypertensive agents and platelet aggregation inhibitors is commonplace among hospice patients with advanced cancer.28 Recent studies also have shown that polypharmacy increases near the end of life, which is fueled not only by symptomatic drugs but also by the continuation of preventive agents until the very last weeks of life.18, 24

The frequent continuation of long-term preventive drugs is indicative of insufficient deprescribing strategies at the end of life. Although the preventive drugs reported in the current study are most often pharmacologically and clinically appropriate in the general population, their use within the context of a limited life expectancy and the palliative goals of care should be examined critically.29, 30 Preventive medicines are not necessarily inappropriate at the end of life, because some may have palliative indications to avert distressing symptoms or to avoid serious complications (eg, anticoagulants for managing cancer-related venous thrombosis). However, the large percentage of older adults with cancer who continue to receive statins, antihypertensives, and vitamins and mineral supplements throughout the last year of life does suggest the existence of routine-based prescribing practices that contribute to low-value care. The current study finding that older adults with cancers with a poor prognosis (eg, those of the brain, lung, liver, or pancreas) were just as likely as those with less aggressive disease to use preventive drugs during their last month of life suggests that there is room for deprescribing.

The question of whether drug treatments should be initiated or continued near the end of life is at the center of the “Choosing Wisely” campaign, which has been endorsed by the American Society of Clinical Oncology, the American Geriatrics Society, and the American Medical Directors Association. For example, it is explicitly recommended to refrain from administering lipid-lowering agents to older patients with a limited life expectancy. Evidence from a recent randomized controlled trial has demonstrated that discontinuing statins in this population is safe and can result in an improved quality of life.12 Three components appear to be essential to reducing the burden of preventive drugs of limited benefit. First, timely physician-patient communication is needed to evaluate whether the prescribed treatments are concordant with the patient’s goals of care. Second, physicians should carefully consider whether the prescribed drugs are likely to achieve their benefit within the patient’s remaining lifetime. Third, the decision to initiate, continue, or discontinue preventive treatments should account for the risk of the patient coming to harm.

From a health economics perspective, it can be argued that drugs account for only a small share of the total health care expenditure, with hospital and long-term care being the major sources of medical spending at the end of life. In the United States, costs related to drugs (including drugs administered during hospital stays) amount to approximately 4% of the entire medical expenditure during the last year of life.31 However, at the patient level, these costs are substantial and may contribute to the “financial toxicity” of treatments, especially in countries with no universal health care insurance coverage.32 It is worth noting that drug prices generally are much lower in Europe compared with the United States, due for the most part to strong price regulation within the European Union. In 2017, pharmaceutical expenditures amounted to $1162 per capita in the United States compared with $479 in Sweden.33 Moreover, indirect costs (eg, the cost of international normalized ratio testing associated with the use of warfarin) and induced costs (eg, hospital expenditures caused by severe adverse drug reactions) of drug prescribing also contribute to the overall burden of drug costs.

To the best of our knowledge, the current study is the first nationwide study to date to explore drug use within the last year of life according to cancer type, and to investigate the costs associated with these drugs. However, we acknowledge several limitations. First, it is possible that a percentage of the patients included in the cohort died of sudden and totally unexpected causes, which could explain why preventive drugs were continued until the time of death. Retrospective cohorts of decedents are indeed prone to confounding-by-indication bias and there is a tendency to underestimate the prognostic uncertainty surrounding end-of-life decisions.34 However, sensitivity analyses were performed in an attempt to separate sudden from nonsudden deaths, and demonstrated only marginal differences regarding patterns of drug use at the end of life. Second, routinely collected data regarding drug dispensing do not allow for an assessment of whether drugs actually are consumed by patients, and do not provide information concerning dose modifications that may occur between 2 refills. It is possible that some drugs were tapered off near the end of life, which the current study data would not reflect. Moreover, the estimations of drug costs relied on the assumption that patients used their treatments according to the prescribed daily dose. Although this assumption is unlikely at the individual level, it is reasonable to assume that, at a population level, variations from one patient to another would cancel each other out. In addition, because drugs administered during hospitalizations are not collected in the Swedish Prescribed Drug Register, the costs attributable to cancer-directed therapy are largely underestimated. Third, although the current study relied on routinely collected health care and administrative data with nationwide coverage in Sweden, the generalizability of its findings may be limited to countries with universal health care coverage and wide access to preventive drugs. Finally, we did not assess the appropriateness of prescribing; some preventive drugs reported in the current study may in specific cases and for specific indications have a meaningful clinical value. For example, the frequent use of bisphosphonates among women with breast cancer could stem from an effort to prevent and control bone metastases.


The use of preventive drugs in the last year of life is common among older adults with cancer, although there is considerable variation in use according to cancer type. In this context, the use of preventive drugs should be reconsidered in light of the patient’s goals of care, values, and preferences. Reducing the therapeutic burden in individuals with advanced cancer has the potential to not only reduce unnecessary adverse effects and improve patient quality of life, but also to reduce the financial burden for patients.

Funding Support

Supported by a grant from the Swedish Research Council for Health, Working Life and Welfare and by the Swedish Research Council. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the article.

Conflict of Interest Disclosures

The authors made no disclosures.

Author Contributions

Lucas Morin conceived and designed the study, performed the statistical analysis, interpreted the data, and drafted and critically revised the article. Adam Todd and Stephen Barclay conceived the study, interpreted the data, and critically revised the article. Jonas W. Wastesson interpreted the data and critically revised the article. Johan Fastbom developed the analytical approach for cost evaluation, interpreted the data, and critically revised the article. Kristina Johnell obtained funding, provided supervision, interpreted the data, and critically revised the article. All authors gave approval for the final version of the article and agreed to be accountable for all aspects of the work.