Current safeguards, risk mitigation, and transparency measures of large language models against the generation of health disinformation: repeated cross sectional analysis.

OBJECTIVES: To evaluate the effectiveness of safeguards to prevent large language models (LLMs) from being misused to generate health disinformation, and to evaluate the transparency of artificial intelligence (AI) developers regarding their risk mitigation processes against observed vulnerabilities. DESIGN: Repeated cross sectional analysis. SETTING: Publicly accessible LLMs. METHODS: In a repeated cross sectional analysis, four LLMs (via chatbots/assistant interfaces) were evaluated: OpenAI's GPT-4 (via ChatGPT and Microsoft's Copilot), Google's PaLM 2 and newly released Gemini Pro (via Bard), Anthropic's Claude 2 (via Poe), and Meta's Llama 2 (via HuggingChat). In September 2023, these LLMs were prompted to generate health disinformation on two topics: sunscreen as a cause of skin cancer and the alkaline diet as a cancer cure. Jailbreaking techniques (ie, attempts to bypass safeguards) were evaluated if required. For LLMs with observed safeguarding vulnerabilities, the processes for reporting outputs of concern were audited. 12 weeks after initial investigations, the disinformation generation capabilities of the LLMs were re-evaluated to assess any subsequent improvements in safeguards. MAIN OUTCOME MEASURES: The main outcome measures were whether safeguards prevented the generation of health disinformation, and the transparency of risk mitigation processes against health disinformation. RESULTS: Claude 2 (via Poe) declined 130 prompts submitted across the two study timepoints requesting the generation of content claiming that sunscreen causes skin cancer or that the alkaline diet is a cure for cancer, even with jailbreaking attempts. GPT-4 (via Copilot) initially refused to generate health disinformation, even with jailbreaking attempts-although this was not the case at 12 weeks. In contrast, GPT-4 (via ChatGPT), PaLM 2/Gemini Pro (via Bard), and Llama 2 (via HuggingChat) consistently generated health disinformation blogs. In September 2023 evaluations, these LLMs facilitated the generation of 113 unique cancer disinformation blogs, totalling more than 40 000 words, without requiring jailbreaking attempts. The refusal rate across the evaluation timepoints for these LLMs was only 5% (7 of 150), and as prompted the LLM generated blogs incorporated attention grabbing titles, authentic looking (fake or fictional) references, fabricated testimonials from patients and clinicians, and they targeted diverse demographic groups. Although each LLM evaluated had mechanisms to report observed outputs of concern, the developers did not respond when observations of vulnerabilities were reported. CONCLUSIONS: This study found that although effective safeguards are feasible to prevent LLMs from being misused to generate health disinformation, they were inconsistently implemented. Furthermore, effective processes for reporting safeguard problems were lacking. Enhanced regulation, transparency, and routine auditing are required to help prevent LLMs from contributing to the mass generation of health disinformation.

Variations in colorectal cancer pattern of care by age and comorbidity in South Australia.

BACKGROUND: Advanced age is associated with decreased likelihood of colorectal cancer treatment. Here, we investigated the extent to which comorbidities are accountable for this lesser treatment. METHODS: Using population-based datasets, the pattern of care among CRC cases in South Australia during 2004-2013 was investigated. Models were used to investigate associations of age with each treatment type, and differences in these associations were explored by comorbidity and cancer site. RESULTS: The presence of comorbidity was associated with a significantly weaker relationship of age with surgery and chemotherapy. The association of age with surgery also varied for colon and rectal primary cancer sites. Individual comorbidity types varied in their associations with each treatment category. For example, dementia was associated with less chemotherapy provision, however, it was not significantly related to the likelihood of surgery. CONCLUSION: This study indicates that the association of age with surgical treatment differed significantly by the CRC subsite. Comorbidity moderated the negative association of age with chemotherapy, and less so, with extent of surgery. Results were novel in indicating associations of multiple individual comorbidity types with CRC treatment modalities. The data suggest that different individual comorbidity types may have different effects on treatment and should be studied separately.

Associations of advanced age with comorbidity, stage and primary subsite as contributors to mortality from colorectal cancer.

Background: Although survival from colorectal cancer (CRC) has improved substantially in recent decades, people with advanced age still have a high likelihood of mortality from this disease. Nonetheless, few studies have investigated how cancer stage, subsite and comorbidities contribute collectively to poor prognosis of older people with CRC. Here, we decided to explore the association of age with mortality measures and how other variables influenced this association. Methods: Using linkage of several administrative datasets, we investigated the risk of death among CRC cases during 2003-2014. Different models were used to explore the association of age with mortality measures and how other variables influenced this association. Results: Our results indicated that people diagnosed at a young age and with lower comorbidity had a lower likelihood of all-cause and CRC-specific mortality. Aging had a greater association with mortality in early-stage CRC, and in rectal cancer, compared that seen with advanced-stage CRC and right colon cancer, respectively. Meanwhile, people with different levels of comorbidity were not significantly different in terms of their increased likelihood of mortality with advanced age. We also found that while most comorbidities were associated with all-cause mortality, only dementia [SHR = 1.43 (1.24-1.64)], Peptic ulcer disease [SHR = 1.12 (1.02-1.24)], kidney disease [SHR = 1.11 (1.04-1.20)] and liver disease [SHR = 1.65 (1.38-1.98)] were risk factors for CRC-specific mortality. Conclusion: This study showed that the positive association of advanced age with mortality in CRC depended on stage and subsite of the disease. We also found only a limited number of comorbidities to be associated with CRC-specific mortality. These novel findings implicate the need for more attention on factors that cause poor prognosis in older people.

Occurrence of comorbidity with colorectal cancer and variations by age and stage at diagnosis.

BACKGROUND: While age and stage at diagnosis are known to affect treatment choices and survival from colorectal cancer (CRC), few studies have investigated the extent to which these effects are influenced by comorbidity. In this study, we describe the occurrence of comorbidity in CRC cases in South Australia and associations of comorbidity with age, stage and the age-stage relationship. Furthermore, we report on the association of individual comorbidities with age and stage at diagnosis. METHODS: The South Australian Cancer Registry (SACR) provided CRC data (C18-C20, ICD-10) for 2004-2013 diagnoses. CRC data were linked with comorbidity data drawn from hospital records and health insurance claims. Logistic regression was used to model associations of comorbidity with age and stage. RESULTS: For the 8462 CRC cases in this study, diabetes, peptic ulcer disease, and previous cancers were the most commonly recorded co-existing conditions. Most comorbidities were associated with older age, although some presented more frequently in younger people. Patients at both ends of the age spectrum (<50 and 80 + years) had an increased likelihood of CRC diagnosis at an advanced stage compared with other ages (50-79 years old). Adjusting for comorbidities moderated the association of older age with advanced stage. Conditions associated with advanced stage included dementia (OR = 1.25 (1.01-1.55)), severe liver disease (OR = 1.68 (1.04-2.70)), and a previous cancer (OR = 1.18 (1.08-1.28)). CONCLUSION: Comorbidities are prevalent with CRC, especially in older people. These comorbidities differ in their associations with age at diagnosis and stage. Dementia and chronic heart failure were associated with older age whereas inflammatory bowel disease and alcohol access were associated with younger onset of the disease. Severe liver disease and dementia were associated with more advanced stage and rheumatic disease with less advanced stage. Comorbidities also interact with age at diagnosis and appear to vary the likelihood of advanced-stage disease. CRC patient have different association of age with stage depending on their comorbidity status.

'Food for Thought'-The Relationship between Diet and Cognition in Breast and Colorectal Cancer Survivors: A Feasibility Study.

Survivors of cancer frequently experience persistent and troublesome cognitive changes. Little is known about the role diet and nutrition plays in survivors' cognition. We explored the feasibility of collecting cross-sectional online data from Australian survivors of breast and colorectal cancer to enable preliminary investigations of the relationships between cognition with fruit and vegetable intake, and the Omega-3 Index (a biomarker of long chain omega 3 fatty acid intake). A total of 76 participants completed online (and postal Omega-3 Index biomarker) data collection (62 breast and 14 colorectal cancer survivors): mean age 57.5 (±10.2) years, mean time since diagnosis 32.6 (±15.6) months. Almost all of the feasibility outcomes were met; however, technical difficulties were reported for online cognitive testing. In hierarchical linear regression models, none of the dietary variables of interest were significant predictors of self-reported or objective cognition. Age, BMI, and length of treatment predicted some of the cognitive outcomes. We demonstrated a viable online/postal data collection method, with participants reporting positive levels of engagement and satisfaction. Fruit, vegetable, and omega-3 intake were not significant predictors of cognition in this sample, however the role of BMI in survivors' cognitive functioning should be further investigated. Future research could adapt this protocol to longitudinally monitor diet and cognition to assess the impact of diet on subsequent cognitive function, and whether cognitive changes impact dietary habits in survivors of cancer.

Patterns of health service utilisation among the Australian population with cancer compared with the general population.

Objective The aim of this study was to describe patterns of health service utilisation among the Australian population with cancer compared with the general population. Methods Data for all respondents aged ≥25 years from two successive National Health Surveys conducted between 2011 and 2014 were analysed. Respondents with a history of cancer were identified as the cancer group, whereas all other respondents who did not report having had a cancer were included in the non-cancer control group. Comparisons were made between the two groups using logistic regression models. Results The population with cancer was more likely to report having consulted their general practitioner, specialist, chemist, dietician, naturopath, nurse, optometrist, dentist, audiologist and other health professionals than the non-cancer population. The cancer population was also more likely to be admitted to hospital and to have visited an out-patient clinic, emergency department and day clinic. The presence of comorbidity and a current cancer were associated with a greater likelihood of receiving health services among the population with cancer. Conclusion The population with cancer used health services significantly more than the non-cancer population. Further studies are urgently needed to identify optimal approaches to delivery of care for this population, including barriers and enablers for their implementation. What is known about the topic? Multimorbidity is highly prevalent among the cancer population due to risk factors shared between cancer and other chronic diseases, and the development of new conditions resulting from cancer treatment and cancer complications. However, the Australian healthcare system is not set up optimally to address issues related to multimorbidity. What does this paper add? This study is the first step in quantifying health services use by the population with cancer compared with the general population without cancer. Cancer survivors have an increased need for specific health services, particularly among those with multimorbidity. What are the implications for practitioners? The development of integrated care models to manage multiple chronic diseases aligned with the Australian National Strategic Framework for Chronic Conditions is warranted. Further studies are urgently needed to identify optimal approaches to delivery of care for this population, including barriers and enablers for their implementation.

Patterns of comorbidities in women with breast cancer: a Canadian population-based study.

PURPOSE: Improving the understanding of co-existing chronic diseases prior to and after the diagnosis of cancer may help to facilitate therapeutic decision making in clinical practice. This study aims to examine patterns of comorbidities in Canadian women with breast cancer. METHODS: We conducted a retrospective cohort study using provincial linked administrative health datasets from British Columbia, Canada, between 2000 and 2013. Women diagnosed with breast cancer between 2005 and 2009 were identified. The index date was defined as the date of diagnosis of breast cancer. Subsets of the breast cancer cohort were identified based on the absence of individual type of comorbidity of interest within 5 years prior to breast cancer diagnosis. For each subset, cases were then individually matched by year of birth at 1:2 ratios with controls without a history of cancer and the individual type of comorbidity of interest within 5 years prior to the assigned index year, matching with the year of breast cancer diagnosis of the corresponding case. Baseline comorbidities were measured over a 1-year period prior to the index date using two comorbidity indices, Rx-Risk-V and Aggregated Diagnosis Groups (ADG). Cox regression model was used to assess the development of seven specific comorbidities after the index date between women with breast cancer and non-cancer women. RESULTS: The most prevalent baseline comorbidity in the breast cancer cohort measured using the Rx-Risk-V model was cardiovascular conditions (39.0%), followed by pain/pain-inflammation (34.8%). The most prevalent category measured using the ADG model was major signs or symptoms (71.8%), followed by stable chronic medical conditions (52.2%). The risks of developing ischemic heart disease, heart failure, depression, diabetes, osteoporosis, and hypothyroidism were higher in women with breast cancer compared to women without cancer, with the hazard ratios ranging from 1.09 (95 CI% 1.03-1.16) for ischemic heart disease to 2.10 (95% CI 1.99-2.21) for osteoporosis in the model adjusted for baseline comorbidity measured using Rx-Risk-V score. CONCLUSION: Women with breast cancer had a higher risk of developing new comorbidities than women without cancer. Development of coordinated care models to manage multiple chronic diseases among breast cancer patients is warranted.

Pharmaceutical Industry Funding of Health Consumer Groups in Australia: A Cross-sectional Analysis.

Relationships between consumer organizations and pharmaceutical manufacturers are the focus of transparency efforts in some jurisdictions, including Australia. This study describes the frequency and nature of industry sponsorship of Australian health consumer organizations and examines the link between sponsorship of the most highly funded organizations and manufacturers' requests for public reimbursement of products for related health conditions. We downloaded 130 transparency reports covering the period January 2013 to December 2016 from the website of Medicines Australia and carried out a descriptive analysis. For the most heavily funded organizations and their sponsors, we examined Public Summary Documents of the Pharmaceutical Benefits Advisory Committee to identify relevant products under consideration for public reimbursement over the study period. Thirty-four pharmaceutical companies provided 1,482 sponsorships to 230 organizations, spending a total of AU$34,507,810. The top clinical areas in terms of amount of funding received were cancer, eye health, and nervous system disorders. The sponsors of the most highly funded groups were companies that in most cases had drugs under review for public reimbursement for conditions covered by these organizations. Interactions between the pharmaceutical industry and consumer organizations are common and require careful management to prevent biases that may favor sponsors' interests above those of patients and the public.

Development of comorbidities in men with prostate cancer treated with androgen deprivation therapy: an Australian population-based cohort study.

BACKGROUND: The increasing use of androgen deprivation therapy has prompted further evaluation of its potential adverse effects as the treatment may exacerbate or increase the risk of developing new comorbid diseases. This study aims to assess the patterns of comorbidities among Australian men with prostate cancer treated with androgen deprivation therapy. METHODS: Pharmaceutical Benefits Scheme (PBS) 10% data between 1 January 2003 and 31 December 2014 was utilised in this retrospective cohort study. Men who had received their first androgen deprivation therapy between 2004 and 2010 were selected as the prostate cancer cohort. Comorbidities were identified using the dispensing claims data and classified with the Rx-Risk-V model. Comparisons were made between the prostate cancer cohort and specific control groups (age-matched and sex-matched without any dispensing of anti-neoplastic agents during the study period and without the individual comorbidity of interest evaluated at baseline at 1:10 ratio) for the development of nine individual comorbidities over time using Cox regression models. RESULTS: The prostate cancer cohort had a significant higher risk of developing cardiovascular conditions (hazard ratio 1.37, 95% CI: 1.26-1.48), depression (1.86, 95% CI: 1.73-2.01), diabetes (1.30, 95% CI: 1.15-1.47), gastric acid disorders (1.48, 95% CI: 1.39-1.57), hyperlipidaemia (1.18, 95% CI: 1.09-1.29), osteoporosis (1.65, 95% CI: 1.48-1.85) and pain/pain-inflammation (1.47, 95% CI: 1.39-1.55) compared to the control groups. The hazard ratios for cardiovascular conditions and depression were highest in the first year and declined over time. There were no significant differences between the two groups for reactive airway diseases and Alzheimer's disease. CONCLUSION: Men with prostate cancer treated with androgen deprivation therapy had a higher likelihood of developing new comorbidities than men who did not receive androgen deprivation therapy. Our results support the need for developing coordinated care models that effectively address multiple chronic diseases experienced by prostate cancer survivors.

Incidence of comorbidities in women with breast cancer treated with tamoxifen or an aromatase inhibitor: an Australian population-based cohort study.

BACKGROUND: The development of comorbidities has become increasingly relevant with longer-term cancer survival. OBJECTIVE: To assess the pattern of comorbidities among Australian women with breast cancer treated with tamoxifen or an aromatase inhibitor. DESIGN: Retrospective cohort study using Pharmaceutical Benefits Scheme (PBS) data (10% sample) from January 2003 to December 2014. Dispensing claims data were used to identify comorbidities and classified with the Rx-Risk-V model. The breast cancer cohort had tamoxifen or an aromatase inhibitor dispensed between 2004 and 2011 with no switching between types of endocrine therapy. Comparisons were made between the breast cancer cohort and specific control groups (age- and sex-matched at 1:10 ratio without any dispensing of anti-neoplastic agents during the study period) for the development of five individual comorbidities over time using Cox regression models. RESULTS: Women treated with tamoxifen had a higher incidence of cardiovascular conditions, diabetes, and pain or pain-inflammation, but a lower incidence of hyperlipidaemia compared with non-cancer control groups, as indicated by PBS data. Women treated with aromatase inhibitors were more likely to develop cardiovascular conditions, osteoporosis, and pain or pain-inflammation compared with non-cancer control groups. The risks of hyperlipidaemia and osteoporosis were significantly lower among tamoxifen users compared with aromatase inhibitor users. CONCLUSIONS: Women with hormone-dependent breast cancer treated with an endocrine therapy had a higher risk of developing specified comorbid conditions than women without cancer, with different comorbidity profiles for those on tamoxifen versus aromatase inhibitors. Further research into the causes and mechanism of development and management of comorbidities after cancer is needed.

Changes in the prevalence of comorbidity in the Australian population with cancer, 2007-2014.

BACKGROUND: Coexistence of multiple chronic diseases is highly prevalent among the cancer population. This study aims to assess changes in the prevalence of chronic conditions among the population with cancer compared to the Australian general population between 2007 and 2014. METHODS: Data from three successive National Health Surveys conducted by the Australian Bureau of Statistics between 2007 and 2014 were utilized. Comparisons were made between the samples of the Australian population aged 25 years and above with a history of cancer and those respondents who did not report having had a cancer using logistics regression models. RESULTS: People with a history of cancer had significantly higher odds of reporting non-infectious comorbidity compared to the non-cancer groups across the three surveys. There were no significant changes in the prevalence of diseases affecting circulatory, musculoskeletal, digestive, nervous system, blood and blood forming organs, eye, skin and infectious and parasitic diseases over time among the population with cancer. The prevalence of mental and behavioural problems, endocrine, nutritional and metabolic diseases, and diseases of respiratory and genitourinary system has increased over time among the cancer survivors. CONCLUSION: Comorbidity is more prevalent among the cancer population than the general population without cancer. The prevalence of comorbidity was fairly stable for most but not all comorbidities in the population with cancer over the eight-year study period. Further studies on the impacts of coordinated care models for the management of multi-morbidity experienced by cancer survivors that align with the 'National Strategic Framework for Chronic Conditions' are needed.

Medicines access programs to cancer medicines in Australia and New Zealand: An exploratory study.

Medicines Access Programs (MAP) offer access to publicly unfunded medicines at the discretion of pharmaceutical companies. Limited literature is available on their extent and scope in Australia and New Zealand. This study aims to identify MAPs for cancer medicines that were operational in 2014-15 in Australia and New Zealand and describe their characteristics. A preliminary list of MAPs was sent to hospital pharmacists in Australia and New Zealand to validate and collect further information. Pharmaceutical companies were contacted directly to provide information regarding MAPs offered. Key stakeholders were interviewed to identify issues with MAPs. Fifty-one MAPs were identified covering a range of indications. The majority of MAPs were provided free of charge to the patient for medicines that were registered or in the process of being registered but were not funded. Variability in the number of MAPs across institutions and characteristics was observed. Australia offered more MAPs than New Zealand. Only two of 17 pharmaceutical companies contacted agreed to provide information on their MAPs. Eight stakeholder interviews were conducted. This identified that while MAPs are widely operational there is lack of clinical monitoring, inequity to access, operational issues and lack of transparency. Our results suggest a need for a standardised and mandated policy to mitigate issues with MAPs.

Comorbidities in Australian women with hormone-dependent breast cancer: a population-based analysis.

OBJECTIVE: To compare how frequently selected chronic diseases developed in women with breast cancer receiving endocrine therapy, and in women without cancer. DESIGN, SETTING AND PARTICIPANTS: Retrospective, rolling cohort study, analysing a random 10% sample of Pharmaceutical Benefits Scheme (PBS) data for the period 1 January 2003 - 31 December 2014. Women with breast cancer who first commenced endocrine therapy between January 2004 and December 2011 were identified, and age- and sex-matched (1:10) by comorbidity with control groups of women who did not have a dispensing record for antineoplastic agents during the study period or the comorbidity of interest at baseline. MAIN OUTCOME MEASURES: Development of any of eight pre-selected comorbidities, identified in PBS claims data with the RxRisk-V model. RESULTS: Women with hormone-dependent breast cancer were significantly more likely than women in the control group to develop depression (overall hazard ratio [HR], 1.36; 95% CI, 1.26-1.46), pain or pain-inflammation (HR, 1.30; 95% CI, 1.23-1.38), osteoporosis (overall HR, 1.27; 95% CI, 1.17-1.39), diabetes (HR, 1.24; 95% CI, 1.10-1.41), cardiovascular disorders (overall HR, 1.22; 95% CI, 1.13-1.32), and gastric acid disorders (HR, 1.20; 95% CI, 1.13-1.28). The hazard ratios for developing cardiovascular disorders, depression and osteoporosis were highest during the first year of endocrine therapy. The risk of hyperlipidaemia was lower among women with breast cancer than in the control group (HR, 0.88; 95% CI, 0.81-0.96). There was no significant difference between the two groups in the risk of reactive airway diseases (HR, 1.05; 95% CI, 0.98-1.13). CONCLUSION: Comorbid conditions are more likely to develop in women who have been diagnosed with hormone-dependent breast cancer than in women without cancer. Our results further support the need to develop appropriate models of care to manage the multiple chronic disorders of breast cancer survivors.

Comorbidity, physical and mental health among cancer patients and survivors: An Australian population-based study.

AIM: To assess the prevalence of comorbidities and measures of physical and mental health among the cancer patients and survivors compared with the general population. METHODS: Data collected by the Australian Bureau of Statistics from 2011-2012 National Health Survey were utilized for this cross-sectional study. Comparisons were made between adults aged 25 years and over with history of cancer (n = 2170) and those respondents who did not report having had a cancer (n = 11 592) using logistic regression models. Analyses were repeated according to cancer status (current cancer vs. cancer survivor). RESULTS: People with history of cancer had significantly higher odds of reporting mental and behavioral problems (overall cancer group adjusted odds ratio 1.36, 95 percent confidence interval 1.20-1.54; current cancer 2.53, 1.97-3.27; cancer survivor 1.20, 1.05-1.38), circulatory conditions (overall cancer group 1.25, 1.12-1.39; current cancer 1.38, 1.08-1.76; cancer survivor 1.22, 1.09-1.38), musculoskeletal conditions (overall cancer group 1.37, 1.24-1.52; current cancer 1.66, 1.30-2.12; cancer survivor 1.33, 1.19-1.48) and endocrine system disorders (overall cancer group 1.19, 1.06-1.34; current cancer 1.29, 1.00-1.66; cancer survivor 1.17, 1.04-1.33) compared with the noncancer group. Cancer patients and survivors were more likely to report poor health status, a higher level of distress, and a greater number of chronic conditions compared with the noncancer group. CONCLUSION: Poor health and comorbidity is more prevalent among cancer patients and survivors than the noncancer population. Our results further support the need to develop models of care that effectively address multiple chronic conditions experienced by the cancer population.

Access to new cancer medicines in Australia: dispelling the myths and informing a public debate.

Despite the high level of spending on cancer medicines in Australia, consumer organisations and the pharmaceutical industry often make claims of delayed or lack of access to new cancer medicines-claims that are frequently supported by prominent coverage in the Australian media. These claims, while morally and psychologically compelling, tend to ignore the complexity of medicines funding decisions. In this commentary we summarise the current situation regarding the registration and funding of cancer medicines in Australia, elucidate the main challenges associated with access to cancer medicines in the Australian context, and describe some of the steps that have been taken to address these challenges.

Cancer drugs in 16 European countries, Australia, and New Zealand: a cross-country price comparison study.

BACKGROUND: Cancer drugs challenge health-care systems because of their high prices. No cross-country price comparison of cancer drugs for a large number of countries has been published. We aimed to survey the prices of cancer drugs in high-income countries (Europe, Australia, and New Zealand). METHODS: Based on comparability in terms of the economic situation and of the pharmaceutical system, we surveyed official list prices per unit at ex-factory price level of 31 originator cancer drugs in 16 European countries, Australia, and New Zealand as of June, 2013. Drug price data for the European countries were provided by the Pharma Price Information (PPI) service; Australian and New Zealand drug price data were retrieved from the respective pharmaceutical schedules. FINDINGS: In Austria, Denmark, Finland, Germany, Italy, Norway, Sweden, and the UK, price information was available for all or all but one drug surveyed whereas the availability of price data was restricted for some drugs in other countries, especially in New Zealand and Portugal. The difference of a drug price between the highest priced country and the lowest priced country varied between 28% and 388%. A few drugs had lower outliers, especially Greek and UK prices, and upper outliers (particularly prices in Switzerland, Germany, and Sweden). Overall, Greek prices ranked at a low level, whereas Sweden, Switzerland, and Germany showed price data in similarly high ranges. INTERPRETATION: Our results showed variations in ex-factory prices of originator cancer drugs in the 18 surveyed countries. However, the surveyed prices do not include discounts negotiated by funding organisations because these discounts are confidential. Because pricing authorities can also only use these official undiscounted prices when they set prices through the common policy of external price referencing, they risk overpaying. Our findings provide an evidence base for policy makers to decide whether further policy measures related to drug prices are needed. FUNDING: None.

Regulatory withdrawal of medicines marketed with uncertain benefits: the bevacizumab case study.

BACKGROUND: Withdrawal of conditional regulatory approval or subsidization of new medicines when subsequent evidence does not confirm early trial results may not be well understood or accepted by the public. OBJECTIVES: We present a case study of the US Food and Drug Administration (FDA)'s decision to withdraw the indication of bevacizumab for the treatment of advanced breast cancer and include an analysis of the reactions of stakeholders with a view to identifying opportunities for improving risk management for new medicines with conditional approval or funding. METHODS: We drew on a range of information sources, including FDA documents, medical journals and media reports, to describe the evidentiary basis of the FDA decisions. We analysed the reactions and perspectives of the stakeholders. RESULTS: In 2008 bevacizumab was granted conditional approval for treatment of advanced breast cancer by the FDA pending submission of supplementary satisfactory evidence. In 2011 the FDA decision to withdraw the indication was met with a hostile reaction from many clinicians and cancer survivors. There were different interpretations of the therapeutic value of bevacizumab with strong beliefs among cancer survivors that the medicine was effective and potential harm was manageable. High expectations of the public may have been encouraged by overly positive media reports and limited understanding by the public of the complexity of the scientific evaluation of new medicines and of the regulatory processes. CONCLUSIONS: Improving understanding and acceptance of approval or coverage schemes conditional to evidence development may require the development of risk management plans by regulatory and funding institutions. They may include a range of strategies such as requirements for formal patient acknowledgment of the conditional availability of the medicine, 'black-triangle' equivalent labels that identify full approval is based on pending evidence, and ongoing communication with the media, public and health professionals.

Charlson and Rx-Risk comorbidity indices were predictive of mortality in the Australian health care setting.

OBJECTIVE: To compare the performance of Charlson index and Rx-Risk score using data from Australian Department of Veterans' Affairs. STUDY DESIGN AND SETTING: A study of older adults (N=94,714) who had both Charlson and Rx-Risk scores based on their hospital diagnoses and prescription medication dispensings during the baseline year (January 2005-December 2005). Predictive ability of 1-year and 3-year mortality was compared by Akaike information criterion model fit statistic and c statistic in logistic regression models. We also compared the scores for identifying specific medical conditions. RESULTS: Both indices were significant predictors of all-cause mortality (P<0.0001). Of the population identified with a condition from either score, Rx-Risk score identified more than 95% of patients with gastric, respiratory, or cardiovascular condition, compared with Charlson index only identifying 2%, 31%, and 14%, respectively. The indices were comparable regarding diabetes. The Charlson index identified 83% of patients with dementia and 67% of those with cancers, whereas Rx-Risk score identified 38% and 43%, respectively. CONCLUSION: Both the Charlson and Rx-Risk scores predict mortality, but neither index identified all comorbidities. Based on data availability, preferences, and research purposes, investigators can use either Charlson index or Rx-Risk score to adjust for comorbidity.