In ORIENT-11, addition of sintilimab to first-line chemotherapy demonstrated improved progression-free survival (PFS), overall survival (OS), and objective response rate (ORR) in Chinese patients with AMnsqNSCLC. This SLR and NMA compared the efficacy and safety of sintilimab+pemetrexed+platinum versus FDA-approved and NCCN-recommended immune checkpoint inhibitor combinations (ICIs) for first-line treatment of AMnsqNSCLC. SLR followed the PRISMA guideline. Eligible studies were phase 2 and 3 RCTs published in English between 1991 and September 2021 comparing efficacy or safety of ICIs with other treatments for AMnsqNSCLC without EGFR/ALK mutations. Bayesian fixed and random effects NMA with independent baseline models were used, and chemotherapy was the common comparator. PFS, OS, ORR, and exposure-adjusted adverse events(AEs) were the key outcomes analyzed. Hazard ratios(HR)<1.0, odds ratios(OR)>1.0, and median exposure-adjusted rate difference(D)>0 favored the reference treatment, sintilimab+pemetrexed+platinum. After screening 6,418 records, 11 RCTs involving a total of 4,979 eligible patients regardless of PD-L1 status were included in the primary evidence network. ICIs with chemotherapy were generally superior to chemotherapy alone. For PFS, pembrolizumab+pemetrexed+platinum(HR 0.96 [95% Credible Interval: 0.71-1.30]) and atezolizumab+bevacizumab+platinum+nab-paclitaxel(0.83 [0.57-1.19]) were not significantly different to sintilimab+pemetrexed+platinum. Sintilimab+pemetrexed+platinum had significantly better PFS compared to atezolizumab+platinum+nab-paclitaxel(0.57[0.40-0.82]) and nivolumab+ipilimumab+pemetrexed+platinum(0.66[0.48-0.92]). Sintilimab+pemetrexed+platinum was not significantly different for OS: atezolizumab+bevacizumab+platinum+nab-paclitaxel(0.85[0.59-1.23]), atezolizumab+platinum+nab-paclitaxel(0.81[0.56-1.17]), pembrolizumab+pemetrexed+platinum(1.08[0.79-1.47]), nivolumab+ipilimumab+pemetrexed+platinum(0.95[0.67-1.34]), and nivolumab+ipilimumab(0.83[0.61-1.13]). Rate of Grade≥3 AEs(D range:-0.35-0.84) and ORR(OR range:0.73-1.91) were comparable across ICIs. This NMA showed that sintilimab+pemetrexed+platinum was associated with comparable efficacy and safety versus standard-of-care ICIs in the US, further supporting the favorable benefit/risk profile of ICI versus chemotherapy among patients with previously untreated AMnsqNSCLC and no EGFR or ALK mutations. The results may aid clinical decision-making and future cost-effectiveness evaluation of ICIs in the absence of head-to-head evidence from RCTs.

Health economic models of type 2 diabetes are increasingly being used to guide formulary decision making in the US. Models informing such decisions must be clinically credible and valid for the populations of interest. The aim of the present analysis was to evaluate the impact of using risk equations derived from different populations on modeled long-term outcomes for a US population with type 2 diabetes. Long-term projections of clinical outcomes were made for a US cohort with type 2 diabetes receiving two hypothetical interventions to reduce HbA1c and BMI (treatment A was superior to treatment B in terms of HbA1c and BMI improvements). In the first scenario, the risk of complications and mortality was modeled using equations from the BRAVO model (derived from a US population) and in a second scenario United Kingdom Prospective Diabetes Study Outcomes Model 2 (UKPDS OMS2, derived from a UK population) equations were used. All other simulation settings and parameter inputs were identical. Future clinical benefits were discounted at 3% annually. Differences in mortality were observed with incremental life expectancy of approximately 0.09 years using BRAVO equations (11.91 versus 11.82 years) compared with 0.03 years using UKPDS OM2 (12.56 versus 12.53 years). Incremental quality adjusted life expectancy estimates were 0.16 QALYs with BRAVO equations (7.52 versus 7.36 QALYs) versus 0.08 QALYs with UKPDS OM2 (7.36 versus 7.28 QALYs). Small differences between treatments were projected in the cumulative incidence of most diabetes-related complications, but overall complication rates were lower with UKPDS OM2 than with BRAVO equations. Different risk equations in an economic model of a US cohort with type 2 diabetes produces different long-term outcomes, which could have a notable influence on the cost-effectiveness of interventions.

Alzheimer’s disease (AD) dementia is a debilitating, progressive condition that often places caregiving responsibility on informal (unpaid) caregivers, causing a substantial burden to caregivers and society. Caregiver utility comprises an important part of this burden and should be considered when assessing interventions for AD dementia. This study aimed to quantify utility through directly eliciting utility values from the general public in the UK. Six health state vignettes were developed using published literature, online patient forums, and interviews with experts (N=5) including clinicians, caregivers, and caregiver advocates. Vignettes varied based on disease severity, caregiver-patient relationship and living situation. Health states were validated through further expert interviews (N=3) with clinicians and a caregiver. A pilot study was conducted with the general public (N=10) before finalization of the health states. Utility values (0 to 1) were elicited through time trade-off (TTO) interviews with members of the general public. Face-to-face interviews were included with 109 respondents for the analysis. Mean TTO scores were elicited for the caregiver living with a patient with AD with mild (0.79; 95% CI: 0.74 to 0.84), moderate (0.65; 95% CI: 0.60 to 0.71) or severe dementia (0.49; 95% CI: 0.44 to 0.55) and for caregivers living separately from a patient with AD with mild (0.74; 95% CI: 0.70 to 0.79) or moderate dementia (0.71; 95% CI: 0.66 to 0.76). The utility of a caregiver of a person with AD with severe dementia being cared for in a nursing home (0.64, 95% CI: 0.58 to 0.71) was comparable to the utility of a caregiver living with a patient with AD with moderate dementia (0.65, 95% CI: 0.60 to 0.71). This study quantifies the substantial burden of AD dementia caregiving across the AD dementia continuum. Overall, caregiver utility decreased with increasing severity of AD dementia.

Dementia is a debilitating, progressive condition that often places caregiving responsibility on unpaid caregivers, causing a substantial burden to caregivers. Caregiver utility comprises an important part of this burden and should be considered when assessing interventions for AD dementia. This study aimed to quantify the utility of AD with MCI and AD dementia caregivers in Japan. Seven health state vignettes were developed using published literature, online patient forums, and interviews with experts (N=4) including clinicians and caregivers. Vignettes varied based on disease severity, caregiver-patient relationship and living situation. Health states were validated through interviews with two clinical experts. A pilot study was conducted with the general public (N=10) before finalising the health states. Utility values (0 to 1) were elicited through time trade-off (TTO) interviews with members of the general public. Face-to-face interviews were included with 116 respondents for the analysis. The mean TTO score for caring for a patient with AD with MCI was 0.75 (95% CI: 0.71 to 0.79). TTO scores were elicited for the caregiver living with a patient with AD with mild (0.77; 95% CI: 0.73 to 0.81), moderate (0.51; 95% CI: 0.45 to 0.56) and severe dementia (0.32; 95% CI: 0.27 to 0.37), or living separately from a patient with AD with mild (0.67; 95% CI: 0.62 to 0.72), moderate (0.54; 95% CI: 0.49 to 0.59) and severe dementia when patient is cared for in a nursing home (0.76; 95% CI: 0.71 to 0.81). This study quantifies the substantial burden of AD dementia caregiving across the AD dementia continuum. Caregiver utility decreased as the severity of the disease increased, except when patients with AD with severe dementia were being cared for in a nursing home. Future research should explore further the differences between AD with MCI and AD with mild dementia.

To examine potential cost-effectiveness of Alzheimer’s disease-modifying therapies (DMTs) of different treatment durations. BACKGROUND: Clinical trial results suggest the disease-modifying impact of a new class of anti-amyloid monoclonal antibodies. Optimal treatment duration is uncertain due to limited follow-up in clinical trials. A recent economic analysis of the first such agent approved questioned the economic value of treatments that slow progression of early symptomatic Alzheimer’s disease (AD) to more advanced states. Results were highly sensitive to therapy costs, which would likely be substantially influenced by treatment duration. We constructed a population-based Markov state-transition simulation model to calculate the cost-effectiveness of hypothetical DMTs. Base-case assumptions were similar to those in the Institute for Clinical and Economic Review evaluation of aducanumab, including a healthcare system perspective and patient lifetime time horizon. We assumed greater treatment efficacy (30% slowing of progression from mild cognitive impairment and from mild AD states), consistent with benefits anticipated from DMTs currently under evaluation. The impact of different treatment durations and therapy costs were examined. A hypothetical DMT used continuously until patients progress to severe AD had an incremental cost-effectiveness ratio (ICER) ranging from $297K/QALY to $586K/QALY as annual treatment cost varied from $28K to $56K, based on the original and subsequently reduced prices set by one manufacturer. A hypothetical DMT with similar efficacy but used for only 18 months yielded ICERs from $89K to $173K/QALY. Use of a diagnostic test (cost $4000) to identify 40% of patients who could discontinue treatment due to amyloid negativity at 6 months reduced total treatment costs by 25% and yielded ICERs from $77K/QALY to $139K/QALY. Results were sensitive to mean age, efficacy assumptions and treatment cost. Efficacious Alzheimer’s DMTs used for limited durations or until amyloid plaque clearance have potential to deliver value consistent with accepted cost-effectiveness thresholds.

Randomized placebo-controlled trials in the development of disease-modifying treatments for Alzheimer’s disease are typically of short duration (12–18 months), and health economic modeling requires extrapolation of treatment effects beyond the trial period.

Background: There is little longitudinal data on resource use and costs associated with Alzheimer's disease (AD) in France. Objectives: To evaluate resource use and societal costs associated with AD in a French cohort of patients and their caregivers and the effect of patient cognitive decline on costs over an 18-month period. Methods: Community-dwelling patients with mild, moderate, or moderately severe/severe AD dementia (n = 419) were followed-up for 18 months. Total societal costs were estimated by applying 2010 unit costs to resource use, including outpatient visits, hospital days, institutionalization, and caregiver hours. Cognitive function was assessed by Mini Mental State Examination scores. Results: Mean cumulative total costs over the 18-month period were 24,140 for patients with mild AD dementia, 34,287 for those with moderate AD dementia, and 44,171 for those with moderately severe/severe AD dementia (P < 0.001; ANOVA comparison between severity groups). The biggest contributor to total societal costs was caregiver informal care ( >50% of total costs at all stages of AD dementia). Cognitive decline (>3-point decrease in Mini-Mental State Examination score or institutionalization) was associated with a 12.5% increase in total costs (P = 0.02). Significant differences were observed across severity groups for caregiver time (P < 0.001); mean monthly caregiver time increased at each time point over the 18 months in each severity group. Conclusions: Increasing severity of AD dementia in France is associated with increased use of resources as well as increased total societal and patient costs; informal care was the greatest cost contributor. Clinically meaningful cognitive decline is associated with significantly increased costs.

Background: While functional loss forms part of the current diagnostic criteria used to identify dementia due to Alzheimer’s disease, the gradual and progressive nature of the disease makes it difficult to recognize clinically relevant signposts that could be helpful in making treatment and management decisions. Having previously observed a significant relationship between stages of functional dependence (the level of assistance patients require consequent to Alzheimer’s disease deficits, derived from the Alzheimer’s Disease Cooperative Study – Activities of Daily Living Scale) and cognitive severity, we investigated whether measures of functional dependence could be utilized to identify clinical milestones of Alzheimer’s disease progression. OBJECTIVES: To describe the patterns of change in dependence over the course of 18 months in groups stratified according to cognitive Alzheimer’s disease dementia severity (determined using the Mini-Mental State Examination score) and to identify characteristics associated with patients showing worsening dependence (progressors) versus those showing no change or improvement (non-progressors). DESIGN: Analysis of longitudinal data from the GERAS study. SETTING: GERAS is an 18-month prospective, multicenter, naturalistic, observational cohort study reflecting the routine care of patients with Alzheimer’s disease in France, Germany, and the United Kingdom. PARTICIPANTS: 1495 community-living patients, aged ≥55 years, diagnosed with probable Alzheimer’s disease dementia, and their caregivers. MEASUREMENTS: Dependence levels, cognitive function, behavioral symptoms, caregiver burden, and cost were assessed at baseline and at 18 months. RESULTS: Of 971 patients having both baseline and 18-month data, 42% (408) were progressors and 563 (58%) were non-progressors. This general pattern held for all three levels of baseline Alzheimer’s disease dementia severity – mild (Mini-Mental State Examination score 21–26), moderate (15–20) or moderately severe/severe (<15) – with 40–45% of each group identified as progressors and 55–60% as non-progressors. No baseline differences were seen between progressors and non-progressors in cognitive scores or behavioral symptoms, although progressors had significantly shorter times since diagnosis and showed milder functional impairment. Baseline factors predictive of increasing dependence over 18 months included more severe cognitive impairment, living with others, and having multiple caregivers. A higher level of initial dependence was associated with less risk of dependence progression. Total societal costs of care also increased with greater dependence. CONCLUSIONS: In this large cohort, 42% of Alzheimer’s disease dementia patients at all levels of cognitive severity became more dependent within 18 months of observation while 58% did not progress. Dependence levels may be considered as meaningful interim clinical milestones that reflect Alzheimer’s disease-related functional deficits, although a time frame that extends beyond 18 months may be necessary to observe changes if used in clinical trials or other longitudinal studies. Recognition of predictors of greater dependence offers opportunities for intervention.

Aim: To analyse baseline data from 552 community-dwelling German patients with mild (n=228), moderate (n=157) or moderately severe/severe (n=167) AD, who participated in a 18-month observational study (GERAS) conducted in three European countries aiming at determining resource use and total costs associated with Alzheimer's disease (AD), stratified by ADseverity at baseline.Method: Patient and caregiver demographics andbaseline characteristics were documented. Resource use information and time spent on informal care by nonprofessional caregivers was obtained using the Resource Utilization in Dementia instrument. Total societal costs at baseline were estimated from patients' healthcare and social care costs, and from caregivers' healthcare and informal care costs in the month prior to baseline.Results: Estimated mean annual total societal costs (2010 values) per patient at baseline differed significantly (p<0.001) between groups with mild (Euro15739), moderate (Euro28941) and moderately severe/severe AD(Euro44662). Informal care costs were the largest contributor to total societal costs, comprising 49, 55 and 64% of the total societal costs of mild, moderate and moderately severe/severe AD, respectively.Conclusion: Total societal costs of ADin community-dwelling individuals increase as disease severity worsens. Caregiver informal care costs were the major cost component at all levels of ADseverity.

BACKGROUND:Because Alzheimer's disease (AD) is characterized by a gradual decline, it can be difficult to identify distinct clinical milestones that signal disease advancement. Adapting a functional scale may be a useful way of staging disease progression that is more informative for healthcare systems.OBJECTIVES:To adapt functional scale scores into discrete levels of dependence as a way of staging disease progression that is more informative to care providers and stakeholders who rely on the functional impact of diseases to determine access to supportive services and interventions.DESIGN:Analysis of data from the GERAS study.SETTING:GERAS is an 18-month prospective, multicenter, naturalistic, observational cohort study reflecting the routine care of patients with AD in France, Germany, and the United Kingdom.PARTICIPANTS:Data were from baseline results of 1497 community-living patients, aged ≥55 years, diagnosed with probable AD and their caregivers.MEASUREMENTS:We used data from the Alzheimer's Disease Cooperative Study Activities of Daily Living Inventory (ADCS-ADL) and mapped items onto established categories of functional dependence, validated using clinical and economic measures. Cognitive function, behavioral symptoms, caregiver burden, and cost were assessed. Based on stages of functional dependence described by the Dependence Scale, individual ADCS-ADL items were used to approximate 6 dependence levels.RESULTS:There was a significant relationship between assigned level of dependence derived from the ADCS-ADL score and cognitive severity category. As the assigned level of dependence increased, the associated clinical and economic indicators demonstrated a pattern of greater disease severity.CONCLUSIONS:This mapping provides initial support for dependence levels as appropriate interim clinical milestones that characterize the functional deficits associated with AD.

The use of indirect comparisons (IC) is now an integral part of the Health Technology Assessment (HTA). When new interventions are assessed, Individual patient data (IPD), if available, can be used for comparison with published aggregated (AGR) data. Methods exist to assess heterogeneity and inconsistency of IC; however in the absence of sufficient studies, IC may be required where differences may exist between inclusion/exclusion criteria, definition of outcomes and patient characteristics. Signorovitch (2010) has proposed the use of MAIC when IPD is available and provided various methods for matching the IPD to the AGR study. Through simulations we assessed different approaches to weight calculations, including weighting as originally proposed by Signorovitch (WS) vs. Entropy Balancing (EB). WS is based on propensity score weights (odds of being in AGR trial) while EB relies on a maximum entropy reweighting scheme. Simulation show the optimal weighting method is to match on covariates against the AGR treatment and control arms separately. In addition, rebalancing on prognostic variables between the IPD arms using EB is beneficial when they are not reported in AGR. For example, with true treatment mean difference between AGR and IPD (IC) of one, six predictive variables in AGR and IPD and three prognostic variables in IPD, the Bucher method gives a biased estimate 0.31(Average Bootstrap 95% Confidence interval: -0.45, 1.08). WS balancing gives 0.99(0.23 –1.73), balancing on each arm separately gives 0.99(0.30-1.68) while rebalancing using EB gives 0.99(0.47-1.52). Also, simulations demonstrate that including placebo response into the weighting in addition to baseline covariates can produce biased results and is not recommended. MAIC can be improved if weighting is performed on each arm separately together with rebalancing of the IPD on the prognostic variables not reported in AGR.

Signorovitch (2010) describes MAIC that focuses on matching one study with individual patient data (IPD) to the covariates in one study with aggregated data (AGR). However in most scenarios there are likely to be multiple studies with IPD and AGR that need to be included in the Indirect Comparison. In addition it may be necessary to extend the network of treatments to include more than the two treatments with a single comparator. We propose a number of potential solutions for including multiple studies and multiple treatments in the MAIC and assess these using simulations with the weighting methods proposed by Signorovitch(2010) as well as with Entropy Balancing Hainmueller (2012) When multiple IPD studies exist then MAIC can be conducted if you consider a)Pooling IPD studies, into one large study and match against the AGR study or b) matching each IPD study against the AGR study. For multiple AGR studies then the IPD data can be matched against a) just one AGR study, b) the average patient characteristics from the AGR studies, c) the average mean and variances from the AGR studies or d) the distribution of patient characteristics using MCMC from the AGR studies. To apply a MAIC in Networks involving multiple studies the choice of study to match on could be an issue so it is important that the assumptions surrounding the NMA are tested, and only if there is no evidence to suggest inconsistency and heterogeneity within the Network, should IPD studies be added to the Network via an MAIC. MAIC can be applied in scenarios where you have multiple studies and treatments if the existing Network satisfies the assumptions around heterogeneity and inconsistency required when performing Network Meta-Analysis.

The objective of this study was to evaluate health care costs over time in patients diagnosed with vascular dementia (VD) who had a prior diagnosis of Alzheimer’s disease (AD) compared with similar patients with no prior diagnosis of AD in patients in Italy. A cohort of patients whose most recent diagnostic records for dementia indicated presence of VD was identified from administrative databases of a sample of Italian Local Health Units. VD patients were required to have no indication for other dementia types between their earliest indication of cognitive decline or earliest VD diagnosis (index date) and their most recent VD diagnoses and to have continuous data visibility for ≥6 months prior to the index date. The cohort was then categorised based on diagnosis of AD prior to their first confirmed VD diagnosis, and followed for up to 3 years post-index. Patients with prior AD were matched to similar patients with no prior AD using propensity score methods. Annual healthcare costs were calculated for the two cohorts. Of the 2652 VD patients, who were primarily identified in a hospital setting, 103 (4%) had prior AD diagnoses. All 103 patients with prior AD were successfully matched (ratio 1:5) for the cost analysis. Patients with prior AD misdiagnosis had higher annual healthcare costs over the follow up period compared to those patients without prior diagnosis. For patients with 3 years of follow-up average health care costs (95% Confidence limits) per patient with prior AD diagnosis were €10,899(7410, 14387), €5256(2578, 7754), €4409 (2090, 6727) in years 1, 2 and 3 respectively, compared to €8196(6652, 9741), €3404(2177, 4632) and €2770(1958, 3582) in the no prior AD cohort. This retrospective analysis suggests that possible misdiagnosis of AD among Italian patients with VD results in greater health care costs.

To compare resource utilization drivers of societal costs for Alzheimer’s disease (AD) over 18-months in three countries participating in an observational study. GERAS is a prospective, multi-centre, non-interventional cohort study in France (n=419), Germany (n=550) and the UK (n=526). Resource use of AD patients and caregivers (including informal caregiving time and institutionalization), contributing to >1% total societal costs (comprising patient health and social care costs and informal caregiver costs - based on 2010 prices) were identified and assessed for country differences using Generalised Linear Models of repeated measures or Cox models, adjusting for key patient and caregiver characteristics. 18-month societal costs per patient: France €33,300, Germany €38,200 and UK €37,900. Caregiver time spent assisting patient with basic and instrumental activities of daily living (ADL) made the largest contribution to total societal costs in each country (55-69%). Caregivers in France spent less time on basic and instrumental ADLs and were less likely to miss work. Patients in France used more community care services and were more likely to spend time in respite care, whereas German patients were less likely to use respite care and had slower time to institutionalization (Hazard Ratio 0.59 (95% CI 0.41-0.84) Germany, 0.84 (0.60-1.18) France, reference UK; p-value 0.0143). UK caregivers spent more time on instrumental ADLs while patients used fewer outpatient resources but were more likely to receive financial support. No country differences in hospital stays or use of AD medication were seen. Caregiver time was consistently the main contributor to societal costs. Although time on basic and instrumental ADLs differed across countries, some of this may be explained by use of community care services and institutionalization. Other resources had different patterns of use across countries, reflecting country-specific health and social care systems.

To use baseline results from a prospective observational study in Alzheimer’s disease (AD) to evaluate methods for dealing with missing longitudinal AD cost data. GERAS is an 18-month observational study of costs associated with AD. Total societal costs included patient health care costs (including hospitalizations, outpatient visits, and medication) and social care costs (including home-care and day-center sessions), and caregiver informal care costs (from time spend on informal care). Missing longitudinal cost data due to patient death/institutionalization was classified as not missing at random (NMAR). Cost data missing for other reasons was classified as missing at random (MAR) or missing completely at random (MCAR). To assess the impact of imputing missing longitudinal cost data, patterns of missing data during follow-up were simulated based on baseline GERAS data to generate 10%, 20%, 30% and 40% missing data for MCAR, MAR and NMAR classifications. Naïve methods (including complete case analysis, mean imputation and regression models), multiple imputation (MI) and a fixed cost were applied to each dataset and %bias assessed using (estimated-actual)/actual cost*100. Total baseline societal costs were available for 1488 (99.4%) of enrolled patients, with a mean monthly cost of €2101(95% CI: €1980-€2222). For MCAR datasets, naïve methods performed as well as MI (20% missing data: 0.6-10.9% bias naïve methods vs 0.2-6.1% MI). For MAR data, MI methods performed better (-3.2% to -14.3% bias) than naïve methods (6.6%-18.0% bias). All approaches were consistently poor with NMAR data (bias range -31.4% to -38.6%); the best performing approach was to impute a fixed value (monthly cost of institutionalisation) with -22.6% bias. For all approaches %bias increased with missing data volume. Methods used to impute missing cost data in AD should be tailored depending on the type of missing data, using sensitivity analysis to assess the impact of any assumptions.