Biased estimation with shared parameter models in the presence of competing dropout mechanisms.

Recently, Thomadakis et al. quantified potential sources of bias that can occur when shared parameter (SP) models are used to jointly model longitudinal trends of a biomarker over time (e.g., a slope) and time-to-dropout in an effort to address concerns over possible informative censoring. Although SP models induce no bias under a missingness completely at random dropout mechanism, the authors demonstrate that bias can occur under a missingness at random (MAR) dropout mechanism wherein dropout depends on the observed biomarker data. To address this, the authors propose including the most recent observed marker value within the hazard function for the time-to-dropout portion of an SP model. They demonstrate via a limited simulation that the proposed model minimizes bias under a specific MAR dropout mechanism and a specific missingness not-at-random dropout mechanism. In the present article, we compare and contrast their work with that of previous authors by illustrating via simulation and an example the degree of bias or lack thereof that can occur when applying SP models, particularly, in the presence of competing dropout mechanisms. We propose the use of a competing risk SP model as a means to minimize bias whenever competing dropout mechanisms are suspected assuming the competing mechanisms result from distinct observable causes of dropout.

Acute Treatment Effects on GFR in Randomized Clinical Trials of Kidney Disease Progression.

BACKGROUND: Acute changes in GFR can occur after initiation of interventions targeting progression of CKD. These acute changes complicate the interpretation of long-term treatment effects. METHODS: To assess the magnitude and consistency of acute effects in randomized clinical trials and explore factors that might affect them, we performed a meta-analysis of 53 randomized clinical trials for CKD progression, enrolling 56,413 participants with at least one estimated GFR measurement by 6 months after randomization. We defined acute treatment effects as the mean difference in GFR slope from baseline to 3 months between randomized groups. We performed univariable and multivariable metaregression to assess the effect of intervention type, disease state, baseline GFR, and albuminuria on the magnitude of acute effects. RESULTS: The mean acute effect across all studies was -0.21 ml/min per 1.73 m2 (95% confidence interval, -0.63 to 0.22) over 3 months, with substantial heterogeneity across interventions (95% coverage interval across studies, -2.50 to +2.08 ml/min per 1.73 m2). We observed negative average acute effects in renin angiotensin system blockade, BP lowering, and sodium-glucose cotransporter 2 inhibitor trials, and positive acute effects in trials of immunosuppressive agents. Larger negative acute effects were observed in trials with a higher mean baseline GFR. CONCLUSION: The magnitude and consistency of acute GFR effects vary across different interventions, and are larger at higher baseline GFR. Understanding the nature and magnitude of acute effects can help inform the optimal design of randomized clinical trials evaluating disease progression in CKD.

GFR Slope as a Surrogate End Point for Kidney Disease Progression in Clinical Trials: A Meta-Analysis of Treatment Effects of Randomized Controlled Trials.

BACKGROUND: Surrogate end points are needed to assess whether treatments are effective in the early stages of CKD. GFR decline leads to kidney failure, but regulators have not approved using differences in the change in GFR from the beginning to the end of a randomized, controlled trial as an end point in CKD because it is not clear whether small changes in the GFR slope will translate to clinical benefits. METHODS: To assess the use of GFR slope as a surrogate end point for CKD progression, we performed a meta-analysis of 47 RCTs that tested 12 interventions in 60,620 subjects. We estimated treatment effects on GFR slope (mean difference in GFR slope between the randomized groups), for the total slope starting at baseline, chronic slope starting at 3 months after randomization, and on the clinical end point (doubling of serum creatinine, GFR<15 ml/min per 1.73 m2, or ESKD) for each study. We used Bayesian mixed-effects analyses to describe the association of treatment effects on GFR slope with the clinical end point and to test how well the GFR slope predicts a treatment's effect on the clinical end point. RESULTS: Across all studies, the treatment effect on 3-year total GFR slope (median R2=0.97; 95% Bayesian credible interval [BCI], 0.78 to 1.00) and on the chronic slope (R2 0.96; 95% BCI, 0.63 to 1.00) accurately predicted treatment effects on the clinical end point. With a sufficient sample size, a treatment effect of 0.75 ml/min per 1.73 m2/yr or greater on total slope over 3 years or chronic slope predicts a clinical benefit on CKD progress with at least 96% probability. CONCLUSIONS: With large enough sample sizes, GFR slope may be a viable surrogate for clinical end points in CKD RCTs.

Performance of GFR Slope as a Surrogate End Point for Kidney Disease Progression in Clinical Trials: A Statistical Simulation.

BACKGROUND: Randomized trials of CKD treatments traditionally use clinical events late in CKD progression as end points. This requires costly studies with large sample sizes and long follow-up. Surrogate end points like GFR slope may speed up the evaluation of new therapies by enabling smaller studies with shorter follow-up. METHODS: We used statistical simulations to identify trial situations where GFR slope provides increased statistical power compared with the clinical end point of doubling of serum creatinine or kidney failure. We simulated GFR trajectories based on data from 47 randomized treatment comparisons. We evaluated the sample size required for adequate statistical power based on GFR slopes calculated from baseline and from 3 months follow-up. RESULTS: In most scenarios where the treatment has no acute effect, analyses of GFR slope provided similar or improved statistical power compared with the clinical end point, often allowing investigators to shorten follow-up by at least half while simultaneously reducing sample size. When patients' GFRs are higher, the power advantages of GFR slope increase. However, acute treatment effects within several months of randomization can increase the risk of false conclusions about therapies based on GFR slope. Care is needed in study design and analysis to avoid such false conclusions. CONCLUSIONS: Use of GFR slope can substantially increase statistical power compared with the clinical end point, particularly when baseline GFR is high and there is no acute effect. The optimum GFR-based end point depends on multiple factors including the rate of GFR decline, type of treatment effect and study design.

Calculating Standard Kt/V during Hemodialysis Based on Urea Mass Removed.

BACKGROUND/AIMS: We derived a novel equation for calculating weekly urea standard Kt/V (stdKt/V) during hemodialysis (HD) based on urea mass removed, comparable to the approach during peritoneal dialysis. METHODS: Theoretical consideration of urea mass balance during HD led to the following equation for stdKt/V, namely, stdKt/V = N × (URR + UFV/V), where N is the number of treatments per week, URR is urea reduction ratio per treatment, UFV is ultrafiltration volume per treatment, and V is postdialysis urea distribution volume. URR required corrections for postdialysis rebound and intradialytic urea generation. We compared the accuracy of this approach with previous equations for stdKt/V by numerical simulations using a 2-compartment model of urea kinetics for thrice-weekly and more frequent HD prescriptions. RESULTS: The proposed equation based on urea mass removed predicted values of stdKt/V that are equivalent to those calculated by previous equations for stdKt/V. CONCLUSION: This work provides a novel approach for calculating stdKt/V during HD and strengthens the theoretical understanding of stdKt/V.

Survival of propensity matched incident peritoneal and hemodialysis patients in a United States health care system.

We sought to compare survival among incident peritoneal dialysis (PD) patients to matched hemodialysis (HD) patients who received pre-dialysis care, including permanent dialysis access placement. Patients starting PD were propensity matched to those starting HD. HD patients who used a central venous catheter during the first 90 days of dialysis were excluded. Stratified Cox proportional hazards models were used to compare patient survival using both intent-to-treat and as-treated analyses. In the intent-to-treat analysis, patients were followed from the date of first dialysis until death and censored at the earliest of the following: renal transplantation, death, renal recovery, loss to follow-up or study end. In the as-treated analysis, patients were also censored at the time of modality change. A total of 1003 matched pairs were obtained from 11,301 incident patients (10,298 HD and 1003 PD). The cumulative hazard ratio for death at one year was 2.38 (95% CI 1.68-3.40) and 2.10 (1.50-2.94) for HD relative to PD patients in the as-treated and intent-to-treat analyses, respectively. The cumulative risk of death, as estimated by the cumulative hazard ratio, favored PD for almost up to 3 years of follow-up in the as-treated analysis and nearly 2 years of follow-up in the intent-to-treat analysis with no differences thereafter. The higher adjusted rate of death observed for HD patients cannot be attributed to initial use of central venous catheters or lack of pre-dialysis care.

PSA enzymatic activity: a new biomarker for assessing prostate cancer aggressiveness.

BACKGROUND: With the advent of widespread prostate-specific antigen (PSA) testing in recent decades, prostate cancer (PCa) has emerged as the most frequently diagnosed non-skin cancer among men in the U.S. and Europe. Greater screening rates coupled with improved detection methods have caused a controversial upsurge in the number of men undergoing prostate biopsy and subsequent treatment. However, current diagnostic techniques generally suffer from limited ability to identify which seemingly indolent cancers are biologically aggressive. METHODS: We collected prostatic fluid from 778 post-radical prostatectomy specimens and randomly selected samples from both the clinically confirmed aggressive (n = 50) and non-aggressive (n = 50) prostate cancer populations. We measured the level of proteolytic enzyme activity of PSA (aPSA) in each sample and used receiver operating characteristic (ROC) analysis to correlate aPSA levels with prostate cancer aggressiveness. RESULTS: We found aPSA in prostatic fluid to be inversely proportional to disease stage, such that patients with the most aggressive PCa have on average significantly reduced aPSA compared to those with less aggressive disease. Significantly, our results suggest that many (22% in our study population) of the diagnosed patients with non-aggressive PCa could have averted or delayed radical prostatectomy. CONCLUSIONS: Given the high level of debate surrounding PSA screening effectiveness [3-5] and the recent U.S. Preventative Services Task Force recommendation to discontinue PSA screening [6], our results provide renewed hope that a clinical monitoring tool may emerge that truly refines PCa treatment decision-making.

Assessing creatinine clearance from modification of diet in renal disease study equations in the ADEMEX cohort: limitations and potential applications.

BACKGROUND AND OBJECTIVES: Twenty-four-hour urine and dialysate collections provide accepted means to assess adequacy in peritoneal dialysis (PD). Recent publications suggest that creatinine clearance (CrCl) estimated from the Modification of Diet in Renal Disease (MDRD) equations (eCrCl) accurately approximates measured CrCl (mCrCl) derived from 24-hour collections of urine and dialysate and might serve as an alternative means to assess small-solute clearance and adequacy in PD. DESIGN, SETTING, PARTICIPANTS, & MEASUREMENTS: Post hoc analysis of data from patients enrolled in ADEMEX was used to assess agreement between mCrCl and eCrCl derived by four- and six-variable MDRD equations (4V-MDRDE and 6V-MDRDE, respectively). Additionally, associations among mCrCl, eCrCl, and survival were determined. RESULTS: Acceptable precision was observed between mCrCl and 4V-MDRDE-eCrCl and 6V-MDRDE-eCrCl for the entire cohort. Precision was markedly diminished when analysis was limited to functionally anuric patients with mCrCl < 12 ml/min per 1.73 m². Although there was no association between survival and mCrCl, for every 1-ml/min per 1.73 m² increase in 4V- and 6V-MDRDE-eCrCl, there was a 6% and 4% increase in risk of death, respectively. There was a negative association between MDRDE-eCrCl and creatinine appearance rates, suggesting MDRDE-eCrCl is significantly confounded by individual differences in muscle mass. CONCLUSIONS: MDRDE-eCrCl provides demographically comparable values to 24-hour urine and dialysate collections across the ADEMEX cohort. However, MDRDEs should not be used to assess small-solute removal or adequacy in individual PD patients or to predict outcome in any cohort of patients over narrow ranges of limited clearance.

What is the true number needed to screen and treat to save a life with prostate-specific antigen testing?

PURPOSE: The European Randomized Study of Screening for Prostate Cancer (ERSPC) reported a 20% mortality reduction with prostate-specific antigen (PSA) screening. However, they estimated a number needed to screen (NNS) of 1,410 and a number needed to treat (NNT) of 48 to prevent one prostate cancer death at 9 years. Although NNS and NNT are useful statistics to assess the benefits and harms of an intervention, in a survival study setting such as the ERSPC, NNS and NNT are time specific, and reporting values at one time point may lead to misinterpretation of results. Our objective was to re-examine the effect of varying follow-up times on NNS and NNT using data extrapolated from the ERSPC report. MATERIALS AND METHODS: On the basis of published ERSPC data, we modeled the cumulative hazard function using a piecewise exponential model, assuming a constant hazard of 0.0002 for the screening and control groups for years 1 to 7 of the trial and different constant rates of 0.00062 and 0.00102 for the screening and control groups, respectively, for years 8 to 12. Annualized cancer detection and drop-out rates were also approximated based on the observed number of individuals at risk in published ERSPC data. RESULTS: According to our model, the NNS and NNT at 9 years were 1,254 and 43, respectively. Subsequently, NNS decreased from 837 at year 10 to 503 at year 12, and NNT decreased from 29 to 18. CONCLUSION: Despite the seemingly simplistic nature of estimating NNT, there is widespread misunderstanding of its pitfalls. With additional follow-up in the ERSPC, if the mortality difference continues to grow, the NNT to save a life with PSA screening will decrease.

Effects of radiotherapy with or without chemotherapy on tongue strength and swallowing in patients with oral cancer.

BACKGROUND: Oral tongue strength and swallowing ability are reduced in patients treated with chemoradiotherapy for oral and oropharyngeal cancer. METHODS: Patients with oral or oropharyngeal cancer treated with high-dose chemoradiotherapy underwent tongue strength, swallowing, and dietary assessments at pretreatment and 1, 3, 6, and 12 months posttreatment. Tongue strength was assessed using the Iowa Oral Performance Instrument (IOPI). Oral and pharyngeal residue was evaluated utilizing videofluoroscopy. RESULTS: Mean maximum tongue strength dropped a nonsignificant amount immediately after treatment, and then increased significantly at 6- and 12-months posttreatment completion. Analyses were adjusted for patient dropout. Tongue strength was not significantly correlated with swallow observations of percentage oral and pharyngeal residue. Ability to eat various diet consistencies was reduced after treatment but improved over time at a rate similar to changes in oral intake and type of diet. CONCLUSIONS: Parallel but not significant changes in oral intake, diet, and tongue strength in the first year post chemoradiation therapy need further study in a larger population.

The reverse epidemiology of plasma total homocysteine as a mortality risk factor is related to the impact of wasting and inflammation.

BACKGROUND: The reason(s) for the apparently paradoxical 'reverse' association in end-stage renal disease (ESRD) patients in whom a low, rather than a high, total plasma total homocysteine (tHcy) level is an indicator of poor outcome remains unclear. The aim of this study was to examine whether the inverse association maintains, mitigates or reverses after comprehensive multivariate adjustment for the presence of wasting and inflammation as well as other potential confounders. METHODS: We studied 317 ESRD patients starting dialysis therapy. Fasting blood samples were taken for the analyses of tHcy, serum albumin, C-reactive protein (CRP), serum creatinine and plasma folate. Nutritional status was assessed by subjective global assessment (SGA). Survival was followed for up to 66 months; 105 patients died. RESULTS: Using Kaplan-Meier analysis, a low tHcy concentration (< or =30 micromol/l) was associated with higher all-cause and cardiovascular (CV) mortality (P < 0.05). Using Cox proportional analysis adjusting for age, gender, glomerular filtration rate = GFR, cardiovascular disease = CVD, plasma folate, total cholesterol and diabetes mellitus, the all-cause and CV mortality still tended to be high for patients with low tHcy. Adding nutritional and inflammation markers (Body mass index = BMI, SGA, serum creatinine, serum albumin and CRP), a low tHcy level was no longer associated with higher mortality but a trend for high tHcy was observed. CONCLUSIONS: The link between wasting inflammation and a low tHcy appears to be responsible for the reverse association between plasma tHcy and clinical outcome in ESRD patients. After adjustment for confounders including nutritional and inflammation markers, a trend towards increased death risk for high, rather than low, tHcy levels was apparent after adjustment.

Modeling of icodextrin in PD Adequest 2.0.

OBJECTIVE: To validate the use of a modified three-pore model for predicting fluid transport during long dwell exchanges that use a 7.5% icodextrin solution. DESIGN: A nonrandomized, single group, repeated measures study. PATIENTS: Ten peritoneal dialysis patients underwent a single 8-hour exchange of a 7.5% icodextrin solution. All patients were naive to icodextrin. MAIN OUTCOME MEASURES: A modified three-pore model was used to model solute and fluid transport during each 8-hour exchange. Concordance correlation coefficients were used to estimate the level of agreement between modeled and measured values of net ultrafiltration (UF) and intraperitoneal volume. METHODS: Each patient underwent a modified 8-hour standard peritoneal permeability analysis using a 2-L 7.5% icodextrin exchange. Dextran 70 was added to the icodextrin solution as volume marker to estimate fluid transport kinetics. Transcapillary UF, fluid absorption, and intraperitoneal volumes were assessed via the volume marker at 0, 5, 15, 30, 60, 120, 240, 300, 360, 420, and 480 minutes. RESULTS: There was strong agreement (concordance correlation = 0.9856) between net UF as measured by the volume marker data and net UF as modeled using the modified three-pore model implemented in PD Adequest (Baxter Healthcare, Deerfield, Illinois, USA). CONCLUSIONS: Net UF and intraperitoneal volumes for long dwell exchanges using a 7.5% icodextrin solution can be accurately modeled with a modified three-pore model. Steady state icodextrin plasma levels are needed to accurately predict net UF for chronic users of icodextrin.

Estrogen and progesterone levels in nipple aspirate fluid of healthy premenopausal women: relationship to steroid precursors and response proteins.

BACKGROUND: Concentrations of estrogen and progesterone within the breast could provide a better reflection of breast cancer risk than levels in the circulation. We developed highly sensitive immunoassays for multiple steroid hormones and proteins in the nipple aspirate fluid (NAF), which can be obtained noninvasively with a simple suction device. Previous studies showed that NAF hormone levels are strongly correlated between breasts and within a single breast over time and are predictably related to hormone replacement therapy or use of oral contraceptives. This study evaluates the relationship of NAF estrogen and progesterone levels to those in serum and saliva, the relationship of NAF estradiol to androgenic and estrogenic precursors in NAF, and the relationship of NAF hormone levels to those of response proteins such as cathepsin D and epidermal growth factor (EGF). METHODS: Normal premenopausal women collected saliva daily and donated blood and NAF in the midluteal phases of menstrual cycles at intervals of 0, 4, 12, and 15 months. Analytes were measured by immunoassays after solvent fractionation. Log-transformed values were fit to repeated measures analysis of covariance models to ascertain associations between analytes. RESULTS: Small nonsignificant associations were found between NAF and serum or salivary estradiol. However, progesterone in NAF was significantly associated with progesterone in serum and saliva (R=0.18 and 0.32, respectively). Within NAF, the estradiol precursors estrone sulfate, androstenedione, and dehydroepiandrosterone were significantly associated with estradiol concentration (P<0.06), and a multiprecursor model explained the majority of variance in NAF estradiol (model R(2)=0.83). Cathepsin D and EGF in NAF could not be predicted from serum or salivary steroid measurements; however, both could be predicted from estradiol and its precursors in NAF (model R(2)=0.70 and 0.93, respectively). CONCLUSIONS: By showing consistent associations between estradiol and its precursors and response proteins, these data provide support for the biological validity of NAF hormone measurements and for the importance of steroid interconversion by aromatase and sulfatase within the breast. The low correlation between estrogen levels in NAF and those in serum or saliva suggests that the degree of association between estrogen or its androgen precursor levels and risk of breast cancer observed in epidemiologic studies using serum estimates might be highly attenuated.

Shared parameter models for the joint analysis of longitudinal data and event times.

Longitudinal studies often gather joint information on time to some event (survival analysis, time to dropout) and serial outcome measures (repeated measures, growth curves). Depending on the purpose of the study, one may wish to estimate and compare serial trends over time while accounting for possibly non-ignorable dropout or one may wish to investigate any associations that may exist between the event time of interest and various longitudinal trends. In this paper, we consider a class of random-effects models known as shared parameter models that are particularly useful for jointly analysing such data; namely repeated measurements and event time data. Specific attention will be given to the longitudinal setting where the primary goal is to estimate and compare serial trends over time while adjusting for possible informative censoring due to patient dropout. Parametric and semi-parametric survival models for event times together with generalized linear or non-linear mixed-effects models for repeated measurements are proposed for jointly modelling serial outcome measures and event times. Methods of estimation are based on a generalized non-linear mixed-effects model that may be easily implemented using existing software. This approach allows for flexible modelling of both the distribution of event times and of the relationship of the longitudinal response variable to the event time of interest. The model and methods are illustrated using data from a multi-centre study of the effects of diet and blood pressure control on progression of renal disease, the modification of diet in renal disease study.

The differential impact of risk factors on mortality in hemodialysis and peritoneal dialysis.

BACKGROUND: While the survival ramifications of dialysis modality selection are still debated, it seems reasonable to postulate that outcome comparisons are not the same for all patients at all times. Trends in available data indicate the relative risk of death with hemodialysis (HD) compared to peritoneal dialysis (PD) varies by time on dialysis and the presence of various risk factors. This study was undertaken to identify key patient characteristics for which the risk of death differs by dialysis modality. METHODS: Analyses utilized incidence data from 398,940 United States Medicare patients initiating dialysis between 1995 and 2000. Proportional hazards regression identified the presence of diabetes, age, and the presence of comorbidity as factors that significantly interact with treatment modality. Stratifying by these factors, proportional and nonproportional hazards models were used to estimate relative risks of death [RR (HD:PD)]. RESULTS: Of the 398,940 patients studied, 11.6% used PD as initial therapy, 45% had diabetes mellitus (DM), 51% were 65 years or older, and 55% had at least one comorbidity. Among the 178,693 (45%) patients with no baseline comorbidity, adjusted mortality rates in nondiabetic (non-DM) patients were significantly higher on HD than on PD [age 18-44: RR (95% CI) = 1.24 (1.07, 1.44); age 45-64: RR = 1.13 (1.02, 1.25); age 65+: RR = 1.13 (1.05, 1.21)]. Among diabetic (DM) patients with no comorbidity, HD was associated with a higher risk of death among younger patients [age 18-44: RR = 1.22(1.05, 1.42)] and a lower risk of death among older patients [age 45-64: RR = 0.92 (0.85, 1.00); age 65+: RR = 0.86 (0.79, 0.93)]. Within the group of 220,247 (55%) patients with baseline comorbidity, adjusted mortality rates were not different between HD and PD among non-DM patients [age 18-44: RR = 1.19 (0.94, 1.50); age 45-64: RR = 1.01 (0.92, 1.11); age 65+: RR = 0.96 (0.91, 1.01)] and younger DM patients [age 18-44: RR = 1.10 (0.92, 1.32)], but were lower with HD among older DM patients with baseline comorbidity [age 45-64: RR = 0.82 (0.77, 0.87); age 65+: RR = 0.80 (0.76, 0.85)]. CONCLUSION: Valid mortality comparisons between HD and PD require patient stratification according to major risk factors known to interact with treatment modality. Survival differences between HD and PD are not constant, but vary substantially according to the underlying cause of ESRD, age, and level of baseline comorbidity. These results may help identify technical advances that will improve outcomes of patients on dialysis.

Hemoglobin levels and erythropoietin doses in hemodialysis and peritoneal dialysis patients in the United States.

Erythropoietic agents, a cornerstone of management, are a major component of the cost of renal replacement therapy. The objectives of this study were to compare (on a month-by-month basis) U.S. hemodialysis and peritoneal dialysis patients in terms of the proportion of patients receiving erythropoietin, erythropoietin doses, and hemoglobin levels after initiation of erythropoietin. Patients studied (hemodialysis, n = 121,970; peritoneal dialysis, n = 7129) began dialysis between 1995 and 2000, had Medicare as their primary payer, were 65 yr old or older at dialysis inception, had no erythropoietin claims before dialysis inception, and did not have a switch in dialysis modality in the first 6 mo of dialysis therapy. Total monthly erythropoietin doses and average monthly hemoglobin levels were calculated from Medicare claims. The proportion of patients who received erythropoietin plateaued at 3 mo in both groups: 25% in peritoneal dialysis patients and 80% in hemodialysis patients. However, monthly erythropoietin doses plateaued at 30,000 units in peritoneal dialysis patients and 60,000 units in hemodialysis patients, a disparity not explicable by differences in baseline characteristics. Among subjects who received erythropoietin, mean hemoglobin levels were similar at steady state in both populations and met the National Kidney Foundation Dialysis Outcomes Quality Initiative hemoglobin target level of 11 to 12 g/dl. Hemoglobin levels in U.S. hemodialysis and peritoneal populations are similar. However, erythropoietin doses are dramatically higher in hemodialysis patients.

Eating ability in head and neck cancer patients after treatment with chemoradiation: a 12-month follow-up study accounting for dropout.

BACKGROUND: Head and neck cancer patients treated with chemoradiation have difficulty eating a normal diet. This study was designed to characterize eating ability over 12 months after chemoradiation treatment. Analyses take patient dropout into account. METHODS: Two hundred fifty-five patients with head and neck cancer treated with chemoradiation were followed for 12 months. Eating ability was analyzed using generalized linear model methods that accounted for non-ignorable dropout. RESULTS: Eating ability was compromised immediately after treatment and improved over 12 months to near pretreatment levels. Ability to eat at most 50% of the diet orally did not return to baseline levels (p <.05). However, the percent of patients eating a normal diet did return to baseline levels. Accounting for dropout modified the results, but the pattern of significance was similar. CONCLUSIONS: Treatment of head and neck cancer with chemoradiation has a significant effect on eating ability, which improves after 12 months after treatment.