Dynamic prediction: A challenge for biostatisticians, but greatly needed by patients, physicians and the public

Prognosis is usually expressed in terms of the probability that a patient will or will not have experienced an event of interest t years after diagnosis of a disease. This quantity, however, is of little informative value for a patient who is still event-free after a number of years. Such a patient would be much more interested in the conditional probability of being event-free in the upcoming t years, given that he/she did not experience the event in the s years after diagnosis, called "conditional survival." It is the simplest form of a dynamic prediction and can be dealt with using straightforward extensions of standard time-to-event analyses in clinical cohort studies. For a healthy individual, a related problem with further complications is the so-called "age-conditional probability of developing cancer" in the next t years. Here, the competing risk of dying from other diseases has to be taken into account. For both situations, the hazard function provides the central dynamic concept, which can be further extended in a natural way to build dynamic prediction models that incorporate both baseline and time-dependent characteristics. Such models are able to exploit the most current information accumulating over time in order to accurately predict the further course or development of a disease. In this article, the biostatistical challenges as well as the relevance and importance of dynamic prediction are illustrated using studies of multiple myeloma, a hematologic malignancy with a formerly rather poor prognosis which has improved over the last few years.

Discussion of “The Evolution of Boosting Algorithms” and “Extending Statistical Boosting”

This article is part of a For-Discussion-Section of Methods of Information in Medicine about the papers “The Evolution of Boosting Algorithms – From Machine Learning to Statistical Modelling” [1] and “Extending Statistical Boosting – An Overview of Recent Methodological Developments” [2], written by Andreas Mayr and co-authors. It is introduced by an editorial. This article contains the combined commentaries invited to independently comment on the Mayr et al. papers. In sub-sequent issues the discussion can continue through letters to the editor.

A concise revised Myeloma Comorbidity Index as a valid prognostic instrument in a large cohort of 801 multiple myeloma patients

With growing numbers of elderly multiple myeloma patients, reliable tools to assess their vulnerability are required. The objective of the analysis herein was to develop and validate an easy to use myeloma risk score (revised Myeloma Comorbidity Index) that allows for risk prediction of overall survival and progression-free survival differences in a large patient cohort. We conducted a comprehensive comorbidity, frailty and disability evaluation in 801 consecutive myeloma patients, including comorbidity risks obtained at diagnosis. The cohort was examined within a training and validation set. Multivariate analysis determined renal, lung and Karnofsky Performance Status impairment, frailty and age as significant risks for overall survival. These were combined in a weighted revised Myeloma Comorbidity Index, allowing for the identification of fit (revised Myeloma Comorbidity Index ≤3 [n=247, 30.8%]), intermediate-fit (revised Myeloma Comorbidity Index 4-6 [n=446, 55.7%]) and frail patients (revised Myeloma Comorbidity Index >6 [n=108, 13.5%]): these subgroups, confirmed via validation analysis, showed median overall survival rates of 10.1, 4.4 and 1.2 years, respectively. The revised Myeloma Comorbidity Index was compared to other commonly used comorbidity indices (Charlson Comorbidity Index, Hematopoietic Cell Transplantation-Specific Comorbidity Index, Kaplan-Feinstein Index): if each were divided in risk groups based on 25% and 75% quartiles, highest hazard ratios, best prediction and Brier scores were achieved with the revised Myeloma Comorbidity Index. The advantages of the revised Myeloma Comorbidity Index include its accurate assessment of patients' physical conditions and simple clinical applicability. We propose the revised Myeloma Comorbidity Index to be tested with the "reference" International Myeloma Working Group frailty score in multicenter analyses and future clinical trials. The study was registered at the German Clinical Trials Register (DRKS-00003868).

Integrating multiple molecular sources into a clinical risk prediction signature by extracting complementary information

BACKGROUND

High-throughput technology allows for genome-wide measurements at different molecular levels for the same patient, e.g. single nucleotide polymorphisms (SNPs) and gene expression. Correspondingly, it might be beneficial to also integrate complementary information from different molecular levels when building multivariable risk prediction models for a clinical endpoint, such as treatment response or survival. Unfortunately, such a high-dimensional modeling task will often be complicated by a limited overlap of molecular measurements at different levels between patients, i.e. measurements from all molecular levels are available only for a smaller proportion of patients.

RESULTS

We propose a sequential strategy for building clinical risk prediction models that integrate genome-wide measurements from two molecular levels in a complementary way. To deal with partial overlap, we develop an imputation approach that allows us to use all available data. This approach is investigated in two acute myeloid leukemia applications combining gene expression with either SNP or DNA methylation data. After obtaining a sparse risk prediction signature e.g. from SNP data, an automatically selected set of prognostic SNPs, by componentwise likelihood-based boosting, imputation is performed for the corresponding linear predictor by a linking model that incorporates e.g. gene expression measurements. The imputed linear predictor is then used for adjustment when building a prognostic signature from the gene expression data. For evaluation, we consider stability, as quantified by inclusion frequencies across resampling data sets. Despite an extremely small overlap in the application example with gene expression and SNPs, several genes are seen to be more stably identified when taking the (imputed) linear predictor from the SNP data into account. In the application with gene expression and DNA methylation, prediction performance with respect to survival also indicates that the proposed approach might work well.

CONCLUSIONS

We consider imputation of linear predictor values to be a feasible and sensible approach for dealing with partial overlap in complementary integrative analysis of molecular measurements at different levels. More generally, these results indicate that a complementary strategy for integrating different molecular levels can result in more stable risk prediction signatures, potentially providing a more reliable insight into the underlying biology.

The nutritional quality of foods carrying health-related claims in Germany, The Netherlands, Spain, Slovenia and the United Kingdom

Backgroung/Objectives:

Compares the nutritional quality of pre-packaged foods carrying health-related claims with foods that do not carry health-related claims.

Subject/Methods:

Cross-sectional survey of pre-packaged foods available in Germany, The Netherlands, Spain, Slovenia and the United Kingdom in 2013. A total of 2034 foods were randomly sampled from three food store types (a supermarket, a neighbourhood store and a discounter). Nutritional information was taken from nutrient declarations present on food labels and assessed through a comparison of mean levels, regression analyses and the application of a nutrient profile model currently used to regulate health claims in Australia and New Zealand (Food Standards Australia New Zealand's Nutrient Profiling Scoring Criterion, FSANZ NPSC).

Results:

Foods carrying health claims had, on average, lower levels, per 100 g, of the following nutrients, energy—29.3 kcal (P<0.05), protein—1.2 g (P<0.01), total sugars—3.1 g (P<0.05), saturated fat—2.4 g (P<0.001), and sodium—842 mg (P<0.001), and higher levels of fibre—0.8 g (P<0.001). A similar pattern was observed for foods carrying nutrition claims. Forty-three percent (confidence interval (CI) 41%, 45%) of foods passed the FSANZ NPSC, with foods carrying health claims more likely to pass (70%, CI 64%, 76%) than foods carrying nutrition claims (61%, CI 57%, 66%) or foods that did not carry either type of claim (36%, CI 34%, 38%).

Conclusions:

Foods carrying health-related claims have marginally better nutrition profiles than those that do not carry claims; these differences would be increased if the FSANZ NPSC was used to regulate health-related claims. It is unclear whether these relatively small differences have significant impacts on health.

Geriatric assessment in multiple myeloma patients: validation of the International Myeloma Working Group (IMWG) score and comparison with other common comorbidity scores

This first validation of the International Myeloma Working Group geriatric assessment in 125 newly diagnosed multiple myeloma patients was performed using the International Myeloma Working Group score based on age, the Charlson Comorbidity Index and cognitive and physical conditions (Activities of Daily Living / Instrumental Activities of Daily Living) to classify patients as fit, intermediate-fit or frail. We verified the International Myeloma Working Group score's impact on outcome, and whether additional tools complement it. Since our prior analyses determined renal, lung and Karnofsky performance impairment as multivariate risks, and the inclusion of frailty, age and cytogenetics complements this, we included the revised myeloma comorbidity index, the Charlson Comorbidity Index, the Hematopoietic Cell Transplantation-Comorbidity Index and the Kaplan-Feinstein Index in this assessment. Multivariate analysis confirmed cytogenetics, Activities of Daily Living, Instrumental Activities of Daily Living and the Charlson Comorbidity Index as risks: 3-year overall survival for fit, intermediate-fit and frail patients was 91%, 77% and 47%, respectively. Using the Charlson Comorbidity Index, the Hematopoietic Cell Transplantation-Comorbidity Index, the Kaplan-Feinstein Index and the revised Myeloma Comorbidity Index allowed us to define fit and frail patients with distinct progression-free and overall survival rates, with the most pronounced differences evidenced via the International Myeloma Working Group score, the Charlson Comorbidity Index and the revised Myeloma Comorbidity Index. Since the Charlson Comorbidity Index is included in the International Myeloma Working Group score, we propose the latter and the revised Myeloma Comorbidity Index for future frailty measurements. Both are useful instruments for identifying myeloma patients with a geriatric risk profile and have a strong prognostic value for functional decline and overall survival. The study was registered as: (clinicaltrials.gov Identifier: 00003686).

Identifying Prognostic SNPs in Clinical Cohorts: Complementing Univariate Analyses by Resampling and Multivariable Modeling

Clinical cohorts with time-to-event endpoints are increasingly characterized by measurements of a number of single nucleotide polymorphisms that is by a magnitude larger than the number of measurements typically considered at the gene level. At the same time, the size of clinical cohorts often is still limited, calling for novel analysis strategies for identifying potentially prognostic SNPs that can help to better characterize disease processes. We propose such a strategy, drawing on univariate testing ideas from epidemiological case-controls studies on the one hand, and multivariable regression techniques as developed for gene expression data on the other hand. In particular, we focus on stable selection of a small set of SNPs and corresponding genes for subsequent validation. For univariate analysis, a permutation-based approach is proposed to test at the gene level. We use regularized multivariable regression models for considering all SNPs simultaneously and selecting a small set of potentially important prognostic SNPs. Stability is judged according to resampling inclusion frequencies for both the univariate and the multivariable approach. The overall strategy is illustrated with data from a cohort of acute myeloid leukemia patients and explored in a simulation study. The multivariable approach is seen to automatically focus on a smaller set of SNPs compared to the univariate approach, roughly in line with blocks of correlated SNPs. This more targeted extraction of SNPs results in more stable selection at the SNP as well as at the gene level. Thus, the multivariable regression approach with resampling provides a perspective in the proposed analysis strategy for SNP data in clinical cohorts highlighting what can be added by regularized regression techniques compared to univariate analyses.

The role of health-related claims and health-related symbols in consumer behaviour: Design and conceptual framework of the CLYMBOL project and initial results

Health claims and symbols are potential aids to help consumers identify foods that are healthier options. However, little is known as to how health claims and symbols are used by consumers in real-world shopping situations, thus making the science-based formulation of new labelling policies and the evaluation of existing ones difficult. The objective of the European Union-funded project Role of health-related CLaims and sYMBOLs in consumer behaviour (CLYMBOL) is to determine how health-related information provided through claims and symbols, in their context, can affect consumer understanding, purchase and consumption. To do this, a wide range of qualitative and quantitative consumer research methods are being used, including product sampling, sorting studies (i.e. how consumers categorise claims and symbols according to concepts such as familiarity and relevance), cross-country surveys, eye-tracking (i.e. what consumers look at and for how long), laboratory and in-store experiments, structured interviews, as well as analysis of population panel data. EU Member States differ with regard to their history of use and regulation of health claims and symbols prior to the harmonisation of 2006. Findings to date indicate the need for more structured and harmonised research on the effects of health claims and symbols on consumer behaviour, particularly taking into account country-wide differences and individual characteristics such as motivation and ability to process health-related information. Based on the studies within CLYMBOL, implications and recommendations for stakeholders such as policymakers will be provided.

Temperaturabhängigkeit der Phosphor-HFS bei Iminophosphoran- Radikalen / Temperature Dependence of ESR-Phosphorus Coupling Constants of Iminophosphorane Radicals

ESR-spectra of iminophosphorane radicals exhibit a strong temperature dependent coupling constant of the phosphorus nucleus. This splitting is described by a superposition of π-σ- and hyperconjugative interactions with the free electron. Because of hindered internal rotation of the iminophosphorane group the hyperconjugation term becomes temperature dependent. Differences in the phosphorus data of ortho- and para-iminophosphorane radicals respectively may be explained by a change of the potential barriers hindering the internal rotation

Lowering the alemtuzumab dose in reduced intensity conditioning allogeneic hematopoietic cell transplantation is associated with a favorable early intense natural killer cell recovery

BACKGROUND AIMS

The anti-CD52 monoclonal antibody alemtuzumab is employed in allogeneic hematopoietic cell transplantation (alloHCT) for the prevention of graft-versus-host disease (GVHD). However, its optimal dosing in this setting has not been determined yet. We compared three different alemtuzumab dose levels in reduced intensity conditioning (RIC) alloHCT with respect to lymphocyte recovery and outcome.

METHODS

In 127 consecutive patients with predominantly advanced stage hematologic malignancies, a first alloHCT after RIC was performed, applying a fludarabine-based protocol (in 93% FBM: fludarabine, bis-chloroethyl-nitrosourea [BCNU], and melphalan). For GVHD prophylaxis, cyclosporine and alemtuzumab at three different dose levels (40 mg, 20 mg, 10 mg) were administered. Recovery of the peripheral blood (PB) lymphocyte sub-populations and clinical outcome were determined with regard to the alemtuzumab dose.

RESULTS

Natural killer (NK) cell concentrations in PB around day +30 correlated inversely with the alemtuzumab dose, whereas other PB lymphocyte subtypes remained essentially unaffected by dosing of alemtuzumab. Lower alemtuzumab doses were associated with a tendency toward improved overall survival mainly during the early post-transplantation months. With regard to the PB NK cell concentration around day +30, "early intense NK cell reconstituters" tended to show an overall survival benefit.

CONCLUSIONS

An alemtuzumab dose reduction to only 10-20 mg provides sufficient GVHD prophylaxis and supports improved NK cell regeneration early after alloHCT in PB ("NK cell saving effect"), which may have a positive effect on overall survival.

Costs of cloud computing for a biometry department. A case study

BACKGROUND

"Cloud" computing providers, such as the Amazon Web Services (AWS), offer stable and scalable computational resources based on hardware virtualization, with short, usually hourly, billing periods. The idea of pay-as-you-use seems appealing for biometry research units which have only limited access to university or corporate data center resources or grids.

OBJECTIVES

This case study compares the costs of an existing heterogeneous on-site hardware pool in a Medical Biometry and Statistics department to a comparable AWS offer.

METHODS

The "total cost of ownership", including all direct costs, is determined for the on-site hardware, and hourly prices are derived, based on actual system utilization during the year 2011. Indirect costs, which are difficult to quantify are not included in this comparison, but nevertheless some rough guidance from our experience is given. To indicate the scale of costs for a methodological research project, a simulation study of a permutation-based statistical approach is performed using AWS and on-site hardware.

RESULTS

In the presented case, with a system utilization of 25-30 percent and 3-5-year amortization, on-site hardware can result in smaller costs, compared to hourly rental in the cloud dependent on the instance chosen. Renting cloud instances with sufficient main memory is a deciding factor in this comparison.

CONCLUSIONS

Costs for on-site hardware may vary, depending on the specific infrastructure at a research unit, but have only moderate impact on the overall comparison and subsequent decision for obtaining affordable scientific computing resources. Overall utilization has a much stronger impact as it determines the actual computing hours needed per year. Taking this into ac count, cloud computing might still be a viable option for projects with limited maturity, or as a supplement for short peaks in demand.

Treatment option of bendamustine in combination with rituximab in elderly and frail patients with aggressive B-non-Hodgkin lymphoma: rational, efficacy, and tolerance

We analyzed the safety and efficacy of rituximab plus bendamustine (R-B) in elderly and frail patients with aggressive B-non-Hodgkin lymphoma (a-B-NHL). Few reports have as yet reported on R-B in a-B-NHL, albeit its value for indolent lymphoma vs. R-CHOP has impressively been shown. We assessed 20 consecutive patients with a-B-NHL receiving R-B as first-line or relapse treatment after (R)-CHOP in our department. Besides patient- and lymphoma-specific characteristics, comorbidity indices were determined. The median patient age was 72 years (51-86), the median Karnofsky performance status was 55 % (40-90 %), and according to the international prognostic index, 15 had high-intermediate or high-risk disease. The comorbidity indices revealed a median Kaplan-Feinstein index of 3 (range 1-3), Charlson comorbidity index of 4 (range 0-9), hematopoietic cell transplantation-specific comorbidity index of 3 (range 0-11), and Freiburg comorbidity index of 2 (range 0-2). Moreover, eight patients had echocardiographic and laboratory signs of cardiac insufficiency, all leading to R-B rather than R-CHOP treatment. The overall response rate was 55 %, with complete response and partial response rates of 20 and 35 %, respectively. In our frail and elderly patient cohort, R-B therapy was well-tolerated. Median progression free survival and overall survival were 8.3 months (95 % confidence interval [CI], 2.8--not reached [n.r.]) and 19.4 months (95 % CI, 4.6--n.r.), respectively. We conclude that R-B is a feasible and safe therapy option in a-B-NHL patients not qualifying for R-CHOP but needs to be further assessed in larger subsequent trials, these currently being under way.

Nutrition labelling - is it effective in encouraging healthy eating?

Abstract

Nutrition labelling has experienced considerable attention in research, combining the expertise as well as general behavioural insights from disciplines such as the social sciences, human medicine and biology, but also business-oriented fields including marketing, communication and general strategy. With a topic as complex as consumption behaviour and its effects on global well-being and health as substantial as they are, many researchers have undertaken studies of quantitative and qualitative nature to explore these relationships. The resulting proliferation of literature has called for reviewing work at many stages, hoping to provide the informational base for actions to be taken on a public health and policy level. Based on such recent changes and developments in the European nutrition labelling landscape, this paper sets out to give an overview on the topic, the reviewing work that has been done so far and the necessary updates on the most recent work in this field. Looking at the question of whether nutrition labelling so far has been effective in encouraging healthy eating, results of this review may be seen as a base for evaluating future achievements of the new EU regulation.

Effects of light availability on leaf gas exchange and expansion in lychee (Litchi chinensis)

Effects of photosynthetic photon flux density (PPFD) on leaf gas exchange of lychee (Litchi chinensis Sonn.) were studied in field-grown "Kwai May Pink" and "Salathiel" orchard trees and young potted "Kwai May Pink" plants during summer in subtropical Queensland (27 degrees S). Variations in PPFD were achieved by shading the trees or plants 1 h before measurement at 0800 h. In a second experiment, potted seedlings of "Kwai May Pink" were grown in a heated greenhouse in 20% of full sun (equivalent to maximum noon PPFD of 200 micromol m(-2)xs(-1)) and their growth over three flush cycles was compared with seedlings grown in full sun (1080 micromol m(-2)xs(-1)). Young potted plants of "Kwai May Pink" were also grown outdoors in artificial shade that provided 20, 40, 70 or 100% of full sun (equivalent to maximum PPFDs of 500, 900, 1400 and 2000 micromol m(-2)xs(-1)) and measured for shoot extension and leaf area development over one flush cycle. Net CO2 assimilation increased asymptotically in response to increasing PPFD in both orchard trees and young potted plants. Maximum rates of CO2 assimilation (11.9 +/- 0.5 versus 6.3 +/- 0.2 micromol CO2 m(-2) s(-1)), dark respiration (1.7 +/- 0.3 versus 0.6 +/- 0.2 micromol CO2 m(-2) s(-1)), quantum yield (0.042 +/- 0.005 versus 0.027 +/- 0.003 mol CO2 mol(-1)) and light saturation point (1155 versus 959 micromol m(-2) s(-1)) were higher in orchard trees than in young potted plants. In potted seedlings grown in a heated greenhouse, shoots and leaves exposed to full sun expanded in a sigmoidal pattern to 69 +/- 12 mm and 497 +/- 105 cm(2) for each flush, compared with 27 +/- 7 mm and 189 +/- 88 cm(2) in shaded seedlings. Shaded seedlings were smaller and had higher shoot:root ratios (3.7 versus 3.1) than seedlings grown in full sun. In the potted plants grown outdoors in 20, 40, 70 or 100% of full sun, final leaf area per shoot was 44 +/- 1, 143 +/- 3, 251 +/- 7 and 362 +/- 8 cm(2), respectively. Shoots were also shorter in plants grown in shade than in plants grown in full sun (66 +/- 5 mm versus 101 +/- 2 mm). Photosynthesis in individual leaves of lychee appeared to be saturated at about half full sun, whereas maximum leaf expansion occurred at higher PPFDs. We conclude that lychee plants can persist as seedlings on the forest floor, but require high PPFDs for optimum growth.

Effects of leaf, shoot and fruit development on photosynthesis of lychee trees (Litchi chinensis)

Changes in gas exchange with leaf age and fruit growth were determined in lychee trees (Litchi chinensis Sonn.) growing in subtropical Queensland (27 degrees S). Leaves expanded in a sigmoid pattern over 50 days during spring, with net CO2 assimilation (A) increasing from -4.1 +/- 0.9 to 8.3 +/- 0.5 micromol m-2 s-1 as the leaves changed from soft and red, to soft and light green, to hard and dark green. Over the same period, dark respiration (Rd) decreased from 5.0 +/- 0.8 to 2.0 +/- 0.1 micromol CO2 m-2 s-1. Net CO2 assimilation was above zero about 30 days after leaf emergence or when the leaves were half fully expanded. Chlorophyll concentrations increased from 0.7 +/- 0.2 mg g-1 in young red leaves to 10.3 +/- 0.7 mg g-1 in dark green leaves, along with stomatal conductance (gs, from 0.16 +/- 0.09 to 0.47 +/- 0.17 mol H2O m-2 s-1). Fruit growth was sigmoidal, with maximum values of fresh mass (29 g), dry mass (6 g) and fruit surface area (39 cm2) occurring 97 to 115 days after fruit set. Fruit CO2 exchange in the light (Rl) and dark (Rd) decreased from fruit set to fruit maturity, whether expressed on a surface area (10 to 3 micromol CO2 m-2 s-1 and 20 to 3 micromol CO2 m-2 s-1, respectively) or on a dry mass basis (24 to 2 nmol CO2 g-1 s-1 and 33 to 2 nmol CO2 g-1 s-1, respectively). Photosynthesis never exceeded respiration, however, the difference between Rl and Rd was greatest in young green fruit (4 to 8 micromol CO2 m-2 s-1). About 90% of the carbon required for fruit growth was accounted for in the dry matter of the fruit, with the remainder required for respiration. Fruit photosynthesis contributed about 3% of the total carbon requirement of the fruit over the season. Fruit growth was mainly dependent on CO2 assimilation in recently expanded dark green leaves.

Shoot development, chlorophyll, gas exchange and carbohydrates in lychee seedlings (Litchi chinensis)

Shoot growth, chlorophyll concentrations, gas exchange and starch concentrations were studied in lychee (Litchi chinensis Sonn.) seedlings of cultivar "Wai Chee" grown in a heated greenhouse at Nambour in subtropical Australia (27 degrees S). We also examined the effects of shoot defoliation and root pruning on leaf expansion. Shoot growth showed a rhythmic cycle under constant greenhouse conditions, with a mean duration of flushing of 20 days and an interval of 10 days over three cycles. Shoots and leaves expanded in a sigmoidal pattern to about 80 mm and 500 cm(2), respectively, for each flush. Starch concentrations of the lower stem and roots decreased as the young red leaves expanded, and increased as the fully expanded leaves turned dark green. Chlorophyll concentrations and net CO(2) assimilation rate were highest in the fully expanded dark green leaves. Removing 50% of the area of each fully expanded leaf had little effect on the expansion of younger leaves, but total biomass of defoliated plants was only 60% of that of controls. In contrast, removing half the roots just before bud swelling reduced final leaf area by 80%. We conclude that the young shoot has relatively low rates of photoassimilation until the leaves are fully expanded and dark green, and depends on assimilates from elsewhere in the plant. During leaf expansion, translocation of assimilates to the shoot occurred at the expense of the roots.