Using fractional polynomials to model the effect of cumulative duration of exposure on outcomes: applications to cohort and nested case-control designs

Association of metformin use and survival in patients with cutaneous melanoma and diabetes

BACKGROUND

Metformin use has been associated with improved survival in patients with different types of cancer, but research regarding the effect of metformin on cutaneous melanoma (CM) survival is sparse and inconclusive.

OBJECTIVES

To investigate the association between metformin use and survival among patients with CM and diabetes.

METHODS

All adult patients with a primary invasive CM between 2007 and 2014 were identified in the Swedish Melanoma Registry and followed until death, or end of follow-up on 31 December 2017 in this population-based cohort study. Patients with both CM and type 2 diabetes mellitus were assessed further. Overall survival (OS) and melanoma-specific survival (MSS) were the primary endpoints. Cox proportional hazard models estimating crude and adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were used comparing peridiagnostic use vs. nonuse of metformin. Dose response was evaluated based on defined daily doses.

RESULTS

Among a total of 23 507 patients, 1162 patients with CM and type 2 diabetes mellitus were included in the final cohort, with a median follow-up time of 4.1 years (interquartile range 2.4-6.1). Peridiagnostic metformin use was associated with a significantly decreased risk of death by any cause (HR 0.68, 95% CI 0.57-0.81). Cumulative pre- and postdiagnostic metformin use was also associated with improved OS: the HR for prediagnostic use was 0.90 (95% CI 0.86-0.95) for every 6 months of use and the HR for postdiagnostic use ranged from 0.98 (95% CI 0.97-0.98) for 0-6 months to 0.59 (0.49-0.70) for 24-30 months of use. No association was found for metformin use and MSS.

CONCLUSIONS

Metformin use was associated with improved OS in patients with CM and diabetes regardless of timing (pre-, post- or peridiagnostic use) and followed a dose-response pattern. However, further research regarding the underlying mechanisms is warranted.

Optimal designs for health risk assessments using fractional polynomial models

Fractional polynomials (FP) have been shown to be more flexible than polynomial models for fitting data from an univariate regression model with a continuous outcome but design issues for FP models have lagged. We focus on FPs with a single variable and construct D-optimal designs for estimating model parameters and I-optimal designs for prediction over a user-specified region of the design space. Some analytic results are given, along with a discussion on model uncertainty. In addition, we provide an applet to facilitate users find tailor made optimal designs for their problems. As applications, we construct optimal designs for three studies that used FPs to model risk assessments of (a) testosterone levels from magnesium accumulation in certain areas of the brains in songbirds, (b) rats subject to exposure of different chemicals, and (c) hormetic effects due to small toxic exposure. In each case, we elaborate the benefits of having an optimal design in terms of cost and quality of the statistical inference.

Characterizing thrombus with multiple red blood cell compositions by optical coherence tomography attenuation coefficient

Embolectomy is one of the emergency procedures performed to remove emboli. Assessing the composition of human blood clots is an important diagnostic factor and could provide guidance for an appropriate treatment strategy for interventional physicians. Immunostaining has been used to identity compositions of clots as a gold-standard procedure, but it is time-consuming and cannot be performed in situ. Here, we proposed that the optical attenuation coefficient of optical coherence tomography (OCT) can be a reliable indicator as a new imaging modality to differentiate clot compositions. Fifteen human blood clots with multiple red blood cell (RBC) compositions from 21% to 95% were prepared using healthy human whole blood. A homogeneous gelatin phantom experiment and numerical simulation based on the Lambert-Beer's law were examined to verify the validity of the attenuation coefficient estimation. The results displayed that optical attenuation coefficients were strongly correlated with RBC compositions. We reported that attenuation coefficients could be a promising biomarker to guide the choice of an appropriate interventional device in a clinical setting and assist in characterizing blood clots.

Levothyroxine dose and risk of atrial fibrillation: A nested case-control study

Contemporary data on the effect of levothyroxine dose on the occurrence of atrial fibrillation (AF) are lacking, particularly in the older population. Our objective was to determine the effect of cumulative levothyroxine exposure on risk of AF and ischemic stroke in older adults.

METHODS

We conducted a population-based observational study using health care databases from Ontario, Canada. We identified adults aged ≥66 years without a history of AF who filled at least 1 levothyroxine prescription between April 1, 2007, and March 31, 2016. Cases were defined as cohort members who had incident AF (emergency room visit or hospitalization) between the date of first levothyroxine prescription and December 31, 2017. Index date was date of AF. Cases were matched with up to 5 controls without AF on the same index date. Secondary outcome was ischemic stroke. Cumulative levothyroxine exposure was estimated based on total milligrams of levothyroxine dispensed in the year prior to index date. Using nested case-control approach, we compared outcomes between older adults who received high (≥0.125 mg/d), medium (0.075-0.125 mg/d), or low (0-0.075 mg/d) cumulative levothyroxine dose. We compared outcomes between current, recent past, and remote past levothyroxine use.

RESULTS

Of 183,360 older adults treated with levothyroxine (mean age 82 years; 72% women), 30,560 (16.1%) had an episode of AF. Compared to low levothyroxine exposure, high and medium exposure was associated with significantly increased risk of AF after adjustment for covariates (adjusted odds ratio [aOR] 1.29, 95% CI 1.23-1.35; aOR 1.08, 95% CI 1.04-1.11; respectively). No association was observed between levothyroxine exposure and ischemic stroke. Compared with current levothyroxine use, older adults with remote levothyroxine use had lower risks of AF (aOR 0.56, 95% CI 0.52-0.59) and ischemic stroke (aOR 0.61, 95% CI 0.56-0.67).

CONCLUSIONS

Among older persons treated with levothyroxine, levothyroxine at doses >0.075 mg/d is associated with an increased risk of AF compared to lower exposure.

Bayesian splines versus fractional polynomials in network meta-analysis

Background

Network meta-analysis (NMA) provides a powerful tool for the simultaneous evaluation of multiple treatments by combining evidence from different studies, allowing for direct and indirect comparisons between treatments. In recent years, NMA is becoming increasingly popular in the medical literature and underlying statistical methodologies are evolving both in the frequentist and Bayesian framework. Traditional NMA models are often based on the comparison of two treatment arms per study. These individual studies may measure outcomes at multiple time points that are not necessarily homogeneous across studies.

Methods

In this article we present a Bayesian model based on B-splines for the simultaneous analysis of outcomes across time points, that allows for indirect comparison of treatments across different longitudinal studies.

Results

We illustrate the proposed approach in simulations as well as on real data examples available in the literature and compare it with a model based on P-splines and one based on fractional polynomials, showing that our approach is flexible and overcomes the limitations of the latter.

Conclusions

The proposed approach is computationally efficient and able to accommodate a large class of temporal treatment effect patterns, allowing for direct and indirect comparisons of widely varying shapes of longitudinal profiles.

A review of the use of time-varying covariates in the Fine-Gray subdistribution hazard competing risk regression model

In survival analysis, time‐varying covariates are covariates whose value can change during follow‐up. Outcomes in medical research are frequently subject to competing risks (events precluding the occurrence of the primary outcome). We review the types of time‐varying covariates and highlight the effect of their inclusion in the subdistribution hazard model. External time‐dependent covariates are external to the subject, can effect the failure process, but are not otherwise involved in the failure mechanism. Internal time‐varying covariates are measured on the subject, can effect the failure process directly, and may also be impacted by the failure mechanism. In the absence of competing risks, a consequence of including internal time‐dependent covariates in the Cox model is that one cannot estimate the survival function or the effect of covariates on the survival function. In the presence of competing risks, the inclusion of internal time‐varying covariates in a subdistribution hazard model results in the loss of the ability to estimate the cumulative incidence function (CIF) or the effect of covariates on the CIF. Furthermore, the definition of the risk set for the subdistribution hazard function can make defining internal time‐varying covariates difficult or impossible. We conducted a review of the use of time‐varying covariates in subdistribution hazard models in articles published in the medical literature in 2015 and in the first 5 months of 2019. Seven percent of articles published included a time‐varying covariate. Several inappropriately described a time‐varying covariate as having an association with the risk of the outcome.

Electronic medical records can be used to emulate target trials of sustained treatment strategies

OBJECTIVE

To emulate three target trials: single treatment vs. no treatment, joint treatment vs. no treatment, and head-to-head comparison of two treatments, we explain how to estimate the observational analogs of intention-to-treat and per-protocol effects, using hazard ratios and survival curves. For per-protocol effects, we describe two methods for adherence adjustment via inverse-probability weighting.

STUDY DESIGN AND SETTING

Prospective observational study using electronic medical records of individuals aged 55-84 with coronary heart disease from >500 practices in the United Kingdom between 2000 and 2010.

RESULTS

The intention-to-treat mortality hazard ratio (95% confidence interval) was 0.90 (0.84, 0.97) for statins vs. no treatment, 0.88 (0.73, 1.06) for statins plus antihypertensives vs. no treatment, and 0.91 (0.77, 1.06) for atorvastatin vs. simvastatin. When censoring nonadherent person-times, the per-protocol mortality hazard ratio was 0.74 (0.64, 0.85) for statins vs. no treatment, 0.55 (0.35, 0.87) for statins plus antihypertensives vs. no treatment, and 1.13 (0.88, 1.45) for atorvastatin vs. simvastatin. We estimated per-protocol hazard ratios for a 5-year treatment using different dose-response marginal structural models and standardized survival curves for each target trial using intention-to-treat and per-protocol analyses.

CONCLUSION

When randomized trials are not available or feasible, observational analyses can emulate a variety of target trials.

A New Model to Predict Benign Histology in Residual Retroperitoneal Masses After Chemotherapy in Nonseminoma

BACKGROUND

Postchemotherapy retroperitoneal lymph node dissection (pcRPLND) is indicated in testicular cancer patients with normalised or plateaued serum tumour markers and residual retroperitoneal lesions >1cm. Challenges remain in predicting postchemotherapy residual mass (pcRM) histology, which may lead to unnecessary surgery.

OBJECTIVE

To develop an accurate model to predict pcRM histology in patients with nonseminomatous germ cell tumours (NSGCTs).

DESIGN, SETTING, AND PARTICIPANTS

A retrospective review of 335 patients undergoing pcRPLND for metastatic NSGCTs to develop a model to predict benign histology in retroperitoneal pcRM. Our model was compared with others and externally validated.

INTERVENTION

Chemotherapy and pcRPLND.

OUTCOME MEASUREMENTS AND STATISTICAL ANALYSIS

Multivariable logistic regression to evaluate the presence of benign histology, and fractional polynomials to allow for a nonlinear association between continuous variables and the outcome. The final Princess Margaret model (PMM) was selected based on the number of variables used, reliability, and discriminative capacity to predict benign pcRM.

RESULTS AND LIMITATIONS

PMM included the presence of teratoma in the orchiectomy, prechemotherapy α-fetoprotein, prechemotherapy mass size, and change in mass size during chemotherapy. Model specificity was 99.3%. Compared with Vergouwe et al's model, PMM had significantly better accuracy (C statistic 0.843 vs 0.783). PMM appropriately identified a larger number of patients for whom pcRPLND can safely be avoided (13.9% vs 0%). Validated in external cohorts, the model retained high discrimination (C statistic 0.88 and 0.80). Larger and prospective studies are needed to further validate this model.

CONCLUSIONS

Our clinical model, externally validated, showed improved discriminative ability in predicting pcRM histology when compared with other models. The higher accuracy and reduced number of variables make this a novel and appealing model to use for patient counselling and treatment strategies.

PATIENT SUMMARY

Princess Margaret model accurately predicted postchemotherapy benign histology. These results might have clinical impact by avoiding unnecessary retroperitoneal lymph node dissection and consequently changing the paradigm of advanced testicular cancer treatment.