What are potential topics for a retrospective cohort study in endocrinology?

Retrospective Cohort Study Proposals in Endocrinology

High-Priority Topics Based on Current Evidence Gaps

Endocrine Therapy and Cognitive Outcomes in Cancer Survivors

Investigate the long-term cognitive effects of aromatase inhibitors versus tamoxifen in breast cancer survivors, stratified by menopausal status and treatment duration. This addresses a critical quality-of-life outcome where current evidence shows conflicting results regarding cognitive sequelae of endocrine therapy 1. The BMJ guidelines highlight that endocrine therapy impacts cognitive function through estrogen deprivation mechanisms, affecting hippocampal neurogenesis and synaptic plasticity 1. A retrospective cohort comparing cognitive decline rates between patients on exemestane versus tamoxifen, with minimum 5-year follow-up, would fill this evidence gap 1.

Key design elements:

• Compare cognitive assessment scores (if available in medical records) between aromatase inhibitor users and tamoxifen users 1
• Stratify by age at treatment initiation (<45 years vs ≥45 years) given the ASTRRA trial findings 1
• Include ovarian function suppression status as a covariate 1
• Assess for dose-response relationships with treatment duration (5 years vs extended therapy) 1

Premature Ovarian Insufficiency After Cancer Treatment

Examine cardiovascular and bone health outcomes in young women with chemotherapy-induced premature ovarian insufficiency (POI) who received hormone replacement therapy versus those who did not. The Blood Reviews guidelines emphasize that POI following gonadotoxic treatments affects approximately 8-10% of female pediatric cancer survivors, yet optimal hormone replacement strategies remain unclear 1. This study would directly impact mortality (cardiovascular events) and morbidity (osteoporosis, fractures) outcomes 1.

Study parameters:

• Include women diagnosed with POI before age 40 following chemotherapy or hematopoietic stem cell transplantation 1
• Primary outcomes: cardiovascular events, bone mineral density changes, fracture rates 1
• Compare different hormone replacement regimens (17β-estradiol with micronized progesterone vs combined oral contraceptives) 1
• Minimum 10-year follow-up to capture long-term cardiovascular outcomes 1

Diabetes Management and Medication Sequencing

Compare mortality and cardiovascular outcomes between second-line diabetes therapies (SGLT2 inhibitors, GLP-1 receptor agonists, sulfonylureas) added to metformin in patients with type 2 diabetes and chronic kidney disease. The American Diabetes Association emphasizes investigating social determinants and real-world effectiveness of diabetes management strategies 2. Multiple studies demonstrate that sulfonylureas as add-on therapy to metformin increase all-cause mortality (HR 1.44,95% CI 1.12-1.84) and major hypoglycemic episodes (HR 2.78,95% CI 1.66-4.66) compared to other oral agents 3.

Critical design considerations:

• Include patients with eGFR 30-60 mL/min/1.73m² at time of second-line therapy initiation 4
• Primary outcomes: all-cause mortality, major adverse cardiovascular events, heart failure hospitalizations 4
• Compare SGLT2 inhibitors (which reduce heart failure hospitalization by 15% in this population) versus sulfonylureas versus GLP-1 receptor agonists 5, 4
• Use propensity score matching to address confounding by indication 2, 3
• Stratify by baseline cardiovascular disease status 5

Common pitfall to avoid: Failing to account for medication adherence and persistence when evaluating treatment outcomes, as this significantly impacts real-world effectiveness 2.

Thyroid Dysfunction Screening and Outcomes

Evaluate the impact of subclinical hypothyroidism treatment timing on cardiovascular outcomes in adults aged 40-70 years with TSH 5-10 mIU/L. The U.S. Preventive Services Task Force notes that a single retrospective cohort study by Razvi et al. showed levothyroxine treatment in patients aged 40-70 years with subclinical hypothyroidism was associated with lower risk of fatal or nonfatal ischemic heart disease events (4.2% vs 6.6%; HR 0.61,95% CI 0.39-0.95) 1. However, this finding requires validation with better control for confounding variables 1.

Study design improvements over existing evidence:

• Adjust for aspirin and lipid-lowering therapy use (major limitation of prior study) 1
• Require two confirmatory TSH measurements separated by 3 months (addressing TSH variability of up to 50%) 1
• Primary outcomes: ischemic heart disease events, all-cause mortality, cardiovascular mortality 1
• Exclude patients over age 70 where treatment benefit was not demonstrated 1
• Include quality-of-life assessments as secondary outcomes 1

Methodological Considerations for All Proposals

Essential Design Elements

• Clearly define inclusion/exclusion criteria to ensure reproducibility, particularly regarding diagnostic criteria and treatment exposure definitions 2
• Use time-varying covariates for medications and comorbidities that change during follow-up 2
• Implement propensity score matching or inverse probability weighting to minimize selection bias inherent in retrospective analyses 2, 3
• Validate administrative data findings with clinical information from medical records when possible 2

Required Subgroup Analyses

• Sex-specific analyses are essential, as the European Association for the Study of Diabetes recommends examining sex differences in response to glucose-lowering medications 2
• Age stratification (particularly for endocrine therapy studies where outcomes differ by age) 1
• Race/ethnicity subgroups to understand disparities in treatment outcomes 2

Critical Pitfalls to Avoid

• Do not rely on single laboratory values for diagnosis (particularly TSH for thyroid dysfunction, which varies by 40-50% on serial measurements) 1
• Account for immortal time bias when defining treatment exposure periods 2
• Include competing risk analyses when mortality is high, as demonstrated in the heart failure hospitalization study where cause-specific hazards were necessary 4
• Validate diagnostic codes with chart review in a subset of patients, as administrative data may misclassify conditions 2