Predicting obesity and smoking using medication data: A machine-learning approach

Does machine learning have a high performance to predict obesity among adults and older adults? A systematic review and meta-analysis

AIM

Machine learning may be a tool with the potential for obesity prediction. This study aims to review the literature on the performance of machine learning models in predicting obesity and to quantify the pooled results through a meta-analysis.

DATA SYNTHESIS

A systematic review and meta-analysis were conducted, including studies that used machine learning to predict obesity. Searches were conducted in October 2023 across databases including LILACS, Web of Science, Scopus, Embase, and CINAHL. We included studies that utilized classification models and reported results in the Area Under the ROC Curve (AUC) (PROSPERO registration: CRD42022306940), without imposing restrictions on the year of publication. The risk of bias was assessed using an adapted version of the Transparent Reporting of a multivariable prediction model for individual Prognosis or Diagnosis (TRIPOD). Meta-analysis was conducted using MedCalc software. A total of 14 studies were included, with the majority demonstrating satisfactory performance for obesity prediction, with AUCs exceeding 0.70. The random forest algorithm emerged as the top performer in obesity prediction, achieving an AUC of 0.86 (95%CI: 0.76-0.96; I2: 99.8%), closely followed by logistic regression with an AUC of 0.85 (95%CI: 0.75-0.95; I2: 99.6%). The least effective model was gradient boosting, with an AUC of 0.77 (95%CI: 0.71-0.82; I2: 98.1%).

CONCLUSION

Machine learning models demonstrated satisfactory predictive performance for obesity. However, future research should utilize more comparable data, larger databases, and a broader range of machine learning models.

Association between unstable diabetes mellitus and risk of pancreatic cancer

BACKGROUND

Deterioration of glycaemic control in people with long-standing diabetes mellitus (diabetes) may be a possible indicator of pancreatic cancer. However, the magnitude of the association between diabetes deterioration and pancreatic cancer has received little attention.

METHODS

We conducted a matched cohort study, nested within a population-based cohort of Australian women with diabetes. Women with unstable diabetes, defined as a change in medication after a 2-year period of stable medication use, were matched by birth year to those with stable diabetes, in a 1:4 ratio. We used flexible parametric survival models to estimate hazard ratios (HRs) and 95% confidence intervals (CI).

RESULTS

We included 134,954 and 539,789 women in the unstable and stable diabetes cohorts, respectively (mean age 68 years). In total, 1,315 pancreatic cancers were diagnosed. Deterioration of stable diabetes was associated with a 2.5-fold increased risk of pancreatic cancer (HR 2.55; 95% CI 2.29-2.85). The risk was particularly high within the first year after diabetes deteriorated. HRs at 3 months, 6 months and 1 year were: 5.76 (95% CI 4.72-7.04); 4.56 (95% CI 3.81-5.46); and 3.33 (95% CI 2.86-3.89), respectively. The risk was no longer significantly different after 7 years.

CONCLUSIONS

Deterioration in glycaemic control in people with previously stable diabetes may be an indicator of pancreatic cancer, suggesting investigations of the pancreas may be appropriate. The weaker longer-term (3-7 years) association between diabetes deterioration and pancreatic cancer may indicate that poor glycaemic control can be a risk factor for pancreatic cancer.

Nitrogen-Based Bisphosphonate Use and Ovarian Cancer Risk in Women Aged 50 Years and Older

BACKGROUND

There are few readily modifiable risk factors for epithelial ovarian cancer; pre-clinical studies suggest bisphosphonates could have chemo-preventive actions. Our study aimed to assess the association between use of nitrogen-based bisphosphonate medicine and risk of epithelial ovarian cancer, overall and by histotype.

METHODS

We conducted a case-control study nested within a large linked administrative dataset including all Australian women enrolled for Medicare, Australia's universal health insurance scheme, between July 2002 and December 2013. We included all women with epithelial ovarian cancer diagnosed at age 50 years and older between 1st July 2004 and 31st December 2013 (n = 9,367) and randomly selected up to five controls per case, individually matched to cases by age, state of residence, area-level socioeconomic status, and remoteness of residence category (n = 46,830). We used prescription records to ascertain use of nitrogen-based bisphosphonates (ever use and duration of use), raloxifene and other osteoporosis medicines (non-nitrogen-based bisphosphonates, strontium and denosumab). We calculated adjusted odds ratios (OR) and 95% confidence intervals (CI) using conditional logistic regression.

RESULTS

Ever use of nitrogen-based bisphosphonates was associated with a reduced risk of epithelial ovarian cancer compared to non-use (OR = 0.81, 95%CI : 0.75-0.88). There was a reduced risk of both endometrioid (OR = 0.51, 95%CI : 0.33-0.79) and serous histotypes (OR = 0.84, 95%CI : 0.75-0.93), but no association with the mucinous or clear cell histotypes.

CONCLUSION

Use of nitrogen-based bisphosphonates was associated with a reduced risk of endometrioid and serous ovarian cancer. This suggests the potential for use for prevention, although validation of our findings is required.