Optimal PMA Protocol Analysis¶
This analysis explores multiple statistical methods to identify the optimal PMA for first oral feeding. Each method provides complementary evidence for protocol development.
IMPORTANT VALIDATION DISCLAIMER: These findings are derived from the same dataset used to develop the protocol. An independent validation study with a separate cohort is required before clinical implementation. The current analysis should be considered hypothesis-generating rather than confirmatory.
Method 1: Slope Comparison¶
Comparing the relationship between PMA at first feed and PMA at full feed for infants starting before vs. after 35 weeks PMA.
Regression Slopes by Group¶
| Group | N | Slope | SE | p-value |
|---|---|---|---|---|
| < 35 weeks PMA | 166 | 0.66 | 0.17 | 0.0002 |
| ≥ 35 weeks PMA | 62 | 1.23 | 0.21 | 0 |
Slope Difference Test¶
- Interaction term p-value: 0.0319
- Interpretation: Significant difference in slopes between groups
Formal Interaction Test¶
Tests whether the feeding relationship differs significantly between groups. - Interaction coefficient: 0.566 - Interaction p-value: 0.0319 - Model R²: 0.400
Model diagnostics note: Potential issues with multicollinearity, non-normal residuals detected in one or both groups.
Slope Coefficients:¶
- Before 35 weeks: Slope = 0.66 (SE = 0.17)
- After 35 weeks: Slope = 1.23 (SE = 0.21)
- Statistical difference: Slopes are significantly different (p = 0.0319)
Slope Comparison Visualization¶
Visual comparison of feeding progression slopes by initiation timing.
Method 2: Optimal Cutpoint¶
Identifying the PMA cutpoint for first feeding that minimizes overall time to full oral feeding.
Cutpoint Comparison¶
| Cutpoint | N Early | N Late | Mean Days Early | Mean Days Late | Difference | p-value |
|---|---|---|---|---|---|---|
| 32.75 | 23.00 | 205.00 | 20.31 | 15.38 | -4.93 | 0.09 |
| 33.00 | 35.00 | 193.00 | 18.07 | 15.48 | -2.60 | 0.28 |
| 33.25 | 48.00 | 180.00 | 18.88 | 15.08 | -3.80 | 0.08 |
| 33.50 | 68.00 | 160.00 | 16.89 | 15.44 | -1.45 | 0.45 |
| 33.75 | 89.00 | 139.00 | 16.64 | 15.39 | -1.25 | 0.49 |
| 34.00 | 110.00 | 118.00 | 16.37 | 15.42 | -0.96 | 0.58 |
| 34.25 | 121.00 | 107.00 | 15.92 | 15.83 | -0.09 | 0.96 |
| 34.50 | 138.00 | 90.00 | 15.83 | 15.95 | 0.12 | 0.95 |
| 34.75 | 152.00 | 76.00 | 15.59 | 16.46 | 0.87 | 0.64 |
| 35.00 | 166.00 | 62.00 | 15.63 | 16.55 | 0.92 | 0.64 |
| 35.25 | 178.00 | 50.00 | 15.84 | 16.02 | 0.18 | 0.93 |
| 35.50 | 191.00 | 37.00 | 15.53 | 17.67 | 2.14 | 0.37 |
| 35.75 | 197.00 | 31.00 | 15.65 | 17.31 | 1.66 | 0.51 |
How to Read This Table:¶
- Cutpoint: PMA threshold being tested
- N Early/Late: Number of infants starting before/after cutpoint
- Mean Days: Average days to achieve full oral feeding
- Difference: How many MORE days late starters take (negative = early starters take longer)
- p-value: Statistical significance of the difference
Optimal Cutpoint: 35.5 weeks PMA¶
- Statistically minimizes average time to full oral feeding for early group
- Statistical significance: p = 0.3665 (not significant)
Cutpoint Analysis Visualization¶
Comparison of outcomes across different PMA cutpoints.
Method 3: Continuous Optimization¶
Modeling the continuous relationship between PMA at first feed and time to full feed to identify optimal timing.
Optimal Timing Results¶
- Optimal PMA: 35.2 weeks
- Expected days to FOF at optimal PMA: 14.4 days
- Optimal zone: 31.5 - 38.8 weeks PMA
- Model R²: 0.021
Model diagnostics note: Multicollinearity, non-normal residuals detected in polynomial model.
Statistical Results:¶
- Optimal PMA: 35.2 weeks (mathematical minimum from polynomial fit)
- Optimal zone: 31.5 - 38.8 weeks PMA (within 7 days of minimum)
- Model fit: R² = 0.021
- Curve pattern: Polynomial relationship with minimum at 35.2 weeks
Continuous Optimization Visualization¶
Polynomial model showing optimal PMA window for feeding initiation.
Method 4: Stratified Outcomes¶
Comparing feeding outcomes across PMA categories to identify groups with best outcomes.
Outcomes by PMA Category¶
| PMA Category | N | Mean Days | SD Days | Median Days | Mean PMA Full | SD PMA Full |
|---|---|---|---|---|---|---|
| <33 weeks | 35 | 18.07 | 12 | 14.79 | 35.12 | 1.57 |
| 33-34 weeks | 75 | 15.58 | 13.01 | 12.5 | 35.77 | 1.86 |
| 34-35 weeks | 56 | 14.16 | 11.96 | 11.62 | 36.54 | 1.77 |
| >35 weeks | 62 | 16.55 | 14.9 | 12.87 | 38.57 | 2.64 |
Statistical Comparison¶
- ANOVA F-statistic: 0.71
- p-value: 0.5478
- Interpretation: No significant differences between PMA categories
PMA Category Interaction Test¶
Tests whether the feeding relationship differs between early (<34 weeks) and later starters. - Interaction p-value: 0.0097 - Interpretation: Significant difference in feeding relationships
Statistical Results Summary:¶
- ANOVA indicates no significant differences between PMA categories
- F-statistic = 0.71, p-value = 0.5478
- Sample sizes range from 35 to 75 across categories
Stratified Outcomes Visualization¶
Comparison of feeding outcomes across PMA categories.
Methods Summary¶
Multiple analytical approaches applied to identify statistical patterns:
- Slope comparison: Regression analysis of feeding progression rates
- Cutpoint analysis: Systematic testing of PMA thresholds
- Continuous optimization: Polynomial modeling for optimal timing identification
- Stratified comparison: ANOVA across PMA categories
Statistical Methods Summary:¶
- Mathematical optimization techniques applied to identify minimum outcome values
- Statistical significance testing performed across multiple grouping strategies
- Model diagnostics and assumption checking included where available