Our Methodology

TwinCalc v2.0 adds conservative heuristic weights to the dizygotic part of a 1.25% spontaneous-twinning baseline, with a 10% model guardrail. Treatment is flagged but not numerically modelled. Every choice, exclusion and primary source is documented below.

Last updated:

Abstract

TwinCalc v2.0 is an educational model for spontaneous twinning, not a clinical prediction model. It separates a 0.40% monozygotic component from a 0.85% dizygotic component and applies conservative heuristic weights only to the latter: P = min(10%, 0.40% + 0.85% × ∏ weights). We do not have individual-level validation data, calibration curves or confidence intervals; the displayed number should therefore be read as a transparent model output, not a measured personal risk.

Baseline

The 1.25% spontaneous baseline combines an approximately 0.40% monozygotic component with a 0.85% dizygotic calibration component. It is a model choice informed by broad population literature, not the current rate for every country. Published sources use different denominators: twin deliveries, individual twin babies, pregnancies or maternities. Our Twin Birth Data Atlas documents comparable HMBD delivery-level series and explains why population rates cannot be inserted directly into an individual conception estimate.

Maternal age

Evidence. Age is consistently associated with spontaneous dizygotic twinning and multifollicular development. Recent US birth data show a much smaller late-age gradient than the former model assumed.

Effect size. ×0.70 (<25), ×0.95 (25–29), ×1.10 (30–34), ×1.20 (35+)

Weighting. Conservative heuristic applied to the dizygotic component only. Evidence grade: moderate to high.

IVF and assisted reproduction

Evidence. Multiple-birth risk varies materially by treatment type, stimulation protocol, embryo-transfer count, age and clinic. In 2024 the UK IVF multiple-birth rate was 3.2%, while US embryo-transfer outcomes and ASRM guidance use different denominators and protocol groupings.

Effect size. No numerical multiplier in model v2.0.

Weighting. A “yes” answer displays a treatment-specific warning and leaves the spontaneous estimate unchanged.

Family history

Evidence. Familial aggregation and genetic studies support a heritable component for spontaneous dizygotic twinning. The clearest quantified association concerns women with a first-degree female relative who had dizygotic twins.

Effect size. ×2.00 when biological family history is reported.

Weighting. One broad, conservative proxy; no separate paternal-side multiplier. Evidence grade: moderate.

Population and ancestry context

Evidence. Large studies show substantial geographic variation in spontaneous dizygotic twinning, while monozygotic twinning remains closer to 4 per 1,000. Broad labels mix ancestry, geography, environment and measurement practices.

Effect size. ×0.50 (East Asian) to ×1.80 (West African context).

Weighting. Coarse heuristic, never an individual genetic classification. Evidence grade: high for population variation; low for individual prediction.

Height & BMI

Evidence. Observational studies report associations between maternal anthropometry and dizygotic twinning after adjustment for several covariates. Residual confounding and cohort differences remain possible.

Effect size. Height ×0.95–1.20; BMI ×0.95–1.15.

Weighting. Two deliberately compressed heuristics on the dizygotic component. Evidence grade: low to moderate.

Previous pregnancies

Evidence. Parity has been associated with dizygotic twinning in multiple datasets, although recent work describes mixed evidence and age is a stronger predictor.

Effect size. ×1.00 (none) to ×1.15 (three or more).

Weighting. Small compressed heuristic. Evidence grade: low to moderate.

Factors deliberately not weighted

Evidence. Recent hormonal-contraception cessation and breastfeeding were present in model v1. The available evidence is old, small or indirect and does not support stable individual coefficients. A paternal-side-only family-history coefficient is also not independently quantified.

Effect size. ×1.00: contraception, breastfeeding and paternal-side-only history are neutral.

Weighting. Removed from the questionnaire and calculation on 10 July 2026. Evidence grade: insufficient for weighting.

Limitations & disclaimers

  • The model is not trained or validated on individual outcomes and provides no confidence interval.
  • Multiplying the heuristic weights assumes interactions that the available evidence cannot fully estimate.
  • Population/ancestry categories reflect aggregate data and do not identify an individual genetic predisposition.
  • Fertility-treatment effects are excluded. Use protocol- and clinic-specific data instead.
  • The 10% guardrail prevents extreme compounding; it is not a biological threshold.
  • This tool does not replace medical advice. Consult a qualified clinician for personal guidance.

Full bibliography

  1. [1] National Center for Health Statistics. (2026). Twin Birth Rates: United States, 2024.
  2. [2] Beemsterboer SN, et al. (2006). The paradox of declining fertility but increasing twinning rates with advancing maternal age. Human Reproduction.
  3. [3] ASRM Practice Committee. (2022). Multiple gestation associated with infertility therapy.
  4. [4] CDC. (2022 data). ART National Summary.
  5. [5] HFEA. (2026). Fertility treatment 2024: trends and figures.
  6. [6] Smits J, Monden C. (2011). Twinning across the developing world. PLOS ONE.
  7. [7] Painter JN, et al. (2010). A genome-wide linkage scan for dizygotic twinning.
  8. [8] Mbarek H, et al. (2016). Identification of common genetic variants influencing spontaneous dizygotic twinning.
  9. [9] Reddy UM, et al. (2005). Maternal height and body mass index as risk factors for twinning. Obstetrics & Gynecology.

A monthly briefing on twin science

One careful email a month — new factor pages, study updates, no spam.