Open a DNA relationship test report and somewhere in the middle you will find a number labeled with three letters: CPI, CSI, or CRI. That number is the heart of the report. Everything else on the page, including the probability of paternity on the final line, is derived from it. Which three letters you see depends on which test you ordered, and families who order more than one test often get confused when the same lab reports different-looking numbers on different pages.
This guide explains what each of the three indexes measures, why they are mathematically the same thing applied to different biological questions, and how to read the number regardless of which letters label it.
The same math, three different biological questions
All three indexes are likelihood ratios. They compare two scenarios: the probability of observing the DNA markers that showed up on the test IF the claimed biological relationship is true, versus the probability of observing those same markers IF the tested people are unrelated. A likelihood ratio of 10,000 means the observed DNA is 10,000 times more likely under the claimed relationship than under random chance.
The letters just indicate which relationship question the ratio is answering:
- CPI (Combined Paternity Index) on a paternity test. Compares "this man is the biological father" versus "a random unrelated man happens to share these markers with the child."
- CSI (Combined Sibling Index) on a sibling test. Compares "these two children share the same biological father" (or father and mother, for full siblings) versus "two random unrelated children happen to share these markers."
- CRI (Combined Relationship Index) on a grandparent, aunt or uncle, or more general kinship test. Compares "this relationship is biologically true" versus "random unrelated people happen to share these markers."
Different name, same math. The International Society for Forensic Genetics publishes the framework labs across the world use for calculating and interpreting these indexes at ISFG.
The interpretation ranges are the same for all three
Regardless of which letters show up on your report, the thresholds for interpreting the number are identical:
- Above 10: supports the claimed relationship.
- Between 0.1 and 10: inconclusive. The current sample cannot resolve the question.
- Below 0.1: does not support the claimed relationship.
The practical ranges differ. On a clean paternity trio (mother, child, alleged father), CPI values are typically in the millions, which translates to a probability of paternity of 99.99% or higher. CSI and CRI values on confirmed relationships are typically much lower, often in the hundreds to tens of thousands, because the biological signal is weaker on relationships one or two steps removed from direct parentage.
A CSI of 500 and a CPI of 500,000 both support their respective relationships. The absolute numbers are not directly comparable (CSI of 500 is a strong confirmation of a sibling relationship, while CPI of 500 would be unusually low for a paternity trio), but the categorical interpretation is identical.
Why kinship values are typically lower than paternity values
Autosomal DNA is inherited half from each biological parent. A child shares about 50% of their autosomal DNA with each biological parent on average, which is a strong signal that a 23-marker panel resolves with high confidence. The CPI reflects that strong signal.
Second-degree relationships (grandparent, aunt or uncle, half-sibling) share only about 25% of autosomal DNA on average, and the variance around that average is wider. A 23-marker panel can still resolve these relationships, but the resulting CSI or CRI value is naturally lower because the biological signal is weaker.
For a full breakdown of what the markers themselves measure, the guide on what autosomal STR markers are and why paternity tests use them walks through the inheritance math.
Where each index appears on your report
A paternity test report shows the CPI value near the top of the result section, often broken down by individual marker (one row per locus with the per-marker likelihood ratio) and then combined at the bottom into the single CPI number. Below the CPI, most labs display the probability of paternity, derived from the CPI assuming a 50% prior probability. A CPI of 10,000 corresponds to a probability of paternity of 99.99%.
A sibling test report shows the CSI value with the same per-marker breakdown. Sibling reports often include a separate interpretation line stating whether the result supports "full siblings," "half siblings," or "not related" based on the shape of the marker evidence, not just the aggregate CSI value.
A grandparent, aunt or uncle, or general kinship report shows the CRI value with the same per-marker breakdown. Kinship reports typically include an interpretation line like "supports biological grandparent relationship" or "does not support the claimed relationship" based on the CRI value and its position relative to the 10 and 0.1 thresholds.
When a result lands in the inconclusive 0.1 to 10 range
The inconclusive range is far more common on kinship tests than on paternity trios. A clean paternity trio essentially never lands in the inconclusive range. A single-grandparent or single-aunt-uncle test with no maternal sample can land there, especially on borderline biological signals or less-common allele combinations.
USDC's protocol when a kinship test lands between 0.1 and 10 on the standard 23 markers is to reflex-run additional autosomal markers on the same sample, bringing the total up to 28 markers. The reflex step resolves most borderline results. If the reflex result is still in the inconclusive range, the lab report explains which additional relatives, if any, would tighten the analysis. The guide on kinship DNA testing when only one relative is available covers the ranking of which relatives help most in which scenarios.
The probability of paternity line is derived, not independent
Many families focus on the probability of paternity line and treat it as a separate number from the CPI. It is not. The probability of paternity is a mathematical transformation of the CPI, assuming a prior probability of 50% before the test (meaning no outside evidence either way). The formula is:
probability of paternity = CPI / (CPI + 1)
Plug in a CPI of 10,000 and the probability of paternity is 10,000 / 10,001 = 99.99%. Plug in a CPI of 1,000,000 and it rounds to 99.9999%. The probability of paternity never reaches 100% because the formula never does, even with astronomically high CPIs.
Sibling and kinship reports typically do not include a derived probability line because the 50% prior assumption that works for paternity does not translate cleanly to kinship questions, where the prior probability depends on how many other possible relationships are being compared.
What to do with the number once you have it
For most families, the categorical interpretation is all that matters. If the index is above 10, the relationship is supported. If it is below 0.1, it is not. If it is in between, the lab's protocol determines the next step before you even see the final report.
The absolute magnitude of the number can be useful for context (a CPI of 100 million is more decisive than a CPI of 100), but the practical answer for ordering decisions is the same: supported or not supported. For a deeper walk through what each number on a paternity report means in sequence, see the guide on how to read paternity test results.
USDC's home paternity test ($79), home sibling test ($139), and home grandparent test ($139) all report results using the appropriate index for the biological question. The underlying math is the same across all three.
0 comments