ICSI is one of the most important technologies in reproductive medicine. Injecting a single sperm directly into an oocyte has enabled treatment for many couples affected by severe male-factor infertility. Over time, however, the method has been used far more broadly—including when semen quality is not severely impaired.
Does a more advanced technique automatically increase the chance of having a child? The latest meta-analysis of randomised trials provides a clear answer: not in a population without severe male-factor infertility.
The most important outcome: live birth
A review published in Human Reproduction included six robust randomised trials. Four studies involving 1,438 participants reported live birth. Live birth occurred in 32.8% of couples assigned to ICSI and 34.5% assigned to conventional IVF. The relative risk was 0.96, with a 95% confidence interval of 0.85–1.09.
The cumulative live-birth rate—covering subsequent transfers from a single oocyte retrieval—was 43.2% after ICSI and 47.4% after cIVF. Again, ICSI showed no advantage. Clinical-pregnancy rates were similarly close: 37.7% and 39.4%, respectively.
This does not prove that ICSI is worse in every individual case. It shows that routinely adding ICSI without an appropriate indication does not improve the most important treatment outcome at population level.
Does ICSI prevent total fertilisation failure?
This concern must not be dismissed. A couple undergoes ovarian stimulation, oocyte retrieval and enormous emotional strain, only to be told the next day that none of the oocytes fertilised. This outcome can occur after conventional IVF, but it can also occur after ICSI.
In the meta-analysis, total fertilisation failure occurred in 4.3% of ICSI cycles and 5.0% of cIVF cycles. The difference was not statistically significant. In the subgroup with normal semen or non-severe male-factor infertility, the corresponding rates were 4.0% and 4.5%.
The large INVICSI trial, conducted in six Danish clinics among women starting their first treatment cycle, produced an even more striking finding: TFF occurred in 4.8% after ICSI and 3.7% after conventional IVF. The confidence interval was wide, so we should not infer that ICSI increases the risk. We can, however, state honestly that ICSI was not an “insurance policy” against fertilisation failure in this trial.
What if only a few oocytes are available?
This is particularly important because the loss of each oocyte matters more when the cohort is small. In a randomised trial of 336 cycles among patients aged at least 40 years and/or with no more than four oocytes, the fertilisation rate was higher after conventional IVF: 61.7% versus 53.4% after ICSI.
The denominator must be described precisely. These figures apply to the entire mixed population—“age ≥40 years and/or ≤4 oocytes”—not solely to an isolated low-oocyte subgroup. In the meta-analysis subgroup limited to ≤4 retrieved oocytes, only implantation was reported, in a very small sample: N=42, 7.2% after ICSI and 10.8% after cIVF; RR 0.67 (95% CI 0.37–1.22). A separate fertilisation rate and live-birth rate per couple were not reported for that subgroup.
There is therefore no basis for routinely choosing ICSI merely because few oocytes were retrieved. At the same time, each decision becomes more individual when only a few oocytes are available, because random variation has a greater influence.
A population result is not every couple’s story
The absence of an average ICSI benefit does not mean that every individual couple would have obtained the same outcome with both methods. In one case, conventional IVF may result in total fertilisation failure even though ICSI could have overcome a barrier to zona penetration. In another, the problem may involve oocyte activation and may not be solved simply by injecting a sperm. TFF can also occur after ICSI.
Randomising couples answers the question of which method should routinely be selected for the population. It cannot reveal the counterfactual outcome that the same couple would have experienced with the other method in the same cycle.
Should oocytes be split in the first cycle?
If a sufficient number of mature oocytes is available and there is genuine uncertainty about fertilisation capacity, splitting the cohort can be considered: some oocytes undergo conventional IVF and the remainder ICSI. Such a split may reduce the consequences of choosing only one method and provide information useful in a subsequent cycle.
This conclusion is not directly established by the meta-analysis, however. The authors deliberately excluded studies that randomised individual oocytes rather than couples. Sibling-oocyte studies are useful for comparing fertilisation, but they are less able to answer the question of cumulative live birth for the patient as a whole.
When only a few oocytes are available, a split may be impractical. Dividing four mature oocytes into two and two leaves each method with a very small sample. Failure in one arm may reflect the biology of those particular oocytes rather than the insemination method. Split IVF/ICSI should therefore not be proclaimed a standard for every first cycle.
How should the decision be made?
The first step is to determine whether a genuine indication for ICSI exists: severe male-factor infertility, surgically retrieved sperm, previous total or very low fertilisation, or a specific embryological indication. The next considerations are the number of mature oocytes, treatment history, semen quality on the day of retrieval and the laboratory’s experience.
Patients should be told two truths at the same time:
- routine ICSI without severe male-factor infertility does not, on average, increase the chance of live birth;
- an individual couple may lose a particular cycle if the selected method produces no fertilisation while the alternative potentially could have done so.
Good medicine is neither the automatic use of a more expensive technology nor its indiscriminate rejection for everyone. It requires distinguishing an indication, population-level evidence and individual risk.
Sources
- Kayimu K. et al., Human Reproduction 2026, DOI 10.1093/humrep/deag066.
- Berntsen S. et al., Nature Medicine 2025, DOI 10.1038/s41591-025-03621-x.
- Fancsovits P. et al., J Gynecol Obstet Hum Reprod 2023, DOI 10.1016/j.jogoh.2023.102625.
Educational material; it does not replace an individual medical consultation.
Illustration: an original visual generated with the assistance of AI. It depicts local oocyte membrane damage as a possible technical complication of ICSI; it does not represent the typical course of the procedure or the frequency of this event.