Reading the research

How A Trial Measures Erectile Function: The IIEF, In Plain English

Almost every trial quoted in this aisle reports its result on the same questionnaire, and almost none of the marketing explains what the numbers on it mean. A supplement can beat placebo with a very small p value and still deliver a change nobody would notice. This article explains how the score is built, where the lines fall, and how to read a result like a three-point gain without being carried away by either the good news or the bad.

By HoneyPower Team · Last updated September 26, 2026 · 12 min read · Reading the research

What the questionnaire is

The International Index of Erectile Function, usually shortened to the IIEF, was published in 1997 by Rosen and colleagues to solve a problem. Researchers testing treatments for erectile dysfunction needed a brief, reliable, self-administered measure that worked across cultures and could detect a change caused by treatment. The team drew on a literature search and on interviews with patients and their partners, ran an initial questionnaire past an international panel of experts, and translated and validated the final version in ten languages.

The result was a 15-item questionnaire. A principal components analysis found five factors: erectile function, orgasmic function, sexual desire, intercourse satisfaction and overall satisfaction. Each of the five domains showed high internal consistency, and the scale as a whole did too. Most importantly for a trial, all five domains showed significant changes between baseline and post-treatment scores in men who responded to treatment and not in men who did not, which is what it means for a questionnaire to be sensitive to treatment.

That history explains why it turns up everywhere. It was built for exactly the job that supplement trials also need done. When you meet a phrase like improved erectile function score in a trial abstract, this is nearly always the instrument, or a shortened version of it, standing behind the phrase. The shorter IIEF-5 version, for example, is scored on a 1 to 25 scale in the Cochrane ginseng review discussed below.

The scale and its severity bands

Trials rarely quote all 15 questions. The workhorse is the erectile function domain, a six-question subscale scored from 6 to 30, with higher meaning better function. A 1999 paper by Cappelleri and colleagues asked how that number should be turned into a diagnosis and a severity grade.

They used data from four clinical trials of sildenafil and two control samples, 1,151 men in all, 1,035 with erectile dysfunction and 116 without, all of whom reported attempting sexual activity. A statistical method for finding optimal cutoffs settled on 25: men scoring 25 or below were classified as having erectile dysfunction, and men scoring above 25 as not. At an assumed prevalence of one half, that cutoff had a sensitivity of 0.97 and a specificity of 0.88. The same method then divided severity into five bands, and the agreement between predicted and actual classes was substantial (weighted kappa 0.80).

Bands from the 1999 diagnostic evaluation of the domain. The authors noted that clinical validation against patients' own severity ratings was still warranted.

The bands are useful for one plain reason: they put a number in context. A man who moves from 12 to 15 is still in the moderate band. A man who moves from 21 to 24 has crossed from mild to moderate into mild. The same three points can be a shrug or a step, depending on where you start, and the next section is about exactly that.

The smallest change that counts

Statistical significance answers whether a difference is probably not chance. It does not answer whether the difference is big enough for a man to notice. For years the field had no agreed answer to the second question, until Rosen and colleagues estimated the minimal clinically important difference for the erectile function domain.

They did it with data from 17 randomised, double-blind, placebo-controlled trials of tadalafil, covering 3,345 patients treated for 12 weeks. The anchor was a single question from the same questionnaire, about how often intercourse was satisfactory over the previous four weeks. Using a subset of 11 studies to develop the thresholds and the remaining six to validate them, the receiver operating characteristic method produced an overall figure of 4 points, with a sensitivity of 0.74 and a specificity of 0.73.

The overall figure hides the more useful finding. The threshold varied significantly with how severe the dysfunction was at the start: 2 points for mild, 5 for moderate and 7 for severe. It did not differ by age group, geographic region or cause. A man at the severe end needs a bigger change before it registers as meaningful than a man at the mild end does. That is intuitive once you say it aloud, since a mild score has less far to go.

The authors were careful about their limits. The analysis rested on tadalafil trials, and the abstract states that the results needed to be replicated with other drugs in the same class and in non-drug interventions. The thresholds are a well-built yardstick, not a law of nature.

Erectile function domain, anchor-based estimates from 17 tadalafil trials over 12 weeks.

Two results, read properly

With a scale, some bands and a threshold in hand, take two published results that are relevant to the names on this label and read them the way a careful reviewer would.

Ginseng, pooled by Cochrane

The Cochrane review of ginseng for erectile dysfunction included nine randomised or quasi-randomised trials with 587 men with mild to moderate erectile dysfunction, aged 20 to 70, all compared with placebo, with follow-up of up to 12 weeks. On the erectile function domain of the 15-item questionnaire, ginseng came out ahead by a mean difference of 3.52 points (95 percent confidence interval 1.79 to 5.25), from three studies, with low certainty. Because that is below the threshold of 4 the reviewers assumed, they described the effect as trivial. On the shorter five-item version, the mean difference was 2.39 points (0.89 to 3.88), moderate certainty, against an assumed threshold of 5, which they also called probably trivial.

Look at what that does. The confidence interval excludes zero, so the difference is statistically real. The point estimate is smaller than the smallest change judged meaningful. Both statements are true at once, and a marketing line that reports only the first is not lying, exactly, but it is leaving out the second. The same review found that ginseng may improve men's self-reported ability to have intercourse (risk ratio 2.55, 95 percent confidence interval 1.76 to 3.69, low certainty), which is a different outcome and deserves to be reported as one. The Cochrane authors concluded that ginseng may only have trivial effects on erectile function or satisfaction compared with placebo on validated instruments, based mostly on low certainty evidence.

That is the most recent systematic review of ginseng for this outcome, and it is a good example of why the ginseng dose article is more cautious than the headline of a single trial.

Tribulus, in a larger single trial

The second example is a single trial rather than a pooled one. Kamenov and colleagues randomised 180 men aged 18 to 65 with mild or moderate erectile dysfunction, with or without hypoactive sexual desire disorder, to a tribulus terrestris extract or placebo for 12 weeks. The extract was a Bulgarian product standardised to furostanol saponins, six tablets a day. The primary outcome was the change in questionnaire score, and the difference between tribulus and placebo was 2.70 points (95 percent confidence interval 1.40 to 4.01) in the intention-to-treat population, with a p value below 0.0001. The trial also reported differences in intercourse satisfaction, orgasmic function, sexual desire and overall satisfaction, and no difference in adverse events.

Whether 2.70 points counts as meaningful depends on two things the abstract leaves partly open. The first is the starting severity, because the smallest meaningful change is 2 points for mild and 5 for moderate, and the trial enrolled both. The second is which score was used: the abstract says the questionnaire score without saying whether it was the erectile function domain, the total or a shortened version, and the thresholds above belong to the domain. So the honest reading is that this is a statistically strong result of a size that may or may not clear the bar, and that the comparison is not exact. That kind of caveat is what the whole article is for.

For the saponin standardisation that made this trial what it was, see the article on the tribulus saponin problem.

What the 2026 position statement asks for

The most recent piece of the puzzle is a position statement from the Sexual Medicine Society of North America, written by Trost and colleagues and published in the Journal of Sexual Medicine in 2026. Its starting observation is blunt: to date, no scoring system had adequately defined clinical efficacy of erectile dysfunction therapies when comparing cohorts or pooling studies.

The statement makes six recommendations. It names the erectile function domain as the preferred standardised instrument and the only one adequate for a primary outcome. It says that to claim efficacy, a treatment must show a difference between groups that reaches the minimal clinically important difference, given as median improvements of 2, 5 or 7 points between groups in mild, moderate and severe cohorts, in prospective, randomised, placebo-controlled and adequately powered studies. For mixed-severity populations it proposes a weighted adjustment.

Two further points are directly relevant to anyone reading supplement research. The statement says it would not support efficacy of a therapy based on statistical significance in a meta-analysis alone, and that studies claiming efficacy through responder analyses, meaning a statistically higher percentage of men reaching the threshold, would not be acceptable unless the median difference between groups on the domain itself achieved the threshold.

Neither statement is a rule that binds a supplement seller, and this article does not suggest that any product breaks one. They describe what a rigorous standard looks like, and they give a reader something concrete to compare a result against. Held up against that standard, the ginseng review's own description of its evidence, short follow-up and mostly low certainty, shows the size of the gap, and that is a statement about the research rather than about any individual man's experience.

The caveats that come with the thresholds

It would be a mistake to walk away from this with a rigid pass or fail rule, so here are the honest limits.

• The thresholds were built on tadalafil trials. The authors of the 2011 paper said explicitly that other drugs and non-drug interventions needed to be tested.
• They are group figures. A man can feel a real change at a smaller difference than the group threshold, and another may notice nothing at a larger one.
• The domain is not the whole picture. Desire, satisfaction and self-reported ability to have intercourse are separate outcomes, and the ginseng review found a signal on the last while calling the domain effect trivial.
• Small trials miss things. A trial of 30 men can fail to reach significance on a difference that a trial of 180 would detect.

Why this matters when the label prints no amounts

HoneyPower is a dietary supplement. It names six ingredients, prints no amount for any of them, and the seller positions it as supporting healthy testosterone, blood circulation, energy and stamina. It makes no claim to change a score on any questionnaire, and this article does not suggest that it does.

What the scale gives a reader is a way to weigh the research figures the ingredients page prints beside each name. A trial that reports a gain of a few points is describing a change in a group, on a scale with bands, against a threshold that depends on where the men began. If the trial does not say where they began, which is common, the result cannot be placed at all. If the label does not say how much of the ingredient is in the capsule, a reader cannot tell which trial the product resembles. Those are two separate gaps, and the questionnaire makes both visible.

If you do try the bottle, it is worth writing down where you are before the first capsule, on whatever scale you choose, and looking again at the point the research used rather than at the end of the first week. The results timeline sets out how the seller's typical wait compares with the length of the trials, and the 180-day guarantee is the term that makes a proper look possible.

Four questions to ask of any result

1. How many points, on which scale? Domain, total or short form. The thresholds attach to the domain.
2. Where did the men start? Two points is meaningful in the mild band and small in the severe one.
3. Against what? Placebo, a waiting list or their own baseline. Only the first two answer the question.
4. Significant or meaningful? A very small p value and a change below the threshold can sit in the same abstract.

References