Combining peptides: rationale, confounding and study design

Discussions of research peptides often move quickly from single compounds to combinations. If peptide A has promising results and so does peptide B, surely both together should do better? This article explains why that inference does not hold. It covers how combination effects are defined, why they are so difficult to attribute, and which study designs can answer questions about combinations.

This is a methodological article. It does not recommend any combination and contains no dosing, protocols or optimisation advice.

Why combinations are studied

Combination therapy is a standard approach in several areas of medicine. The potential advantages include greater efficacy, lower toxicity and a reduced risk of drug resistance. To claim those advantages, a combination has to be shown to be superior to its components given alone^[1].

There can also be a physiological reason to combine agents. Growth hormone secretion is controlled by growth hormone-releasing hormone (GHRH), growth hormone-releasing peptides (GHRPs) and somatostatin. Giving GHRH and a GHRP together stimulates pulsatile growth hormone secretion markedly, an effect described as synergy. Even this well-characterised interaction varies between individuals. In one study of 47 men, it weakened with age and abdominal visceral fat^[2]. A plausible mechanism is a reason to study a combination, not a result.

Additivity, synergy and antagonism

These terms have precise meanings in pharmacology.

• Additivity is the baseline expectation. Each constituent contributes to the effect in accordance with its own potency^[3].
• Synergy is an effect greater than additivity predicts.
• Antagonism is an effect smaller than additivity predicts.

This means "the combination worked better than either alone" does not establish synergy. Two compounds with purely additive effects will usually outperform each one alone. To detect synergy, researchers need each compound's own dose–response relationship and a reference model for what additivity would look like.

The choice of reference model matters. Effect-based approaches compare the observed combined effect with an expected effect calculated from the individual effects. Dose–effect-based approaches instead compare doses that produce equal effects. No single method suits every situation^[1]. Isobolograms offer a visual assessment but still require formal statistical analysis. The relative potency of two agents need not be constant at every effect level^[3]. The widely used combination index expresses the result as a single number, with values below 1 indicating synergy, 1 additivity and above 1 antagonism^[4]. It relies on its own modelling assumptions.

Apparent benefit can also arise without any interaction at all. Analysing cancer drug combinations, researchers showed that variation between patients, together with each drug acting independently, was enough to explain why many approved combinations outperformed single drugs. In that situation, each patient benefits from whichever drug their tumour responds to best, with no added benefit from the other^[5]. A combination can therefore beat its components in a trial without the drugs enhancing each other.

Confounding and attribution

When two compounds are given together and outcomes are observed without a control group, the result cannot be attributed. The improvement may reflect either compound, both, neither, or factors unrelated to treatment:

• Natural recovery. Many injuries and painful conditions improve over time without treatment.
• Regression to the mean. People often seek treatment when symptoms are at their worst, so some improvement is expected on later measurement.
• Placebo and expectation effects, which are strongest for subjective outcomes such as pain.
• Co-interventions, such as rest, physiotherapy or other medicines introduced at the same time.
• Selection. Those given a combination may differ systematically from those given one compound, or none.

A published example illustrates the problem. In a retrospective review of knee pain treated in one clinic, 4 of 16 patients had received BPC-157 together with thymosin β4. Outcomes were patients' telephone recollections of their pain, with no control group and no standardised measures^[6]. Whatever the result, a design of this kind cannot say what either compound contributed.

The factorial design

The standard design for studying two interventions and their interaction is the factorial trial. In a 2 × 2 factorial study, participants or animals are randomised to one of four groups^[7]:

Only this full comparison can show whether the combined effect exceeds, matches or falls short of what the separate effects predict. However, factorial trials are usually powered to detect the main effect of each intervention. Detecting an interaction reliably requires a much larger sample^[7]. Simulations of interaction tests in randomised trials make the point concrete. There, the interaction was between treatment and a patient subgroup, but the statistics are the same. A trial with 80% power to detect an overall effect had only 29% power to detect an interaction of the same size. Samples needed to be about four times larger to detect it with the same power^[8].

Reporting matters too. A systematic review of factorial trials found that only 8 of 44 were designed to assess the benefit of combining treatments; the rest used the design for efficiency. Only 59% reported testing for interaction. The authors concluded that accurate interpretation depends on reporting the data for every treatment group separately^[9].

Controls and bias

A well-designed combination study needs more than four groups. It also needs the safeguards expected of any experiment:

• Matched vehicle controls, given by the same route and on the same schedule.
• Random allocation and, where possible, blinding of those giving treatment and assessing outcomes.
• Predefined primary outcomes and a sample-size calculation.
• Complete reporting of every group, including negative results.

The ARRIVE 2.0 guidelines set out the minimum information an animal study must report for readers to judge its rigour^[10]. A survey of life-science publications found that measures to reduce bias were reported only to a limited extent^[11]. Combination studies multiply the opportunities for bias, because there are more groups, more comparisons and more ways to choose which result to emphasise.

Dose–response complexity

A single dose of each compound tells us about one point in a much larger space. Interactions can differ with the ratio of the two compounds and with the size of the effect being measured^[3][4]. The arithmetic grows quickly. Testing four dose levels of each compound in every pairing produces 16 combination groups, plus the single-compound and vehicle groups. This is why rigorous combination studies are rare and expensive, and why a single positive result at one dose pairing says little about the combination in general.

Why separate evidence does not add up

Evidence that compound A and compound B each have effects does not establish evidence for A plus B:

1. Interactions are unknown. The combination could be additive, synergistic or antagonistic. Without a direct test, none of these can be assumed.
2. Safety does not transfer. Adverse effects can interact, and exposure to one compound may alter the handling of the other. Separate safety data cannot rule this out.
3. The evidence may not match. Each compound may have been studied in different species, models, tissues and outcomes. Combining results from different contexts produces an inference, not evidence.
4. Evidence quality differs. One compound may have human trial data and the other only animal studies. The combination can be no better supported than its least-studied component, and is often less well supported than either.
5. Physical compatibility is untested. Mixing compounds in one solution raises chemical and physical stability questions that single-compound data do not address.

Summary

Combinations are studied for good reasons, including complementary mechanisms and the hope of greater benefit or less toxicity. Showing a genuine combination effect is demanding. It requires each compound's own dose–response relationship, a clear reference model for additivity, a factorial design with appropriate controls, and a sample large enough to detect an interaction. Uncontrolled reports of combined use cannot attribute outcomes to either compound, and evidence for two compounds studied separately cannot be summed into evidence for the pair.