Positive Outcomes Influence the Rate and Time to Publication, but Not the Impact Factor of Publications of Clinical Trial Results

Pilar Suñé, Josep María Suñé, J. Bruno Montoro•View original

OverviewBalancedalloy voice

Imagine you're a clinician skimming the latest journals before rounds. A shiny new drug looks amazing. It has a big effect, a clean p-value, and it’s published in a top journal. It's tempting to trust that glow. But here's the catch: what you see in print might be a curated slice of reality. That's publication bias — when studies with favorable results are more likely to be written up and published, while those with neutral or unfavorable results stall, hide, or trickle out years later. If what gets published depends on how it turned out, we end up overestimating benefits, underestimating harms, and making worse decisions. To move this from an abstract worry to something you can measure, Suné, Suné, and Montoro took a census of trials passing through a single, very busy ethics committee in Barcelona. They didn't cherry-pick specialties or journals. Instead, they followed everything drug-related that the committee approved from 1997 to 2004 and kept following those trials through early 2010 to see what happened. They investigated not just whether there was a paper, but whether the trial started, was finished, and where the results showed up, if anywhere. This provided a rare, ground-level look at the pipeline from protocol to publication. They were meticulous about what counted as "published." A trial only crossed that line when its global results appeared in a peer-reviewed medical journal. Meeting abstracts, sponsor synopses posted online, and final reports sent back to the ethics committee were logged — useful for finding out what happened — but they didn't count as a publication. That distinction matters, as it separates knowledge that's easy for clinicians and meta-analysts to find from information that lives in obscure or temporary channels. They also defined outcomes in plain, clinical terms. Results were considered positive if the statistics favored the experimental drug; negative if there was no significant difference or if the control did better; and descriptive for non-controlled designs. For equivalence or noninferiority trials — where the goal is to show two options are close enough — a positive result meant they met that standard; a negative result meant they didn't. If a published paper emphasized a different primary outcome than the protocol, they classified the study by what the paper actually led with, reflecting what readers would take away. Matching a given paper to a given trial can be tricky, so they used a tight set of anchors: the drug and its dose and schedule, the condition and population studied, the sample size, the main outcome and how it was analyzed, and the sponsor or funder. If the protocol number, a clinical trials registry ID, or the hospital and principal investigator showed up, that was even better. On the analysis side, they treated time to publication like a survival curve — the "event" was the appearance of a journal article — and they used Kaplan–Meier estimates and Cox regression to compare groups while adjusting for who sponsored the work, the trial phase, sample size, and medical specialty. Here's the landscape they mapped. The backbone of the analysis consists of 785 completed trials. Across that entire group, 380 ended up in peer-reviewed journals — a publication rate of 48 percent. Now, if we loosen the lens to ask, "Could we find the results anywhere?" they could identify outcomes for 541 of those 785. That's better, but it still leaves almost a third of finished trials with no accessible result in any channel they checked. Outcome mattered a lot for whether a trial made it into print. Among the trials where they could find results, 84 percent of positive studies were published in journals, compared with 69 percent of negative ones. Descriptive results were lower still, at 59 percent. The pattern is stark: when a drug looked favorable, it was far more likely to become part of the permanent, citable record. When it didn't, the chances dropped. Speed mattered too. If you're waiting for evidence to update guidelines — or to decide what to put on your hospital's formulary — time is not a trivial detail. Positive trials moved through the pipeline about a year faster. The median time from study completion to journal publication was just over two years for positive results and just over three years for negative results; descriptive studies took closer to four. Put another way, when they modeled the whole dataset — controlling for sponsor, phase, sample size, and specialty — a positive result was associated with nearly double the speed to publication compared with a negative one. The hazard ratio sat at 1.99, and the gap was statistically clear. That acceleration wasn't confined to early, exploratory work. In phase three and four trials — the ones that most directly inform clinical practice — positive results still got to press faster, with a hazard ratio of 2.11. And when the trials involved drugs or uses already approved — the scenarios closest to day-to-day prescribing — the tilt was even steeper, with a hazard ratio of 2.43. Those are the studies you want promptly and fully on the record because they shift real patient care. Yet they, too, were filtered by outcome. Two other features nudged papers along. Non-industry sponsorship — think academic or public funding — correlated with quicker publication, roughly doubling the pace compared with industry-backed trials. Studies with very large sample sizes, more than one thousand participants, also moved faster, with a hazard ratio around 2.5. That makes sense: big trials are often coordinated through networks with publication plans baked in, and their results are inherently newsworthy. Now, you might expect that positive results not only get out more and faster, but also land in fancier journals. The team tested that by looking at journal impact factors, a rough — and imperfect — proxy for journal prestige. Across the 380 published trials, the median impact factor was about 6.4. Comparing positive with negative publications, there wasn't a meaningful difference overall: positive trials had a median impact factor around 6.3, negative ones around 8.3, and the gap wasn't statistically significant. So, once a study cleared the hurdle into the literature, the tier of journal didn't systematically favor positive results. There were exceptions in the details. Descriptive studies tended to appear in lower-impact-factor outlets. And in phase three and four trials, the surprising twist was that negative results landed in higher-impact-factor journals than positives. The median was about 9.9 for negatives and 6.0 for positives in that subset. You can read that two ways: one is that top journals make room for well-designed null results when they matter for practice. The other is that, for less definitive work, a quiet bias in opportunity still leans toward exciting, favorable effects. If you're wondering whether reform efforts — trial registration and prospective protocols — were visible in this cohort, the answer is mixed. Out of nine hundred and forty-five trials in the broader dataset, only two hundred and fifty-nine were registered on the public clinical trials registry. That reflects the era: these studies began in the late 1990s and early 2000s, when registration was gaining steam but not yet universal. The authors searched widely — PubMed, ISI, the Cochrane Library, Spanish databases — and checked sponsor websites and a now-defunct industry results portal. They did what a determined meta-analyst would do. Still, for two hundred and forty-four of the seven hundred and eighty-five completed trials, they couldn't find results anywhere. That's a lot of missing evidence. Where did the studies that did publish end up? All over the map: three hundred and eighty articles spread across one hundred and twenty-five journals. The New England Journal of Medicine led the pack with forty-seven publications, Journal of Clinical Oncology had thirty-eight, The Lancet had twenty-three, and Annals of Oncology had twelve. That spread reinforces the point about impact factor: positive and negative results both found homes in influential outlets once they were written up. Let's pause on the limitations because they sharpen how you interpret all this. This is a single-center cohort from Barcelona, which means local practices and sponsors shape the mix. The team used multiple sources to track results, but some full texts were inaccessible, and matching a paper to a protocol is never perfect. They made careful use of drug, dose, population, outcomes, and sponsor to link records, but mismatches remain possible. The censoring date — searches closed in March 2010 — will miss very late publications. And crucially, they didn't dig into why certain sponsors or investigators chose not to publish. The patterns are clear; the motives are not. Even with those caveats, the takeaways are hard to ignore. First, half of completed trials made it to a journal, which means half did not. Second, among the studies where results could be found, positive outcomes were more likely to be published and reached print roughly a year sooner. Third, once a trial was in the literature, journal prestige didn't generally distinguish positive from negative outcomes — except in specific subgroups — suggesting the bigger filter is the decision to write and submit in the first place. Why does that matter? Because evidence-based medicine is only as good as the evidence base. If neutral or negative trials are slower to publish or never appear, meta-analyses overestimate benefit. Clinicians base choices on a lit set that's brighter than reality. Patients face risks we didn't see coming because the unflattering studies were late or lost. There's a silver lining here. The pattern that Suné and colleagues expose is exactly what trial registration and mandatory results reporting aim to fix. When every trial is registered up front, with outcomes declared prospectively, and every result must be reported within a set window, you decouple dissemination from desirability. Some of that infrastructure was just coming online during their study years. Today, it's stronger, but enforcement is uneven, and cultural habits are sticky. If you're a researcher, the message is plain: write up your null results. If you're an editor, make room for them. And if you're a policymaker or a funder, tie dollars and approvals to timely, complete reporting — not just for the hits, but for the misses, too. A good evidence base isn't a highlight reel. It's the whole season. One last thought on that phase three and four twist, where negative results hit higher-impact journals. I think it tells us the appetite is there when the null really matters — when a big, definitive trial says, "No, this doesn't help," top journals pay attention. That's encouraging. The real work is pushing that same transparency upstream, through the slew of smaller or less glamorous studies where a positive glow still seems to speed the path to publication, or block it when the glow is gone. So the next time you see a gleaming new therapy in print, ask yourself two questions. How fast did the good news arrive? And what are we still not seeing? The Barcelona cohort doesn't preach cynicism. It argues for completeness. And in medicine, completeness is not a luxury. It's the difference between being dazzled and being right.

Imagine you're a clinician skimming the latest journals before rounds. A shiny new drug looks amazing. It has a big effect, a clean p-value, and it’s published in a top journal.

It's tempting to trust that glow. But here's the catch: what you see in print might be a curated slice of reality. That's publication bias — when studies with favorable results are more likely to be written up and published, while those with neutral or unfavorable results stall, hide, or trickle out years later.

If what gets published depends on how it turned out, we end up overestimating benefits, underestimating harms, and making worse decisions.

To move this from an abstract worry to something you can measure, Suné, Suné, and Montoro took a census of trials passing through a single, very busy ethics committee in Barcelona. They didn't cherry-pick specialties or journals. Instead, they followed everything drug-related that the committee approved from 1997 to 2004 and kept following those trials through early 2010 to see what happened.

They investigated not just whether there was a paper, but whether the trial started, was finished, and where the results showed up, if anywhere. This provided a rare, ground-level look at the pipeline from protocol to publication.

They were meticulous about what counted as "published." A trial only crossed that line when its global results appeared in a peer-reviewed medical journal. Meeting abstracts, sponsor synopses posted online, and final reports sent back to the ethics committee were logged — useful for finding out what happened — but they didn't count as a publication. That distinction matters, as it separates knowledge that's easy for clinicians and meta-analysts to find from information that lives in obscure or temporary channels.

They also defined outcomes in plain, clinical terms. Results were considered positive if the statistics favored the experimental drug; negative if there was no significant difference or if the control did better; and descriptive for non-controlled designs. For equivalence or noninferiority trials — where the goal is to show two options are close enough — a positive result meant they met that standard; a negative result meant they didn't.

If a published paper emphasized a different primary outcome than the protocol, they classified the study by what the paper actually led with, reflecting what readers would take away.

Matching a given paper to a given trial can be tricky, so they used a tight set of anchors: the drug and its dose and schedule, the condition and population studied, the sample size, the main outcome and how it was analyzed, and the sponsor or funder. If the protocol number, a clinical trials registry ID, or the hospital and principal investigator showed up, that was even better. On the analysis side, they treated time to publication like a survival curve — the "event" was the appearance of a journal article — and they used Kaplan–Meier estimates and Cox regression to compare groups while adjusting for who sponsored the work, the trial phase, sample size, and medical specialty.

Here's the landscape they mapped. The backbone of the analysis consists of 785 completed trials. Across that entire group, 380 ended up in peer-reviewed journals — a publication rate of 48 percent.

Now, if we loosen the lens to ask, "Could we find the results anywhere?" they could identify outcomes for 541 of those 785. That's better, but it still leaves almost a third of finished trials with no accessible result in any channel they checked.

Outcome mattered a lot for whether a trial made it into print. Among the trials where they could find results, 84 percent of positive studies were published in journals, compared with 69 percent of negative ones. Descriptive results were lower still, at 59 percent.

The pattern is stark: when a drug looked favorable, it was far more likely to become part of the permanent, citable record. When it didn't, the chances dropped.

Speed mattered too. If you're waiting for evidence to update guidelines — or to decide what to put on your hospital's formulary — time is not a trivial detail. Positive trials moved through the pipeline about a year faster.

The median time from study completion to journal publication was just over two years for positive results and just over three years for negative results; descriptive studies took closer to four. Put another way, when they modeled the whole dataset — controlling for sponsor, phase, sample size, and specialty — a positive result was associated with nearly double the speed to publication compared with a negative one. The hazard ratio sat at 1.99, and the gap was statistically clear.

That acceleration wasn't confined to early, exploratory work. In phase three and four trials — the ones that most directly inform clinical practice — positive results still got to press faster, with a hazard ratio of 2.11. And when the trials involved drugs or uses already approved — the scenarios closest to day-to-day prescribing — the tilt was even steeper, with a hazard ratio of 2.43.

Those are the studies you want promptly and fully on the record because they shift real patient care. Yet they, too, were filtered by outcome.

Two other features nudged papers along. Non-industry sponsorship — think academic or public funding — correlated with quicker publication, roughly doubling the pace compared with industry-backed trials. Studies with very large sample sizes, more than one thousand participants, also moved faster, with a hazard ratio around 2.5.

That makes sense: big trials are often coordinated through networks with publication plans baked in, and their results are inherently newsworthy.

Now, you might expect that positive results not only get out more and faster, but also land in fancier journals. The team tested that by looking at journal impact factors, a rough — and imperfect — proxy for journal prestige. Across the 380 published trials, the median impact factor was about 6.4.

Comparing positive with negative publications, there wasn't a meaningful difference overall: positive trials had a median impact factor around 6.3, negative ones around 8.3, and the gap wasn't statistically significant. So, once a study cleared the hurdle into the literature, the tier of journal didn't systematically favor positive results.

There were exceptions in the details. Descriptive studies tended to appear in lower-impact-factor outlets. And in phase three and four trials, the surprising twist was that negative results landed in higher-impact-factor journals than positives.

The median was about 9.9 for negatives and 6.0 for positives in that subset. You can read that two ways: one is that top journals make room for well-designed null results when they matter for practice. The other is that, for less definitive work, a quiet bias in opportunity still leans toward exciting, favorable effects.

If you're wondering whether reform efforts — trial registration and prospective protocols — were visible in this cohort, the answer is mixed. Out of nine hundred and forty-five trials in the broader dataset, only two hundred and fifty-nine were registered on the public clinical trials registry. That reflects the era: these studies began in the late 1990s and early 2000s, when registration was gaining steam but not yet universal.

The authors searched widely — PubMed, ISI, the Cochrane Library, Spanish databases — and checked sponsor websites and a now-defunct industry results portal. They did what a determined meta-analyst would do. Still, for two hundred and forty-four of the seven hundred and eighty-five completed trials, they couldn't find results anywhere. That's a lot of missing evidence.

Where did the studies that did publish end up? All over the map: three hundred and eighty articles spread across one hundred and twenty-five journals. The New England Journal of Medicine led the pack with forty-seven publications, Journal of Clinical Oncology had thirty-eight, The Lancet had twenty-three, and Annals of Oncology had twelve.

That spread reinforces the point about impact factor: positive and negative results both found homes in influential outlets once they were written up.

Let's pause on the limitations because they sharpen how you interpret all this. This is a single-center cohort from Barcelona, which means local practices and sponsors shape the mix. The team used multiple sources to track results, but some full texts were inaccessible, and matching a paper to a protocol is never perfect.

They made careful use of drug, dose, population, outcomes, and sponsor to link records, but mismatches remain possible. The censoring date — searches closed in March 2010 — will miss very late publications. And crucially, they didn't dig into why certain sponsors or investigators chose not to publish. The patterns are clear; the motives are not.

Even with those caveats, the takeaways are hard to ignore. First, half of completed trials made it to a journal, which means half did not. Second, among the studies where results could be found, positive outcomes were more likely to be published and reached print roughly a year sooner.

Third, once a trial was in the literature, journal prestige didn't generally distinguish positive from negative outcomes — except in specific subgroups — suggesting the bigger filter is the decision to write and submit in the first place.

Why does that matter? Because evidence-based medicine is only as good as the evidence base. If neutral or negative trials are slower to publish or never appear, meta-analyses overestimate benefit.

Clinicians base choices on a lit set that's brighter than reality. Patients face risks we didn't see coming because the unflattering studies were late or lost.

There's a silver lining here. The pattern that Suné and colleagues expose is exactly what trial registration and mandatory results reporting aim to fix. When every trial is registered up front, with outcomes declared prospectively, and every result must be reported within a set window, you decouple dissemination from desirability.

Some of that infrastructure was just coming online during their study years. Today, it's stronger, but enforcement is uneven, and cultural habits are sticky.

If you're a researcher, the message is plain: write up your null results. If you're an editor, make room for them. And if you're a policymaker or a funder, tie dollars and approvals to timely, complete reporting — not just for the hits, but for the misses, too. A good evidence base isn't a highlight reel. It's the whole season.

One last thought on that phase three and four twist, where negative results hit higher-impact journals. I think it tells us the appetite is there when the null really matters — when a big, definitive trial says, "No, this doesn't help," top journals pay attention. That's encouraging.

The real work is pushing that same transparency upstream, through the slew of smaller or less glamorous studies where a positive glow still seems to speed the path to publication, or block it when the glow is gone.

So the next time you see a gleaming new therapy in print, ask yourself two questions. How fast did the good news arrive? And what are we still not seeing?

The Barcelona cohort doesn't preach cynicism. It argues for completeness. And in medicine, completeness is not a luxury. It's the difference between being dazzled and being right.

Related lectures