Adverse events by booster class — what is common, and when it appears
Answering “how often do side effects happen?” takes numbers. But when we gathered the material, bruising in the same trial, in the same patients, was reported both as 63.2% and as 0.0%. What had changed was neither the product nor the patients but who did the counting. So before laying out the numbers class by class, this piece sets down how to read those numbers. And at the end it leaves the largest limitation in this field exactly as it stands — not one trial has compared the classes directly.
The conclusion, first
Incidence figures are governed more by the method of measurement than by the preparation. In a registration trial in one regulator's approval documents, in the same trial, in the same patients, bruising was 63.2% in the diaries the patients kept themselves and 0.0% in the records the physicians wrote. Swelling was 69.2% against 4.3%, erythema 66.7% against 5.1%. So putting numbers from different studies into one table and reading it as “A is safer than B” is almost always wrong — in fact 17.2% from a prospective trial and 0.048% from post-marketing spontaneous reports are about 350 times apart, and that is not a difference between preparations but a difference in how they were counted. What is confirmed for each class runs like this — the common reactions (bruising · swelling · erythema · pain) generally appear at 1–3 days and are gone within 1–2 weeks, and the rare ones (delayed-onset nodules · granulomas) appear months later. And the most important line in this piece — no two of the seven classes have been compared directly for adverse events in any trial. So “which preparation is safer” cannot be answered on the current evidence; what can be answered instead is which area, and with what instrument.
Same trial, same patients, bruising 63.2% against 0.0%
The regulator's approval documents for a CaHA preparation carry the adverse events of a randomised controlled trial in the nasolabial folds (117 people) in two forms: the diary the patients kept themselves for 14 days, and what the physicians recorded at the visits.
| Adverse event | Patient diary | Physician record |
|---|---|---|
| Bruising | 63.2% | 0.0% |
| Swelling | 69.2% | 4.3% |
| Erythema | 66.7% | 5.1% |
| Pain | 28.2% | 1.7% |
| Nodule | 0.9% | 0.0% |
| Granuloma | 0% | 0% |
The product is the same and the patients are the same. What changed is only who did the counting. In another trial in the same documents (100 people), swelling was 99.0% in the patient diary and 8.0% in the physician record. So when you look at any figure, the first thing to ask is not “what percentage?” but “who counted, and how?”
There are four kinds of denominator — they must not be mixed
| Kind | Character | Typical value |
|---|---|---|
| Prospective trial + patient diary | Actively collected daily. Comes out highest | 60–99% |
| Prospective trial + physician record | Comes out more than tenfold lower even in the same trial | 0–9% |
| Retrospective chart review | Only what was written down. In between | 0.3–3% |
| Post-marketing spontaneous reports | The denominator is an estimate of units sold. Under-reported | 0.0x% |
What the 350-fold difference really is. For PLLA, physician-counted nodules · papules in a prospective trial came to 17.2% (20/116), while for microsphere PCL total adverse events in post-marketing spontaneous reports came to 0.048%. Set side by side, PCL looks 358 times safer. It must not be read that way. The first is a number asked about actively at every visit; the second is what someone reported voluntarily, divided by units sold.
What is confirmed for each class
We put them in one table, but always with the character of the denominator alongside. This table was not made for comparing classes with one another; it is a table showing what has been confirmed for each class and what is empty.
| Class | Confirmed figures | Nature of the material |
|---|---|---|
| Cross-linked HA (BYRYZN) | Local adverse events 92% (control 82%) · tenderness 83 · pain 82 · swelling 71 · erythema 51 · bruising 48 · itching 40 · serious adverse events 0 · all resolved spontaneously within 2 weeks · no nodules over 48 weeks | Prospective · multicentre · randomised · double-blind split-face, n=100, 48 weeks (patient diary) |
| Cross-linked HA (delayed-onset nodules) | 0.33% (7/2,139) | Single-centre retrospective |
| PN (Rejuran) | We did not find an incidence figure | The most widely cited safety paper is co-authored by manufacturer employees · an expert opinion supported by the manufacturer, and that paper itself states that there is no post-marketing surveillance data |
| hADM (Re2O · CellREDM) | Pain 40.7% (control 10.1%) · swelling 45.4% (13.5%) · raised skin temperature 8.1% (1.1%), all p<0.05 · serious adverse events 0 · early reactions resolved within a week | Double-blind · multicentre randomised controlled trial, 175 completers, 6 months. There is no delayed-onset data |
| CaHA (Radiesse · DCLASSY) | Total adverse events 3% (173/5,081 procedures), of which 96% were nodules (166). 49% of the nodules were in areas of high movement | 21 papers pooled, 2,779 people |
| Liquid PCL (GOURI) | There is no incidence data. The published clinical evidence is 2 patients · 12 weeks, both with swelling · erythema lasting under 48 hours | Descriptive case series (n=2) |
| Microsphere PCL | Post-marketing 0.048% · in a 3-year retrospective series of 1,111 procedures, persistent oedema 4.5% · bruising 2.7% · transient lump 0.45%, with 0 nodules · granulomas · infections | Post-marketing spontaneous reports + single-centre retrospective (the denominator is procedures, not patients) |
| PDLLA (Juvelook) | We did not find an incidence figure. That said, a nodule management guideline and a diagnostic · treatment flowchart have been published separately | Case reports · consensus recommendations |
| (Comparator) PLLA | Physician-recorded nodules 8.6% + papules 8.6% = 17.2% (20/116) in a prospective trial | Randomised · multicentre · evaluator-blinded, n=233, 25 months |
The cells in this table most worth looking at are the ones that say “we did not find”. For PN · liquid PCL · PDLLA there is no incidence data at all. And these three classes are the ones talked about in the market as having “almost no side effects”. “There are no reports” is not “it does not happen”. Where there is no surveillance system in particular, no reports is the expected outcome.
A note — among the products we use, for BYRYZN and DCLASSY we did not find peer-reviewed clinical literature searching by product name. That does not mean none exists; it means we could not confirm it, and we have written the class-level material (cross-linked HA · CaHA) in their place.
The timeline — when it appears is what separates one thing from another
“When does it happen” is far more useful than “what happens”. A different time point means a different cause and a different response.
| Time point | What | Basis |
|---|---|---|
| During · immediately after the procedure | Vascular occlusion | Most blindness cases are immediate, during the procedure |
| 1–3 days | Bruising · swelling · erythema · pain | In the CaHA material most adverse events fall on days 1–3 |
| ~1 week | The above resolve / hADM early reactions resolve | The approval documents' “short-term reactions of under 7 days” |
| 1 day–1 week | Acute bacterial infection | PLLA post-marketing material |
| ~2 weeks | Cross-linked HA local reactions all resolve spontaneously | 48-week randomised trial |
| Median 55 days | PLLA papules | Approval documents (median duration 110 days) |
| Median 160 days (about 5 months) | PLLA nodules | Approval documents (median duration 100 days, most resolve spontaneously) |
| Mean 16.2 months (range 3 months–6 years) | Delayed inflammatory nodules generally | Single-centre retrospective, 61 people |
| Up to 2 years | Refractory nodules — cases that went as far as surgical excision | PLLA post-marketing |
The centre of gravity on the time axis differs by class. With cross-linked hyaluronic acid, most of it is immediate and it is settled within 2 weeks. CaHA also clusters at 1–3 days, except that when an adverse event does occur, 96% of them are nodules. PLLA is the reverse, with delayed events as the main event, and nodules appearing at around 5 months. hADM has stronger early reactions than the control but they settle within a week, and there is no delayed-onset data beyond 6 months.
Infection is usually within a week and delayed inflammatory nodules months later, so the timing helps in telling them apart. That said, the range for delayed nodules is wide, from 3 months to 6 years, so timing alone cannot separate them.
hADM — material that runs against the belief that “human-derived means gentle”
Re2O and CellREDM use as their material acellular dermal matrix, human skin with the cells removed. So the expectation that “being human-derived, it will be gentle” is common. The result of the double-blind multicentre randomised controlled trial went the other way.
| Adverse event | hADM (n=86) | Control (n=89) |
|---|---|---|
| Pain | 40.7% (35 people) | 10.1% (9 people) |
| Swelling | 45.4% (39 people) | 13.5% (12 people) |
| Raised skin temperature | 8.1% (7 people) | 1.1% (1 person) |
| Erythema · induration · bruising | The two groups were similar (p>0.05) | |
| Treatment-related serious adverse events | 0 in both groups | |
All three of the items above were statistically significant (p<0.05). The early reactions settled within a week with conservative treatment and there were no serious events. That is, it does not mean the material is dangerous; it means the expectation that it is “gentle” is not supported by the data.
And for this class there is nowhere for adverse events to be reported
Re2O · CellREDM are neither medical devices nor drugs but human tissue under the Act on Safety and Management of Human Tissue, distributed through tissue banks. Two things follow from that.
First, they can reach the market without marketing approval or a clinical trial. That is different from Rejuran or Juvelook, which went through medical-device clinical studies. Second, there is no public database collecting adverse events. The reporting system that exists for medical devices does not exist for this classification. So for this class, “there are no adverse event reports” cannot serve as evidence of safety — because there is nowhere to report them to.
Whether it qualifies as “minimal manipulation” is also under discussion. The developers' side takes the view that the basic structure is preserved and only micronised, while the other side points to overseas criteria under which enzymatic · mechanical breakdown is hard to regard as minimal manipulation, and calls it unclear. The MFDS stated in July 2026 that it would begin reviewing the classification and regulatory framework for this class. When that is settled we will update this paragraph.
Why we use this material and how far the evidence goes is set out separately in how far hADM has actually been confirmed as a material. This piece deals only with the adverse events part of it.
Vascular complications — the area rather than the preparation, and the instrument
The heaviest complication is loss of vision. Pooling the cases reported from 1906 to 2023 gives a cumulative 511, and there is material analysing the 365 from 2018–2023 separately.
| Area | Share | Preparation | Share |
|---|---|---|---|
| Nose | 40.6% | Hyaluronic acid | 79.6% |
| Forehead | 27.7% | Autologous fat | 13.5% |
| Glabella | 19.0% | CaHA | 1.6% |
| Temple · periorbital · cheek and others | The remainder | PLLA | 1.1% |
The two right-hand columns must not be read as a ranking of risk. Hyaluronic acid is at 79.6% because hyaluronic acid is used overwhelmingly more often. Since we do not know how many procedures were done with each preparation (the denominator), comparing risk between preparations is impossible. The distribution by area on the left, by contrast, can be interpreted — the nose · forehead · glabella together make 87.3%, which matches exactly the anatomy in which the vessels of those areas connect to the arteries that go to the eye. “Which area” is an evidence-based division of risk, “which preparation” is not.
Recovery was poor — among the 318 cases with vision data, complete recovery 6.0% · partial improvement 25.8% · no recovery 68.2%. Having some vision remaining at the outset predicted improvement (p<0.001).
Material that has a denominator — needle against cannula
Widen the view to vascular occlusion as a whole and there is material with a denominator. In a retrospective cohort covering 1.7 million syringes over ten years there were 189 vascular occlusions, and they split by instrument like this.
| Needle | 1 / 6,410 syringes (176 cases) |
|---|---|
| Cannula | 1 / 40,882 syringes (13 cases) |
| Difference | Odds 77.1% lower with a cannula |
This is a risk-reduction measure with clearer evidence behind it than choosing a preparation. That said, the authors stated that the material is self-reported and that minor events may be under-reported. An independent, separate prospective dataset also came out similarly, at about 1/6,558 for needles. The criteria for choosing an instrument are set out in needle or cannula.
Whether a late nodule is infection or inflammation — still not separable
For a nodule that appears months later, “is it bacteria or an immune reaction?” is the most important fork in practice. And with the current methods it does not separate well.
In a single-centre dataset of 61 people with late · delayed complications, of the 18 who went to surgery only 7 (38.9%) were culture-positive. The authors stated explicitly that “a negative result does not exclude the diagnosis”. An international consensus body likewise could not conclude whether bacteria are poorly isolated on standard testing because the tests are insensitive or because the pathogen has already been overwhelmed by the immune response.
So the claim that “you can tell infection from inflammation just by palpating and using ultrasound” has no evidence behind it. Even taking tissue and culturing it, more than 60% are negative. In practice we work on the premise that they cannot be told apart, and respond in a direction that covers both possibilities while looking at the timing · warmth · systemic symptoms · course together.
These 61 people are material with a numerator and no denominator — since the total number of procedures over the same period is not available, an incidence cannot be calculated from it. Nodules themselves are covered in more detail in nodules that appear after a procedure.
Why we cannot say “A is safer than B”
We have laid out the numbers class by class up to here, but those numbers cannot be used to rank the classes. The reason is simple.
No two of the seven classes have been compared directly for adverse events in any trial. The closest study we found compared hyaluronic acid alone against hyaluronic acid mixed with CaHA, and it had 20 participants and did not quantify adverse events. A trial comparing PLLA and CaHA directly is registered but has not yet reported results.
So the numbers in the tables in this piece are simply things gathered into one place from different trials · different denominators · different reporters · different eras. 92% and 3% and 0.048% were not measured with the same ruler.
Even so, there are things that can be said. The area (the nose · forehead · glabella are dangerous), the instrument (a cannula has fewer occlusions than a needle), the timing (within 2 weeks and months later are different events), and whether it can be reversed (only the hyaluronic acid family can). These four stand on evidence without any comparison between classes.
In summary — what is confirmed, what we could not confirm
| Category | Content |
|---|---|
| Confirmed | In the same trial · the same patients, bruising 63.2% against 0.0% (depending on the reporter) · cross-linked HA local reactions all resolved spontaneously within 2 weeks, no nodules over 48 weeks · hADM had 4.0 times the pain · 3.4 times the swelling of the control (p<0.05) · CaHA adverse events 3%, of which 96% were nodules · PLLA nodules at a median of 160 days · vascular occlusion needle 1/6,410 against cannula 1/40,882 · 87.3% of blindness cases were nose · forehead · glabella |
| Could not confirm | The incidence for PN · liquid PCL · PDLLA · delayed-onset data beyond 6 months and the safety of repeat injection for hADM · the incidence of skin necrosis · the incidence of biofilm nodules · the vascular complication risk of each preparation (no denominator) · peer-reviewed clinical literature for BYRYZN · DCLASSY |
| Material pointing the other way | hADM had stronger early reactions than the control — the opposite of “human-derived, therefore gentle” · post-marketing 0.048% and prospective 17.2% are 350 times apart, but that is a difference of method, not of preparation |
| Not established | The claim that filler causes an autoimmune syndrome — the evidence on the supporting side is entirely case reports, there is no cohort or incidence data, and there is a separate paper disputing the existence of the concept itself. It cannot be written as established |
| Tried to confirm and could not | On the widely cited claim that “three filler patients had facial swelling in a vaccine trial”, we opened two of the regulator's briefing documents ourselves and did not find any statement relating to filler. That is not an assertion that none exists; it means we could not confirm it |
The one line this piece is meant to leave. Whatever adverse event figure you are looking at, ask two things first — who counted, and what the denominator is. Checking just those two shows that a good many product comparison tables are not comparisons at all.
Frequently asked questions
How often do booster side effects happen?
It cannot be answered with a single number. In the same trial, in the same patients, bruising was 63.2% in the patient diary and 0.0% in the physician record. The common reactions (bruising · swelling · erythema · pain) occur to some degree in almost everyone and are gone within 1–2 weeks. The rare ones (delayed-onset nodules · granulomas) appear months later depending on the class, and the confirmed incidences are about 0.33% for cross-linked hyaluronic acid and 3% of adverse events for CaHA.
I have heard GOURI (liquid PCL) has few side effects.
It has not been confirmed as few; there is simply nothing yet to count. The published clinical evidence for liquid PCL is a case series of 2 patients · 12 weeks, which is all we found. Both had swelling and erythema lasting under 48 hours and there were no other reports. But an incidence cannot be stated from 2 people. On the microsphere PCL side there is a large denominator of 320,000 syringes · 1,111 procedures, but that is material for a different formulation.
I was told Rejuran has no side effects.
Not “none” but “none reported”. And the two are different. The most widely cited safety paper for the PN class is an expert opinion co-authored by manufacturer employees with the writing supported by the manufacturer, and that paper itself states that “there is no post-marketing surveillance data”. Where there is no surveillance system, no reports is the expected outcome. The common local reactions (swelling · pain · bruising · erythema) are reported descriptively, but we did not find a figure with a numerator and a denominator attached.
Re2O and CellREDM come from human skin, so are they gentler?
The data went the other way. In the double-blind multicentre randomised controlled trial, pain 40.7% against 10.1%, swelling 45.4% against 13.5%, raised skin temperature 8.1% against 1.1% — more on the hADM side, and all statistically significant. That said, there were no serious adverse events in either group and the early reactions settled within a week. It does not mean the material is dangerous; it means the expectation that it is “gentle” is not supported by the data.
Where do I report a side effect? Does it differ by product?
It does. Products approved as medical devices (Rejuran · Juvelook · Radiesse · GOURI and so on) have an adverse event reporting system. But Re2O · CellREDM are not medical devices; they are classified as human tissue under the Act on Safety and Management of Human Tissue, and this classification has no public database collecting adverse events. The MFDS stated in July 2026 that it would begin reviewing the classification and regulatory framework for this class.
Which product is the safest?
It cannot be answered on the current evidence. No two of the seven classes have been compared directly for adverse events in any trial. The closest study we found also had 20 participants and did not quantify adverse events. The numbers for each class come from different trials · different denominators · different reporters, so they were not measured with the same ruler. What does stand on evidence without any comparison between classes is four things — the area · the instrument · the timing · whether it can be reversed.
Which products cause the big accidents like blindness?
79.6% of the reported cases were hyaluronic acid, but this must not be read as a ranking of risk — hyaluronic acid is used overwhelmingly more often, and we do not know how many procedures were done with each preparation. What can be interpreted is the area — of the 365 cases, nose 40.6% · forehead 27.7% · glabella 19.0%, three areas making 87.3%. That matches the anatomy in which the vessels of those areas connect to the arteries that go to the eye.
Which is safer, a needle or a cannula?
On vascular occlusion, there is material with a denominator. In a retrospective cohort covering 1.7 million syringes over ten years, it was 1/6,410 for needles and 1/40,882 for cannulas, with the odds 77.1% lower on the cannula side. This is a risk-reduction measure with clearer evidence behind it than choosing a preparation. That said, depending on the area and the plane there are cases where a cannula is the less suitable tool, so we choose the instrument to fit the purpose.
A lump appeared a few months later. Is it an infection?
With the current methods it does not separate well. Of the 18 who went to surgery, 7 (38.9%) were culture-positive, and the authors stated explicitly that a negative does not exclude infection. An international consensus body also left this distinction unresolved. So in practice we work on the premise that they cannot be told apart and respond in a direction that covers both possibilities while looking at the timing · warmth · systemic symptoms · course together. If it suddenly becomes red, hot and painful, please come in quickly.
Is it true that filler causes autoimmune disease?
It is not established. The material claiming so is, as far as we have confirmed, entirely case reports, with no controlled studies and no incidence data. Conversely there is a separate paper disputing the existence of the concept itself (we did not reach the abstract). The core point of contention is the criticism that the diagnostic criteria are non-specific. Neither “this does not happen” nor “it does” can be said on the current evidence.
Who wrote this
Written and reviewed by Lee Chi-Hak, MD, medical director of Miso Clinic in Daegu, South Korea. Every study cited above is given together with its design, its size and the limitations the authors themselves recorded, and where we could not find data, we have said that we could not find any.
| Medical director | Lee Chi-Hak, MD |
|---|---|
| Address | 4F Bombom Building, 125 Dongdeok-ro, Jung-gu, Daegu, South Korea · Exit 1, Kyungpook National University Hospital Station |
| Phone | +82-53-428-2700 |
| Hours | Weekdays 11:00–19:00 (lunch 13:00–14:00) / Saturday 10:00–16:00 (no lunch break) / Closed Sundays and public holidays |
| Columns | All clinical columns |
| Reference library | All booster and device references |
References
- The reporter effect and the CaHA registration trial — the instructions for use in the regulator's approval documents for RADIESSE (PMA P050037). Nasolabial fold randomised controlled trial n=117, 14-day patient diary against physician report. Bruising 63.2% against 0.0%, swelling 69.2% against 4.3%, erythema 66.7% against 5.1%, pain 28.2% against 1.7%, nodules 0.9% against 0.0%, granulomas 0. HIV lipoatrophy trial (n=100) swelling 99.0% against 8.0%. Most adverse events on days 1–3, “short-term, under 7 days”.
- PLLA registration trial — the regulator's approval documents for SCULPTRA (PMA P030050/S002). Randomised · multicentre · evaluator-blinded, n=233 (PLLA 116 / control 117), 13 months + a 25-month extension. Physician-reported nodules 8.6% (10/116) + papules 8.6% (10/116) = 17.2%. Nodule onset median 160 days (mean 209 days) · duration median 100 days, papules median 55 days · duration 110 days. Most resolved spontaneously, with only 1 case receiving intralesional steroid. Post-marketing — nodules at 1–14 months, some persisting 2 years before surgical excision, severe infection at 1 day–1 week.
- Cross-linked hyaluronic acid — Lee SH, Park E, Won CH, Aesthetic Surgery Journal 2026;46(5):569. Prospective · multicentre · randomised · double-blind · split-face non-inferiority, 100 randomised (FAS 93), 48 weeks. Local adverse events 92% against 82%, systemic 47%, serious 0, all local adverse events resolved spontaneously within 2 weeks, no nodules reported over 48 weeks.
- Cross-linked hyaluronic acid delayed-onset nodules 0.33% — single-centre retrospective, 2,139 people (7 cases). The 1 case among 131 people in the skin-booster form (0.8%, on day 122) is from prospective single-arm material.
- hADM — Zhang et al., Aesthetic Plastic Surgery 2025. Double-blind · multicentre · randomised · non-inferiority, 202 randomised (175 completers: hADM 86 / collagen 89), 6 months. Pain 40.7% against 10.1%, swelling 45.4% against 13.5%, raised skin temperature 8.1% against 1.1% (all p<0.05). Erythema · induration · bruising similar in the two groups (p>0.05). Treatment-related serious adverse events 0, early reactions resolved within a week. The split-face trial of a Korean hADM skin booster (n=20, 20 weeks) states only “no serious adverse events, mild transient erythema · swelling”, a description with no numerator · denominator.
- The regulatory status of hADM — classified as human tissue under the Act on Safety and Management of Human Tissue and distributed through tissue banks, and unlike medical devices it has no public adverse event database. There is academic disagreement over whether it qualifies as “minimal manipulation”, and the MFDS stated in July 2026 that it would begin a classification review and an overhaul of the regulatory framework. This paragraph describes the regulatory situation, so we will update it if it changes.
- CaHA — Kadouch JA, Journal of Cosmetic Dermatology 2017. 21 papers pooled, 5,081 procedures · 2,779 people. Total adverse events 173 (3%), of which nodules 166 (96%), persistent inflammation · swelling 4, persistent erythema 2, overcorrection 1. 49% of the nodules were in areas of high movement.
- Microsphere PCL — Lin SL & Christen MO, Journal of Cosmetic Dermatology 2020;19(8):1907. 780 patients · 1,111 procedures, 3 years. Persistent oedema 4.5% · bruising 2.7% · maxillary swelling 0.72% · transient lump 0.45%, with 0 nodules · granulomas · infections · intravascular injections. The denominator is procedures, not patients. The post-marketing material is 155 adverse events in 323,726 syringes (0.048%) — spontaneous reports, so it is captured lower than the reality.
- Liquid PCL — Brito K & Ong D, Annals of Case Reports 2023. 2 patients, 12 weeks. Both had local swelling · erythema lasting under 48 hours, and 1 needed an oral anti-inflammatory and a short course of steroid. No bruising · nodules · vascular occlusion · granulomas reported (but n=2). There is no incidence.
- PN — Rho NK et al., Clinical, Cosmetic and Investigational Dermatology 2026. This is an Expert Opinion and not a systematic review. Some of the authors are manufacturer employees and the rest are manufacturer advisers, and the medical writing was supported by the manufacturer. The limitations the authors stated — the evidence base is limited, more randomised trials and histological verification are needed, and there is no long-term safety registry or post-marketing surveillance data.
- PDLLA — we confirmed that a clinical management guideline for nodules (including a diagnostic · treatment flowchart) has been published in Journal of Cosmetic Dermatology 2025, but we did not reach the full text. We did not find incidence data.
- Vision loss — Doyon VC et al., Aesthetic Surgery Journal 2024;44(10):1091. Systematic review, 365 new cases (2018–2023) · cumulative 511 (1906–2023). Areas: nose 40.6% · forehead 27.7% · glabella 19.0%. Preparations: HA 79.6% · autologous fat 13.5% · CaHA 1.6% · PLLA 1.1%. Vision recovery (n=318) complete 6.0% · partial 25.8% · none 68.2%. This is a pooling of case reports, so it has no denominator and cannot be used to compare risk between preparations. The geographical distribution is also skewed, at 61.9% China · 25.5% Korea.
- Incidence of vascular occlusion — Alam M et al., JAMA Dermatology 2021. Retrospective cohort, 10 years · 1.7 million syringes, 189 occlusions. Needle 1/6,410 (176 cases), cannula 1/40,882 (13 cases), cannula odds 77.1% lower. The limitations the authors stated — self-reported and not audited, with minor occlusions possibly under-reported. An independent prospective dataset came out similarly at about 1/6,558, but the denominator of that dataset is a survey-based extrapolation.
- Delayed complications and telling infection apart — Zhang YL et al., Frontiers in Microbiology 2023;14:1297948. Single-centre retrospective, 61 people with late · delayed complications. Nodules 68.9% · chronic inflammation 24.6% · secondary infection 7.5%. Onset at a mean of 16.2 months (range 3 months–6 years). Of the 18 who went to surgery, 7 were culture-positive. The authors state — “a negative result does not exclude the diagnosis”. An international consensus body (Goodman GJ et al., Aesthetic Surgery Journal 2023, level of evidence 4) also left the distinction between infection and inflammation unresolved.
- Direct comparison between classes — the closest study we found compares hyaluronic acid alone against hyaluronic acid + CaHA mixed (Aesthetic Surgery Journal Open Forum 2025, n=20, 90 days), and it did not quantify adverse events. A trial comparing PLLA against CaHA is registered but no results have been reported.
- Claims relating to autoimmunity — the material on the supporting side is, as far as we have confirmed, entirely case reports · case series, and we did not find controlled studies · cohorts · incidence data. There is a 2017 paper disputing the existence of the concept, but we did not reach the abstract (a secondary citation).
- Tried to confirm and could not — on the widely cited claim that “three people who had had filler developed facial swelling in a vaccine trial”, we read two of the regulator's advisory committee briefing documents directly but did not find any statement relating to filler. What we did confirm was only 3 cases of Bell's palsy against 1 on placebo. That is not an assertion that none exists; it means we could not confirm it.
- For BYRYZN · DCLASSY we did not find peer-reviewed clinical literature searching by product name. The corresponding cells in this piece use class-level material (cross-linked hyaluronic acid · CaHA) in their place.
Everything in this column is general information and does not replace medical diagnosis or treatment. Effects and side effects vary with individual skin condition, age and underlying illness, and the same result is not guaranteed for everyone. Any decision to proceed should be made in an in-person consultation with a physician.
← Clinical columns · Reference library · Miso Clinic home
한국어 · English · 日本語 · 简体中文 · Español · Tiếng Việt · ภาษาไทย · Bahasa Indonesia