Can All Oxygen Molecules Accumulate in One Corner of the Room? Treating Pathological Narcissism Requires High Standard Randomized Clinical Trials. A Commentary on the Trial by Mohajerin et al.

ABSTRACT

Narcissistic personality disorder (NPD) and pathological narcissism (PN) urgently need empirically supported treatment options. Mohajerin et al. randomized controlled trial yielded promise in addressing this gap. However, in this commentary, an international group of personality disorder experts identify a number of issues with the trial methodology, reporting and interpretation: (i) the unlikeliness that consecutive sample recruitment would result in almost equal number of participants in grandiose vs. vulnerable subgroups; (ii) the unlikeliness that consecutive sample recruitment would result in almost equal number of participating men and women; (iii) the unlikely possibility that grandiose and vulnerable narcissism would appear as uncorrelated subtypes; (iv) dissimilar treatment responses to schema‐focused therapy vs. Unified Protocol; (v) absence of dropouts; (vi) unclear randomization procedure; (vii) poor screening and recruitment procedures and limited transparency in trial registration; (viii) exclusive reliance on self‐report measures; (ix) limited transparency in supervision and treatment fidelity procedures; (x) nonequivalence of treatment dose and duration; and (xi) lack of clarity on how the disclosure of NPD diagnosis to participants was handled in the trial. In light of these concerns, it is difficult to reconcile the results of this trial with the inherent challenges of treating NPD and PN. We therefore warrant against considering it a benchmark in NPD treatment innovation.

Summary

• Narcissistic personality disorder and pathological narcissism are treatable, but still require methodologically solid randomized controlled trials.

In statistical physics, there is a well‐known paradox, known as Loschmidt's paradox (Loschmidt 1876), namely, given the random movement of oxygen molecules, it is statistically possible that all oxygen molecules would accumulate on one side of the room. While it is possible, it has never been observed. The scenario of all oxygen molecules accumulating in one corner of the room calls for urgent attention. However, the impossibility of such outcome invites one to ignore it. In this commentary, we address several concerns related to the trial published by Mohajerin et al. (2026) all of which appear to show such phenomena despite their unlikely nature.

Pathological narcissism (PN), including NPD, presents a significant public health concern due to elevated risk of co‐occurrence of psychiatric disorders, persistence and treatment resistance of comorbid disorders, elevated risk of suicide as well as legal, vocational and marital problems and an increased risk of inflicting emotional distress to others (Weinberg and Ronningstam 2020). In the last two decades, there has been an exponential growth of psychological and psychiatric research regarding the nature of PN. Several psychological therapies have been developed for patients with PN, and yet, until recently, there has been an unfortunate absence of randomized clinical trials (RCTs) assessing the beneficial and harmful effects of psychotherapy for narcissistic personality disorder (NPD) or PN. The absence of such trials has been acknowledged for several decades (e.g., Dhawan et al. 2010) and is in stark contrast to numerous RCTs for closely associated disorders such as BPD (for a review, see Storebø et al. 2020) or ASPD (Fonagy et al. 2025; for a review, see Gibbon et al. 2020). Thus far, the treatability of patients with NPD has been supported by clinical opinions of experts (Dimaggio 2022; Weinberg and Ronningstam 2020; Weinberg 2024; Dimaggio and Weinberg 2024), not empirical findings. Consequently, many experts continue to uphold the unsupported notion that patients with NPD cannot be effectively treated (Finch and Mellen 2025). The absence of formal scientific evidence is related to several factors, including the following: (i) NPD presents unique treatment challenges, leading researchers to study other disorders; (ii) NPD patients tend to leave treatment prematurely—yet another barrier to trial completion; (iii) a large proportion of NPD patients are not treatment seeking, posing challenges to recruitment of the patients and to the representativeness of treatment seeking NPD samples; (iv) NPD patients tend to evoke strong negative feelings in their therapists, thus posing challenges in the selection and retention of trial therapists; (v) the process of randomization may evoke difficult reactions in NPD patients (e.g., due to their sensitivity to rejection and need for control), thus challenging feasibility of RCTs in this population (Weinberg et al. 2024). There is an urgent need for research on NPD and PN as these are conditions with significant prevalence and personal psychological pain with some seeking effective treatment, although often for comorbid disorders. Others come to treatment with time‐limited motivation due to external pressure, such as current crises or an ultimatum from significant others (Weinberg 2024). Other challenges are (i) the need to incorporate interview and functional measures, in addition to self‐report assessments, (ii) fluid and context‐dependent nature of expression of PN and (iii) discontinuity between symptom measures of change and dimensional and personality trait measures (Vater et al. 2014; Ronningstam and Weinberg 2023).

A recent RCT by Mohajerin et al. (2026) concluded that NPD can be treated. At first glance, that could be seen as a welcome advancement in the field. However, upon closer examination of the trial, several serious concerns emerge regarding its credibility, internal validity and the interpretability of its findings, which prompted this group of experts in NPD and personality disorders in general to provide a detailed commentary. Specifically, the trial appears to present multiple threats to internal validity, construct validity and clinical plausibility, which question the validity of the results. The reasons for these concerns are outlined below.

1. Frequencies of Grandiose and Vulnerable Narcissism

The authors report that their sampling strategy required all consecutively recruited participants to meet criteria for NPD. Thereafter, the participants were divided into two groups—grandiose and vulnerable, based on self‐report measures of these constructs. The authors reported an almost equal number of participants in these groups (N [grandiose] = 56, N [vulnerable] = 58). We find these groups atypical compared to NPD patients described in previous literature. In particular, the NPD diagnosis is considered by clinicians and researchers as an ultimate description of grandiose narcissism. Research shows that grandiose narcissism correlates with NPD, whereas vulnerable narcissism is associated with other personality disorders (Dickinson and Pincus 2003; Stanton and Zimmerman 2018). It would be expected that including only patients that meet criteria for NPD would make the sample predominantly grandiose, and vulnerable narcissism would be underrepresented in the sample. The almost 50/50 distribution of grandiose/vulnerable division of Mohajerin et al.'s (2026) sample is very different from all previous publications in the field. This calls into question the generalizability of the results reported for this sample but also suggests patients identified with vulnerable narcissism in this study are not representative of the majority of patients with vulnerable narcissism (since many of them do not meet criteria for NPD, but rather for other disorders). Beyond generalizability concerns, the group sizes invite a re‐evaluation of recruitment strategy as consecutive recruitment is unlikely to produce near‐perfect group sizes.

2. Gender Distribution

We have similar concerns regarding the gender distribution of grandiose and vulnerable narcissism. The literature reports that, whereas vulnerable narcissism is relatively equally distributed among men and women, grandiose narcissism is more prevalent among men (for a review, see Grijalva et al. 2015). However, the randomly selected sample in the trial included a relatively equal number of men and women in both groups of narcissism, raising questions about the representativeness of this sample, as well as concerns regarding diagnostic validity, construct operationalization and recruitment strategy.

3. Grandiose and Vulnerable Narcissism as Fully Separate Subtypes

An even more serious concern arises regarding the notion of grandiose and vulnerable subgroups. The authors selected individuals that met categorical criteria for NPD and further divided the sample into grandiose and vulnerable narcissism. In doing so, the authors viewed grandiose and vulnerable narcissism as mutually exclusive conditions or constructs. However, this is inconsistent with results of correlational and ecological assessment studies that find that grandiose and vulnerable narcissism correlate and co‐occur in individuals high in grandiose narcissism—such as clinical NPD (Jauk et al. 2022; Gore and Widiger 2016; Edershile and Wright 2021; Edershile et al. 2024). Yet in some individuals, vulnerable narcissism does not correlate with grandiose narcissism; these people have vulnerable narcissism without grandiose traits (Jauk et al. 2022). Thus, people with grandiose narcissism are very likely to display high levels of vulnerable narcissism, making it highly unlikely to meet the study's categorical selection strategy for grandiose vs. vulnerable subgroups, since many individuals may meet criteria for both groups. Similarly, those individuals who are predominantly vulnerable and do not have high grandiose traits are a minority. This finding makes the almost 50/50 distribution between grandiose and vulnerable groups highly unlikely. Taken together, these considerations highlight that the Mohajerin et al. describe a data set that does not fit what we know about PN, including the dimensional and often fluctuating nature of narcissistic grandiosity and vulnerability.

4. Dissimilar Response to Schema‐Focused Therapy (SFT) vs. Unified Protocol (UP)

With such interrelationship between grandiose and vulnerable facets of PN, one would expect far less pronounced differential effects of treatments, unlike the trial findings. It would be anticipated that the correlation between grandiose and vulnerable manifestations would most likely make response to SFT and UP similar. However, the marked differential effect of these treatments for these manifestations of narcissism seems puzzling as it suggests that these manifestations of narcissism are more dissimilar than they are in reality. This further underscores concerns about the sample, given the overlap between vulnerability and grandiosity, and the response of these patients to treatment.

5. Absence of Dropouts

Another striking finding is the complete absence of treatment dropout in both groups. In psychotherapy research—especially in personality disorders and narcissistic pathology—dropout rates are typically substantial, often exceeding 40%–60%. The absence of any attrition raises serious concerns regarding (1) sample selection procedures; (2) reporting transparency; or (3) data completeness. This issue is further amplified by the absence of reported treatment‐interfering behaviours, which are widely documented as central to the treatment of PN. The zero dropout rate is an improbable finding for any treatment outcome study and even more atypical for a sample of NPD patients, since these patients are known to have high dropout rate (63%–64%; Hilsenroth et al. 1998; Gamache et al. 2018). It is unclear how the researchers accomplished 100% retention in a sample. Along the same lines, treatments of patients with PN, including NPD, commonly involve challenges such as treatment‐interfering behaviours and impasses (Weinberg and Ronningstam 2020; Weinberg 2024; Kernberg 2007). The current trial documented no such reactions.

The zero dropout rate is even more puzzling since they utilized an unmodified UP—a treatment developed to address neuroticism (Longley and Gleiser 2023; Sauer‐Zavala and Barlow 2021) rather than PN. Although vulnerable narcissism tends to be highly correlated with measures of neuroticism, it also correlated with antagonism (Crowe et al. 2019; Miller et al. 2018). Indeed, antagonism has been observed across both grandiose and vulnerable narcissism and may account for many of the commonly described treatment challenges with these patients (Weinberg and Ronningstam 2020; Weinberg 2024; Kernberg 2007). Antagonism, along with elevated use of maladaptive self‐regulatory behaviours, increases the probability of unmodified treatments resulting in treatment challenges such as stalled or incomplete treatment, irregular attendance, reduced adherence to in‐session and homework tasks and diminished therapeutic alliance. To address this challenge, utilization of treatments developed for other disorders among patients with NPD have commonly required adaptations of the original protocols. This has been a common practice and resulted in such adaptations as TFP for NPD (Diamond et al. 2021), MBT for NPD (Drozek et al. 2023; Choi‐Kain et al. 2022) or GPM for NPD (Weinberg et al. 2019). When unmodified protocols are utilized with patients with NPD, they typically evoke challenging reactions and increase the risk of non‐compliance and treatment discontinuation. Surprisingly, no modifications of the UP were implemented, but neither were treatment challenges nor dropouts reported.

6. Unclear Randomization Procedure

Regarding the randomization procedure in the trial, the authors only write that they ‘randomly assigned participants to either the UP or SFT condition based on a 1:1 allocation ratio’, but the authors do not mention how the allocation sequence was generated, or if it was concealed from the investigators. The lack of such information raises concerns, as readers can question if the trials were in fact adequately randomized. Additional information about the randomization procedure is warranted.

7. Poor Screening and Recruitment Procedures and Limited Transparency in Trial Registration

A key requirement for strengthening the credibility of psychotherapy research is prospective preregistration, which helps prevent issues such as selective outcome reporting and publication bias (Inchausti et al. 2026). In the present trial, registration appears to have been conducted retrospectively, which limits confidence in the transparency and robustness of the analytic strategy. Of note, none of the authors of this commentary could find a public record of the registration.

In addition, reporting lacks a comprehensive CONSORT flow diagram detailing the full trajectory of participants—from initial contact and screening through allocation, treatment and follow‐up. This omission makes it difficult to adequately evaluate sample selection processes and raises concerns regarding the integrity of intention‐to‐treat analyses.

Further concerns arise regarding the screening and recruitment strategy, as participants were reportedly recruited through social media advertisements. While this approach may increase accessibility, it raises important questions about ecological validity and sample representativeness. Individuals with PN rarely seek treatment explicitly for narcissistic difficulties; rather, they typically present with comorbid conditions (e.g., substance use and behavioural addictions) or with symptoms such as anxiety, depression or interpersonal crises.

It is therefore unclear how participants would have self‐referred specifically for the treatment of NPD or PN. This raises concerns that the recruited sample may not adequately reflect the clinical population typically encountered in routine practice.

8. Exclusive Reliance on Self‐Report Measures

The trial relied almost entirely on self‐report instruments, including its primary outcome measure. Although Mohajerin et al. acknowledge this as one of their study limitations, we believe its implications for interpreting the findings deserve considerably greater emphasis in light of the impaired self‐awareness (Dimaggio 2022) and realistic self‐reporting that characterize narcissistic personality pathology. This represents a major limitation in the context of PN, where self‐perception is often distorted, defensive and context‐dependent. Furthermore, when participants are unblinded in a trial (i.e., they know which treatment they were randomly allocated to), their self‐reported outcomes become at increased risk of bias due to the lack of blinding (Juul et al. 2021). The absence of (1) clinician‐rated outcomes, (2) informant‐based assessments or (3) objective indicators of functioning severely limits the validity of the findings and raises the possibility of systematic response bias.

9. Limited Transparency in Supervision and Treatment Fidelity Procedures

The trial reported that therapists received supervision, and that treatment adherence was assessed based on ratings of recorded sessions. However, key aspects of supervision and fidelity monitoring remain insufficiently specified, limiting confidence in the robustness of these procedures. In particular, it is unclear whether supervision was conducted by independent experts or individuals involved in the study design, whether adherence ratings were performed by blind raters, whether inter‐rater reliability was established and whether fidelity assessments captured qualitative aspects of therapeutic process, beyond protocol compliance. This is especially relevant in the context of PN, where treatment typically involves complex relational dynamics, alliance ruptures and therapist countertransference reactions.

It is possible that high reported therapist adherence rates, in the absence of detailed methodological safeguards, may therefore reflect procedural compliance rather than meaningful therapeutic fidelity and should be interpreted with caution.

10. Nonequivalence of Treatment Dose and Duration

One of the major methodological concerns is the substantial asymmetry in treatment dose and duration between conditions. Even though Mohajerin et al. list this as one of their study limitations, we believe that there are more serious implications related to this limitation. The UP was delivered in approximately 28–36 sessions over 7–9 months, whereas SFT involved approximately 48–53 sessions over 14 months. This introduces a systematic confound between treatment model and treatment exposure, rendering between‐group comparisons difficult to interpret. Differences attributed to treatment mechanisms may instead reflect dose, intensity or therapeutic contact time. Although the authors statistically controlled for number of sessions, such adjustment cannot compensate for structural nonequivalence of interventions, particularly in psychotherapy research where treatment duration is inherently tied to theoretical model and therapeutic processes. Studies document that PN changes, though gradually and only slowly (for a review see Orth et al. 2024). Different facets of PN improve at a different, though somewhat comparable pace (Orth et al. 2024), and spontaneous or rapid remissions were not observed in longitudinal studies. Pre–post study of NPD patients in treatment document that NPD remits over the course of several years, not months (Weinberg et al. 2024). Thus, the rapid improvement reported in the study by Mohajerin et al. (2026) seems unlikely given existing evidence and, if accurate, calls for an explanation of the mechanisms underpinning such swift recoveries.

11. Lack of Clarity Regarding Diagnostic Disclosure

The published study does not describe how disclosing an NPD was handled. Since NPD was a requirement for participation in the study, one might expect that researchers disclosed the diagnosis to the participants. This is because the participants needed to provide informed consent to participate in the study and therefore needed to be informed what the study was about. However, disclosure of NPD diagnosis is a complicated clinical issue. There are reasons to consider the impact of the diagnostic disclosure, e.g., patients could feel that their struggles get validated, and their problems in relationships, work and family do not come from their ‘bad’ nature but from a disorder that can be treated. There are also reasons for not disclosing the diagnosis: Patients could feel stigmatized, or that behaviour and attitudes that they experience as ego‐syntonic are considered pathological by the treating clinician in the absence of a solid therapeutic alliance. Currently, there are no guidelines for whether and how such disclosure should be implemented, and clinicians are advised to approach these questions on a case‐by‐case basis (Hersh et al. 2019). It is therefore unclear how this issue was handled in the study, if and how an NPD diagnosis was disclosed, how participants reacted and how diagnostic disclosure affected participants' consent to participate in the study.

12. Conclusions

Taken together, the convergence of multiple concerns raises serious doubts regarding the internal and external validity of the trial. After reviewing the methodology and reported findings, we are reminded of Loschmidt's (1876) paradox: While certain outcomes may be statistically possible, their occurrence can be so unlikely that they challenge plausibility in real‐world conditions. In a similar vein, the results reported in this trial appear difficult to reconcile with the expected complexity, variability and clinical challenges inherent to PN and its treatment.

While the effort to conduct an RCT in this field is commendable, the present design and reported findings fall short of the standards required to support the strength of the conclusions drawn. We therefore encourage readers to interpret these results with the utmost caution, particularly considering the atypical nature of the sample and the unusually consistent pattern of outcomes reported.

In addition, we respectfully request that the authors provide detailed clarifications regarding the issues presented in this commentary, and that the journal considers whether the current evidence base justifies the claims made. We identified the following concerns regarding the trial: (i) the unlikely possibility that consecutive sample recruitment would result in almost equal number of participants in grandiose vs. vulnerable subgroups; (ii) the unlikelihood that consecutive sample recruitment would result in almost equal number of participating men and women; (iii) the unlikelihood that grandiose and vulnerable narcissism would appear as uncorrelated subtypes; (iv) dissimilar treatment responses to SFT vs. UP; (v) absence of dropouts; (vi) unclear randomization procedure; (vii) poor screening and recruitment procedure and limited transparency regarding trial registration; (viii) exclusive reliance on self‐report measures; (ix) limited transparency in supervision and treatment fidelity procedures; (x) nonequivalence of treatment dose and duration; and (xi) lack of clarity on how the disclosure of NPD diagnosis to participants was handled in the trial.

Finally, we urge editors of journals to make use of gold standard review procedures, particularly if data may be used in later meta‐analyses, to ensure the appropriate dissemination of new and clinically useful information, which is desperately needed in the field. Reflecting on our comments, the authors of this commentary agreed that a separate paper is required, which outlines typical challenges in conducting RCTs with patients with PN. This is especially timely, given the high public health significance of development of effective interventions for patients with PN, proliferation of newly developed psychotherapies for PN and paucity of treatment outcome studies in this field. From the perspective of assessment, (i) future studies should combine self‐report measures with clinical interviews and functional assessments, and (ii) interviewers should be blind to the group assignment and not the study therapists. From the sample recruitment perspective, sampling should also include participants with clinically significant levels of grandiosity and vulnerability. Dimensional facets of PN (e.g., grandiosity vs. vulnerability) can be assessed in addition to more categorical measures (e.g., psychiatric diagnosis). In addition, samples should be stratified for gender given the uneven gender distribution of PN and gender dimorphism in the manifestation of PN. From the perspective of retention and attrition, studies should document the strategies used to retain samples, such as attending to the ongoing negotiation of the therapeutic contract. In addition, given the high emotional burden on therapists, future studies should implement strategies to help therapists, including supervision, psychoeducation, support and limiting the number of patients treated by a single therapist.

We return to Loschmidt's paradox, expressing scepticism that Mohajerin et al. reported outcomes represent a clinically realistic therapeutic process in the treatment of narcissistic personality pathology. The proposed unlikelihood of these results, along with the methodological shortfalls of the study itself, may represent the actual challenges of recruitment, assessment and diagnoses and therapeutic engagement and retention that have limited the necessary randomized controlled trials to date. More collaborative dialogue is needed as well as methodological adaptation to develop the much‐needed evidence‐based therapeutic interventions for PN.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The authors have nothing to report.