Multiplex PCR panel dynamics: implications for therapy duration and methodological considerations

Dear editor,

We read with great interest your recent article by Dessajan et al. [1], which evaluated the potential utility of syndromic multiplex PCR of lower respiratory tract samples for predicting clinical outcomes in patients with ventilated hospital-acquired pneumonia (HAP) and ventilator-associated pneumonia (VAP). The authors conclude their findings suggest that multiplex PCR semi-quant values associated with detected pathogens did not predict clinical success.

This area of inquiry is particularly relevant given the ongoing controversies surrounding the optimal duration of therapy for non-lactose fermenting gram-negative bacilli (NLF GNB) such as Pseudomonas aeruginosa and Acinetobacter baumannii complex. Both IDSA and ERS/ESICM/ESCMID/ALAT HAP/VAP guidelines endorse 7-day antibiotic courses, including for NLF GNB, while acknowledging that longer durations may be needed in select individuals [2, 3]. While definitions of recurrence and relapse vary across studies (e.g., some require cessation of antibiotics before a new episode while others require documented clinical cure), these outcomes fundamentally represent patients who initially appeared to respond successfully to treatment but subsequently developed new pneumonia episodes. A 2023 systematic review and meta-analysis of randomized controlled trials comparing short- vs. long-course therapy in VAP reported no statistically significant difference in recurrence (OR 1.90, 95% CI 0.99–3.64) or relapse (OR 1.76, 95% CI 0.93–3.33) for NLF GNB [4]. Bayesian meta-analysis using established empirical priors (Turner et al. for heterogeneity [5], uninformative for overall effect) shows a 95% posterior probability of at least 10% increased odds of recurrence and 94.7% probability for relapse with 7-day therapy (Fig. 1). These findings suggest a tangible risk of clinical deterioration in subgroups that may be obscured by dichotomous significance thresholds for short-course therapy. Despite current guidelines favoring short-course therapy based on equivalent traditional endpoints (ventilator-free days, ICU length-of-stay, mortality), reducing recurrence through diagnostic-guided personalized duration could represent a clinically meaningful advance that is analogous to how differences in global cure rates have influenced treatment recommendations in other infectious syndromes [6].54

Fig. 1.

Bayesian meta-analysis posterior distributions showing probability of increased recurrence (A) and relapse (B) risk with short-course antibiotic therapy in VAP for NLF GNBs

These data highlight the need for robust evaluation of molecular diagnostics that may enable individualized treatment decisions. While the study by Dessajan et al. included only 28 patients (21%) with Pseudomonas aeruginosa and 8 patients (5.8%) with Acinetobacter baumannii complex, the reported median therapy duration was 7 days across the cohort. The limited sample size likely precluded detection of subgroup effects. Additionally, the authors did not report which specific pathogens were associated with recurrence or relapse events, which would have been particularly interesting for the NLF GNB given the above-mentioned controversies. Nonetheless, prior conference proceedings from this research group provide intriguing signals. In non-COVID cases, Pseudomonas aeruginosa remained elevated on semi-quantitative multiplex PCR for a median of 18 days, compared to only 8.3 days by culture [7]. This discordance may offer a window into patient-specific response and risk of relapse. Previous research has reflected increased bacterial burden is correlated to fewer ventilator free days but is also impacted by what bacteria are present and severity of illness among other factors [8]. Though differentiating colonization from true infection remains a fundamental challenge, integrating quantitative molecular dynamics with host biomarkers may provide more nuanced discrimination performance than either approach alone. Establishing consensus clinical thresholds will require substantial validation with semi-quantification research [9]. Further research is needed for integrative diagnostic–prognostic models that go beyond molecular quantification to incorporate host factors, resistance mechanisms, and probabilistic reasoning.

Finally, the authors employed mixed ordinal logistic regression to assess the association between repeated ordinal PCR measures and both time and clinical success. While appropriate for ordinal outcomes, this approach assumes a linear effect on the log-odds. However, many biological processes are non-linear—e.g., the U-shaped relationship between white blood cell count and clinical outcome [10]. Moreover, in the author’s study the waterfall plots reflect an asymmetry around zero (i.e. higher rates of success with −1, −2, and +2 log changes than 0 or +1 log changes), a plateau of effect at large decreases (i.e. −1 log to −2 log does not substantially change success rate), and a non-monotonic transition at zero where success and failure are nearly balanced. These patterns suggest threshold effects (i.e., when the independent variable’s impact on a dependent variable only becomes apparent at a certain level) and biological saturation that linear models cannot capture. Such non-linearities may explain why the linear model failed to detect significant associations, as averaging across complex relationships can mask clinically meaningful signals. Given these observations, we are curious if the authors attempted to fit non-linear models such as generalized additive models.

We commend the authors for advancing the field of molecular diagnostics in pneumonia and hope our observations contribute meaningfully to ongoing discussions about determining how best to personalize therapy in critically ill patients.

Author contributions

The manuscript was written and revised by the four authors listed on the publication. T.T.T. performed analyses using existing publication data. All authors read and approved the final manuscript.

Funding

None.

Data availability

Data publicly available in original systematic review referenced.

Declarations

Ethics approval and consent to participate

Not applicable as this is an opinion paper not involving any patient, nor any patient data.

Consent for publication

Not applicable, as there are not data, no figures.

Competing interests

BH and AMP are employed by bioMérieux. TTT and TK declare that they have no competing interests.