Introduction

Parkinson’s disease (PD) affects a highly heterogeneous group of patients. Several subtyping schemes have been proposed to predict patient outcomes, such as postural instability and gait Disturbance (PIGD) versus tremor-dominant Parkinson’s disease (TDPD) and the akinetic-rigid (AR) subtype versus TDPD1,2,3. These subtypes suggest underlying circuitry dysfunction4 and are associated with different outcomes1. However, motor subtypes are not consistently stable over time, and patients may be classified differently depending on the scheme used5.

Including non-motor symptoms, such as olfactory dysfunction, might improve the association with disease severity6. Olfactory dysfunction or hyposmia is a common prodromal and non-motor symptom in PD that correlates with dopaminergic transporter uptake in early PD1,7,8. Hyposmia also increases the risk of synucleinopathy in patients with isolated rapid eye movement sleep behavior disorder (iRBD)9, making it a potential biomarker for disease progression.

Studies suggest that patients with the TDPD subtype tend to have better olfactory function compared to those with PIGD or AR subtypes10,11,12,13,14,15. Additionally, olfactory dysfunction correlates with axial symptoms like freezing of gait and camptocormia16,17, but this association is not universally observed18. Since the AR subtype reflects a more severe dopaminergic deficit19, and PIGD is linked to faster cognitive decline and more prominent non-motor symptoms1, combinations of olfactory and motor subtypes might indicate a more widespread neurotransmitter dysfunction. The aim of this study was to determine whether integrating dichotomized olfactory subtypes into motor subtype classification improves the assessment of overall symptoms, motor symptoms, cognition, quality of life, and dopaminergic imaging in PD.

Results

Patient characteristics

A total of 157 participants were included. The median age was 67 years, and the mean score on the University of Pennsylvania Smell Identification Test (UPSIT) was 16, as shown in Table 1. Among all participants, 55.4% were classified as PIGD, 37.6% were classified as TDPD, and 7% were classified as indeterminate. The majority (59.9%) had total anosmia (TA), while 40.1% had non-anosmia (NA). There was no significant difference in the distribution of motor subtypes between the anosmia and non-anosmia groups. The p-value for Fisher’s test, which was 0.604, which indicates no significant difference between the groups.

Table 1 Clinical characteristics of the study population (total n = 157).
Full size table

Comparison among motor and olfactory subtypes

There were 55, 32, 34, and 25 participants in the PIGD TA, PIGD NA, TDPD TA, and TDPD NA groups, respectively. The Kolmogorov-Smirnov test indicated a normal distribution for only the scores on the Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale (M-UPDRS) part III, but not any other clinical measures. Significant differences were identified between subgroups in total M-UPDRS scores with the highest occurring for PIGD TA, followed by PIGD NA, TDPD TA, and TDPD NA (59.0, 47.5, 43.0, and 36.0, respectively; H = 23.798, p < 0.001), as well as in the 39-item Parkinson’s Disease Questionnaire Summary index (PDQ-39 SI) scores (22.4, 22.8, 9.6, and 9.0; H = 24.864, p < 0.001), as shown in Table 2; Fig. 1.

Table 2 Clinical characteristics of the different motor and olfactory subtypes (n = 146).
Full size table
Fig. 1
figure 1

This figure shows clinical measurement scores across the four combined motor and olfactory subtypes: PIGD TA, PIGD NA, TDPD TA, and TDPD NA. Panels display (A) M-UPDRS total, (B) M-UPDRS part III, (C) MoCA, and (D) PDQ-39 SI scores. One-way ANOVA was used for M-UPDRS part III analysis, while the Kruskal-Wallis test was applied to other scores. Error bars represent the 95% confidence interval for the median. Clinical severity, as measured by M-UPDRS total scores, followed the order of PIGD TA, PIGD NA, TDPD TA, and TDPD NA. MoCA, Montreal Cognitive Assessment; M-UPDRS, Movement Disorder Society-Sponsored Revision of the Unified Parkinson’s Disease Rating Scale; NA, Non-anosmia; PDQ-39 SI, 39-item Parkinson’s Disease Questionnaire Summary index; PIGD, postural instability and gait disturbance; TA, total-anosmia; TDPD, Tremor-dominant Parkinson’s disease;

Full size image

Those with PIGD TA had the highest disease burden, while those with TDPD NA had the lowest. The Montreal Cognitive Assessment (MoCA) scores also varied significantly, with those who had PIGD TA scoring the lowest and those with TDPD NA scoring the highest (24.0, 25.5, 24.5, and 28.0; H = 15.183, p = 0.002). M-UPDRS part III scores displayed a trend toward significance across the groups (31.5, 30.5, 31.0, and 24.0; F = 2.514, p = 0.061), but no significant differences were observed in TRODAT sum scores (5.0, 4.0, 5.0, and 5.0; p = 0.133).

A general linear model (GLM) analysis was performed to investigate the trends further. Table 3 presents the results adjusted for age, sex, and disease duration, while Table 4 shows the results with additional adjustments for levodopa equivalent daily dose (LEDD). Bonferroni correction was applied to account for multiple subgroup comparisons, and the adjusted significance level was set as α = 0.0167 for each variable (i.e., when using different groups as the reference). To account for the four clinical outcomes analyzed, the overall adjusted threshold was α = 0.0042.

Table 3 Effect of motor and olfactory subtype interaction on estimated values: a GLM analysis adjusted for age, sex, and disease duration.
Full size table
Table 4 Effect of motor and olfactory subtype interaction on estimated values: a GLM analysis adjusted for age, sex, disease duration, and LEDD.
Full size table

The GLM analysis showed consistent trends in the total M-UPDRS score and PDQ-39 SI scores, indicating the highest disease burden for PIGD TA, followed by PIGD NA, TDPD TA, and TDPD NA, when TDPD NA was used as the reference. PIGD TA exhibited the highest disease burden, which was reflected in the total M-UPDRS score (B = 19.646, p = 0.0002) and PDQ-39 SI score (B = 13.983, p = 0.0013). In comparison, TDPD NA had the most favorable outcomes. For total M-UPDRS, PIGD NA had B = 16.400 (p = 0.0029), and TDPD TA had B = 8.412 (p = 0.1216).

For PDQ-39 SI, PIGD NA had B = 9.641 (p = 0.0347), and TDPD TA had B = 1.041 (p = 0.8183). Although the differences between PIGD NA and TDPD TA were not statistically significant, PIGD NA showed a trend toward higher PDQ-39 SI scores compared to TDPD TA, suggesting greater impairment (B = 8.601, p = 0.0418). For MoCA, a trend was observed for worse scores in TDPD TA compared to TDPD NA (B = − 2.603, p = 0.0387). These findings are summarized in Table 3.

Table 4 confirmed these trends. For total M-UPDRS scores, using TDPD NA as the reference, PIGD TA had B = 15.528 (p = 0.0024), PIGD NA had B = 13.024 (p = 0.0143), and TDPD TA had B = 8.069 (p = 0.1205). For PDQ-39 SI, PIGD TA had B = 9.456 (p = 0.0191), PIGD NA had B = 5.770 (p = 0.1710), and TDPD TA had B = 0.668 (p = 0.8717). Similarly, for MoCA, a trend was observed for worse scores in TDPD TA compared to TDPD NA (B = − 2.606, p = 0.0392).

Effect of olfactory categorization on clinical parameters by motor subtype

Olfactory subtyping showed a trend toward better MoCA scores for NA compared to TA within the TDPD subtype (B = 2.614, p = 0.051), but no significant differences were found for the PIGD subtype. No significant differences in M-UPDRS, PDQ-39 SI, or TRODAT scores were found between NA and TA groups. These results are shown in Table 5.

Table 5 Effect of olfactory subtype on clinical parameters by motor subtype: a GLM analysis.
Full size table

Post-hoc power analyses

Stratifying the sample into four subgroups reduced the sample size within each group, potentially affecting the statistical power. To address this concern, we conducted post-hoc power analyses using the effect sizes (f) derived from the Kruskal-Wallis test, and the results are presented in Supplementary Table 1 (Table S1). These results indicate that the overall sample size was adequate for most measures, except for M-UPDRS part III, which showed a small effect size (f = 0.095) and higher sample-size requirement (sample size needed = 1208).

Discussion

This study has explored the interaction between motor and olfactory subtypes in PD, which revealed significant differences in disease severity, motor and cognitive function, and quality of life across subgroups. PIGD TA showed the most severe symptoms, while TDPD NA had the least. However, there were no significant differences between PIGD NA and TDPD TA. Interestingly, olfactory subtyping appeared more relevant in the TDPD group, with TA showing a trend toward worse cognitive function than the NA group.

Although different motor subtype schemes have been employed, their consistency over time remains variable2,3,5,20,21,22,23. Despite this variability, a consistent theme emerges: olfactory dysfunction measured across various tests is generally worse in PIGD and AR subtypes compared to TDPD. For example, three studies10,11,12 found that patients with TDPD had higher identification scores than patients with PIGD or AR subtypes, indicating less olfactory impairment. Similarly, olfactory thresholds were lower in AR compared to TDPD13. Other studies11,14,15 demonstrated a significant association between hyposmia and PIGD or AR subtypes, with patients exhibiting more pronounced olfactory deficits than those with TDPD. These findings are summarized in Table S2.

Interestingly, one study24 reported no cross-sectional difference in olfactory function between motor subtypes. However, the proportion of PIGD increased in the hyposmia group during the second year. Although axial symptoms such as freezing of gait have been linked to abnormal olfaction16,17, the proportion of patients with abnormal olfaction in cases of PIGD has not been consistently higher across studies16,18,25, including our own.

From the perspective of the “brain-first” and “body-first” alpha-synuclein propagation hypothesis, “body-first” PD may exhibit severe hyposmia in early stages, while “brain-first” PD typically shows hyposmia at more advanced stages26. Studies have found that the AR or PIGD subtypes, often associated with severe hyposmia, tend to involve patients with shorter disease durations or drug-naïve cases11,14,15. Conversely, olfactory dysfunction has been correlated with TDPD in drug-naïve patients in one study27. This suggests that as brain-first PD progresses in both clinical severity and olfactory dysfunction, the initial association between PIGD and olfactory dysfunction may lessen over time.

Research suggests that combining olfactory subtypes with non-motor features correlates with worse motor outcomes and a higher prevalence of PIGD7,18. In our study, integrating olfactory subtypes with motor subtypes was linked to distinct outcomes. PIGD TA had the poorest motor, cognitive, and quality-of-life outcomes, while TDPD NA consistently had the best scores. No significant differences were noted between PIGD NA and TDPD TA, although a trend toward better PDQ-39 SI scores in PIGD NA was observed. However, no significant differences were found between PIGD TA and PIGD NA, which could indicate that olfactory dysfunction does not exacerbate motor deterioration once postural instability and gait impairment are present.

This is consistent with previous findings suggesting that olfactory function is more closely linked to appendicular activities of daily living (ADLs), such as buttoning clothes, than to axial ADLs28. As a result, combining olfactory dysfunction with motor subtypes might not produce significant differences, particularly in the PIGD subtype. Future longitudinal studies could further clarify the prognostic value of this combined classification in predicting disease progression.

Validation in both de novo and early to moderate PD has been obtained for the recent classification of PD as mild motor-predominant, diffuse malignant, and intermediate subtypes, based on motor and non-motor symptoms, such as orthostatic hypotension, mild cognitive impairment, REM sleep behavior disorder (RBD), depression, anxiety, and motor scores6,29,30,31. Interestingly, olfactory function was not identified as a key factor in this scheme. One of the studies demonstrated that olfactory impairment was most severe in the diffuse malignant subtype, followed by the intermediate subtype, and least severe in the mild motor-predominant subtype30.

We observed a trend of increasing severity in M-UPDRS and PDQ-39 SI scores across subtypes, with PIGD TA showing the highest burden, followed by PIGD NA, TDPD TA, and TDPD NA. However, these differences did not reach statistical significance after multiple comparisons, suggesting a limited role for incorporating olfactory subtypes into this classification system. Notably, within the TDPD group, worse olfaction was associated with poorer cognition, indicating that olfactory evaluation may have greater utility for patients with milder motor manifestations. We hypothesize that olfactory function exhibits a ceiling effect: in patients with preserved olfaction, it may correlate with motor progression32, while in those with significant olfactory impairment, motor progression is likely influenced by other factors. As such, tracking olfactory function may be more informative in early-stage or mild motor-predominant PD, where it might correlate with motor progression and cognitive deterioration.

The lack of significant differences in 99mTc-TRODAT-1 scan results between the four subgroups suggests that this imaging modality may not sufficiently differentiate clinical phenotypes arising from the combined effects of motor and olfactory subtypes. Studies using the 99mTc-TRODAT-1 have shown differences at various stages of PD33. Another study using [123I] fluoropropyl–carbomethoxy-3β single-photon emission computed tomography (FP-CIT SPECT) revealed distinct patterns of dopaminergic loss, with AR subtypes showing more significant depletion in the dorsal putamen, and TDPD showing reductions in the lateral putamen and caudate nucleus34. Other research has linked motor subtypes (tremor vs. non-tremor) to caudate nucleus changes35.

Additionally, olfactory identification has been strongly correlated with putamen uptake in 99mTc-TRODAT-1 scans36, and striatal binding ratios correlate with olfactory function in TDPD, but not PIGD subtypes37. While these findings suggest a differential dopaminergic uptake pattern in PIGD TA compared to TDPD NA, our study did not demonstrate such differences. Future research may benefit from more regionally precise dopaminergic assessments using 99mTc-TRODAT-1 or other tracers.

When considering olfactory subtyping within motor subtypes, our data indicate that TA is associated with worse cognitive function in TDPD, but not in PIGD. This implies that olfactory subtyping may be more useful in differentiating cognitive outcomes in TDPD, and TDPD NA likely reflects milder dopaminergic and non-dopaminergic dysfunction8,19. In contrast, PIGD is more closely linked to cognitive decline2, which may limit the added value of olfactory subtyping in this group. Studies have highlighted a relationship between olfactory function and executive functioning within the MoCA domains, possibly pointing to underlying frontal lobe pathology38. This connection could further support the brain-first hypothesis, where Lewy body pathology accumulates in the cortices26.

This study has several strengths. First, it is the first to investigate the role of olfactory subtypes in conjunction with motor subtypes in PD and offers new insights into the complex heterogeneity of the disease. By adjusting for key modifiers such as age, sex, disease duration, and LEDD, we ensured a robust analysis. Second, we have provided a comprehensive review of the existing literature on motor subtypes and olfactory function, which could serve as a valuable resource for future studies.

However, several limitations must be acknowledged. First, the dichotomization of olfactory function is a simplified approach that may not capture the full spectrum of olfactory impairment. Although a cutoff score of 19 was used to define TA, this threshold may not reflect the nuances of olfactory decline across different ages32,39 or fully represent its association with synucleinopathy progression9, Furthermore, a single cutoff score does not reflect the entire spectrum of olfactory impairment40,41. Another limitation is that PIGD may represent more of a trait marker than a stable subtype, and its classification may change over time5,42. The cross-sectional design of our study limits the ability to determine causality or track symptom progression. The inclusion of olfactory biomarkers is a preliminary step, and longitudinal studies are needed to assess the stability of these subtypes.

We did not include scores for RBD or autonomic dysfunction, which are both critical for PD subtype characterization. Nevertheless, our primary objective was to assess the impact of an olfactory subtype in combination with motor subtypes. This study was designed as an exploratory analysis aimed at identifying trends and generating hypotheses. However, the COVID-19 restrictions in Taiwan (April 2020 to December 2022) coincided with our study period (October 2016 to May 2021), which significantly impacted the feasibility of olfactory evaluations due to mask mandates and limiting patient follow-up.

Additionally, the overlap among subgroups, the relatively small sample size, and the inclusion of numerous variables introduce limitations that may have affected the robustness of the findings. Studies with larger, more representative cohorts and refined subgroup classifications are essential to validate these trends and to better elucidate the clinical implications of integrating motor and olfactory subtypes in PD. Finally, the visual rating of 99mTc-TRODAT-1 scans may introduce subjectivity and limit the precision of dopaminergic assessments. Future studies should consider using more objective imaging techniques to improve accuracy.

In conclusion, integrating motor and olfactory subtypes may enhance the classification of PD severity. PIGD TA represented the most severe group, while TDPD NA showed the least severity. Subtyping based on olfactory function, particularly within TDPD, may provide insights into cognitive outcomes. Longitudinal studies are needed to confirm the stability and clinical relevance of these subtypes.

Methods

Study designs

This study was conducted at Taichung Veterans General Hospital between October 2016 and May 2021. Participants met the clinical diagnostic criteria for PD of the International Parkinson and Movement Disorder Society43. Clinical assessments included the University of Pennsylvania Smell Identification Test (UPSIT)44,45, the Movement Disorder Society-sponsored revision of the Unified Parkinson’s Disease Rating Scale (M-UPDRS)46, Montreal Cognitive Assessment (MoCA)47, the 39-Item Parkinson’s Disease Questionnaire Summary index (PDQ-39 SI)48,49, and 99mTc-TRODAT-1 single-photon emission computed tomography (SPECT) imaging.

This study was approved by the Institutional Review Board and Ethics Committee of Taichung Veterans General Hospital (No. CE16171B). Written informed consent was obtained from all participants prior to enrollment in accordance with the ethical standards outlined in the Declaration of Helsinki. A cross-sectional design was used to evaluate the interaction between olfactory and motor subtypes. Exclusion criteria included upper respiratory infection, nasal surgery, and incomplete assessments.

Patients’ olfactory subtypes were classified based on UPSIT scores as either the total anosmia (TA) group or non-anosmia (NA). A threshold of UPSIT < 19 for anosmia was selected for the following reasons: (1) it has been widely used to classify anosmia across age groups40, (2) it has been employed in studies to identify individuals at risk for synucleinopathy in patients with REM Sleep Behavior Disorder (RBD)9, and (3) it aligns closely with the average UPSIT values observed in patients with Parkinson’s Disease (PD)50,51. Statistical analyses included data from patients whose M-UPDRS and TRODAT scan evaluations occurred within one year of each other.

Assessment of clinical parameters

Olfactory function was evaluated using the traditional Chinese version of the UPSIT, with scores ranging from 0 to 40. The test consists of 40 microencapsulated odors, each requiring participants to choose the correct odor from four possible options. Disease severity and motor symptoms were assessed using the M-UPDRS (0–260) in the ON state, while cognition was measured by MoCA (0–30). Quality of life was evaluated using the PDQ-39 SI (0–100). Higher scores indicated worse symptoms, except for UPSIT and MoCA, where higher scores reflected better function.

Motor subtypes were defined using M-UPDRS scores23. The tremor score was derived as the mean of 11 UPDRS items (2.10, 3.15–3.18), while the PIGD score was calculated as the mean of five UPDRS items (2.12, 2.13, 3.10–3.12). Patients were classified as TDPD if the tremor-to-PIGD ratio was ≥ 1.15, as PIGD if the ratio was ≤ 0.90, and as indeterminate if the ratio was in between these values.

Dopamine transporter scan

SPECT using 99mTc-TRODAT-1 (TRODAT) was utilized to evaluate dopamine transporter binding. Patients received a slow intravenous injection of a single bolus of 26.7 nCi of 99mTc-TRODAT-1. The TRODAT scan was performed 3 h after the injection. The visual rating scale for the TRODAT scan assessment was adapted from previous studies33,52,53, with modifications to assess each side of the brain separately rather than as a whole.

The rating scale was as follows: 0 signified normal striatal uptake; 1 indicated less than 50% reduction in putaminal uptake with intact caudate uptake; 2 represented more than 50% reduction in putaminal uptake but with preserved caudate uptake; 3 signified a decrease of over 50% in both caudate and putaminal uptake; and 4 denoted complete absence of striatal uptake. Points were allocated separately for each side, and their sum formed the total score. The ratings were conducted by two neurologists who were blinded to the clinical details.

Statistical analysis

Statistical analyses were performed using SPSS version 22.0 for Windows (SPSS Inc., Chicago, IL) and MedCalc® Statistical Software version 22.023 (MedCalc Software Ltd, Ostend, Belgium). Kolmogorov-Smirnov tests were used to assessed data distributions. Parametric or non-parametric comparisons among olfactory and motor subtypes were performed using analysis of variance (ANOVA), Kruskal-Wallis, and Fisher’s exact tests. General linear models (GLMs) were applied to estimate beta coefficients and to assess differences in clinical parameter among groups. Adjustments were made for age, sex, disease duration, and LEDD. A two-tailed test with a significance level of p < 0.05 was applied unless specified otherwise.

Post-hoc power analyses

The data simulation for the post-hoc power analysis was based on the median and interquartile range (IQR), assuming normality, with the standard deviation estimated as \(\:SD=\frac{IQR}{1.349}\:\), following the methodology proposed by Higgins and Keller-McNulty54. While this assumption simplifies the simulation process, it does not affect the validity of the Kruskal-Wallis test results or the effect size calculations (\(\:{\eta\:}^{2}=\frac{(H-k\:+\:1)}{(N-k)}\)), which are rank-based and inherently non-parametric, as outlined by Tomczak and Tomczak55. Additionally, a conservative effect size estimate \(\:{(\epsilon\:}^{2}=\frac{H}{{(N}^{2}-1)}\)) appropriate for non-parametric tests was derived.

The G*Power software was employed to conduct the analysis using \(\:f=\sqrt{\frac{{\eta\:}^{2}}{(1\:-{\eta\:}^{2})\:}}\), selecting “ANOVA: Fixed effects, omnibus, one-way” as the test type. This approach does not rely on normality assumptions. Simulated data were utilized to estimate statistical power and contextualize the effect sizes within the study.

A preliminary power analysis was not conducted due to practical constraints, as the sample size was determined by participant availability during the study period. Given the exploratory nature of the research, maximizing participant recruitment was prioritized over predefined sample sizes. While post-hoc power analysis has recognized limitations, recent studies suggest it can yield valuable insights when applied carefully56. We acknowledge the importance of accounting for sampling variability and incorporating confidence intervals when interpreting the results.