Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Individualized prognosis of cognitive decline and dementia in mild cognitive impairment based on plasma biomarker combinations

Abstract

We developed models for individualized risk prediction of cognitive decline in mild cognitive impairment (MCI) using plasma biomarkers of β-amyloid (Aβ), tau and neurodegeneration. A total of 573 patients with MCI from the Swedish BioFINDER study and the Alzheimer’s Disease Neuroimaging Initiative (ADNI) were included in the study. The primary outcomes were longitudinal cognition and conversion to Alzheimer’s disease (AD) dementia. A model combining tau phosphorylated at threonine 181 (P-tau181) and neurofilament light (NfL), but not Aβ42/Aβ40, had the best prognosis performance of all models (area under the curve = 0.88 for 4-year conversion to AD in BioFINDER, validated in ADNI), was stronger than a basic model of age, sex, education and baseline cognition, and performed similarly to cerebrospinal fluid biomarkers. A publicly available online tool for individualized prognosis in MCI based on our combined plasma biomarker models is introduced. Combination of plasma biomarkers may be of high value to identify individuals with MCI who will progress to AD dementia in clinical trials and in clinical practice.

Main

About 50 million people live with dementia globally, with the prevalence expected to double by 2030 (ref. 1). Fifty to seventy percent of all dementia cases are caused by Alzheimer’s disease (AD)2. The dementia stage is preceded by mild cognitive impairment (MCI). Accurate prognosis is important in MCI, since it may either lead to cognitive decline and dementia (due to AD or other diseases) or be benign and stable3. If disease-modifying treatments became available for AD4, accurate prognostics may be important to guide treatment in patients with MCI.

Even at the MCI stage, key pathological hallmarks of AD can be detected in vivo using cerebrospinal fluid (CSF) biomarkers (for example, the ratio of Aβ42 to Aβ40, and tau phosphorylated at threonine-181 (P-tau181)5,6) or positron emission tomography (PET) of Aβ and tau7,8. However, the use of these technologies is limited due the perceived invasiveness of lumbar punctures and the high cost and low availability of PET imaging. Blood-based biomarkers could overcome these hurdles.

Blood-based biomarkers of Aβ (A), tau (T) and neurodegeneration (N) in AD (that is, the ATN biomarkers)9 include the Aβ42/Aβ40 ratio10, P-tau181 (refs. 11,12,13) and neurofilament light (NfL)14,15, respectively. Aβ42/Aβ40 and P-tau181 in plasma correlate with Aβ-PET and tau-PET findings, respectively, and can distinguish AD dementia from controls and non-AD neurodegenerative disorders10,11,12,13,16. Blood-based NfL is associated with cortical atrophy and cognitive decline in AD17. P-tau217 in plasma has also recently been described18. However, most studies on blood-based AD biomarkers report findings at the group level. There is a gap in our understanding of how well these biomarkers predict clinical outcomes at the individual patient level and how they compare with more basic prediction models. Individualized assessment has recently been applied using CSF and related imaging biomarkers in MCI19,20. A similar study is lacking for blood-based biomarkers. It could be of great value for clinical practice and trials to investigate whether plasma ATN biomarkers perform as well as CSF biomarkers and better than more basic prediction models. We have previously performed a study with a multivariate approach to examine plasma biomarkers and the risk of progression from MCI to AD dementia11, but most other studies focused on evaluating the biomarkers individually10,12,13. None of these studies, however, applied the ATN classification system9, nor did they systematically aim to find the best subset of ATN biomarkers for individualized predictions.

We measured plasma Aβ42/Aβ40, P-tau181 and NfL in patients with MCI from two large cohorts and tested which subset of plasma biomarkers best predicted individual risk for cognitive decline and progression to AD dementia. We compared the prognostic ability of plasma biomarkers with the same biomarkers measured in CSF, as well as with a more basic prediction model, and cross-validated our individual-based risk assessment models both within and across cohorts.

Results

Study population characteristics (model selection sample)

A total of 148 patients with MCI from the Biomarkers for Identifying Neurodegenerative Disorders Early and Reliably (BioFINDER) cohort for whom all three plasma and CSF biomarker measurements were available, along with at least one of the primary or secondary outcomes, were used for model selection (Table 1). The mean age was 71.4 years, the mean education duration was 11.2 years, 36.5% were female and the mean mini-mental state examination (MMSE) score was 27.2 ± 1.7 at the baseline. Moreover, the mean MMSE score was 21.8 ± 5.2 at 4 years after the baseline and conversion to AD dementia was 59.8% within 4 years of the baseline. There was a significant negative correlation between plasma Aβ42/Aβ40 and plasma P-tau181 (coefficient of determination (R2) = −0.30; P < 0.0001; see Extended Data Fig. 1) and a significant positive correlation between plasma P-tau181 and plasma NfL (R2 = 0.33; P < 0.0001; Extended Data Fig. 1), but no significant correlation was observed between plasma Aβ42/Aβ40 and plasma NfL (R2 = −0.08; P = 0.31; Extended Data Fig. 1). The associations between corresponding CSF biomarkers in the BioFINDER model selection sample are shown in Extended Data Fig. 2. The study procedures are outlined in Fig. 1.

Table 1 Study participant characteristics (model selection)
Fig. 1: Flow chart of study procedures.
figure1

In the ADNI cohort, 425 participants had data for plasma P-tau181 and NfL. Of these, 86 also had data for plasma Aβ42/Aβ40. The number of participants available for different outcomes differed for both BioFINDER and ADNI. Asterisks in the section ’Model selection’ represent the ADNI model selection sample, which included individuals in the ADNI cohort for whom data were available for all plasma biomarkers (that is, Aβ42/Aβ40, P-tau181 and NfL). In the section ‘Individualized prediction using best models’, ‘internal prognostic validation’ refers to fivefold cross-validation for the internal prognostic validation step, whereas ‘external prognostic validation’ refers to validating the model selection (that is, validating that the best-performing model for BioFINDER was the best-performing model for ADNI, and vice versa). Asterisks in the section ‘Individualized prediction using best models’ represent the ADNI prognostic validation sample, which included individuals in the ADNI cohort for whom data were available for P-tau181 and NfL.

The data from 86 patients with MCI from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) cohort for whom all three plasma biomarker measurements were available alongside at least one of the primary or secondary outcomes were used to replicate model selection (Table 1). The mean age was 71.5 years, the mean education duration was 16.4 years, 51.2% were female and the mean MMSE score was 28.3 ± 1.7 at the baseline. Moreover, the mean MMSE score was 27.6 ± 2.9 at 4 years after the baseline and conversion to AD dementia was 10.8% within 4 years of the baseline. There was a significant negative correlation between plasma Aβ42/Aβ40 and plasma P-tau181 (R2 = −0.31; P = 0.0023; Extended Data Fig. 2) but no significant correlation between plasma P-tau181 and plasma NfL (R2 = 0.15; P = 0.16; Extended Data Fig. 2) or between plasma Aβ42/Aβ40 and plasma NfL (R2 = −0.16; P = 0.11; Extended Data Fig. 2).

To validate our prognostic models, we used data from 425 patients with MCI from the ADNI cohort for whom plasma P-tau181 and NfL measurements were available alongside at least one of the primary or secondary outcomes available. The mean age of this cohort was 71.0 years, the mean education duration was 16.1 years, 51.8% were female and the mean MMSE score was 28.2 ± 1.7 at the baseline. Moreover, the mean MMSE was 26.6 ± 4.1 4 years after the baseline and conversion to AD dementia was 33.1% within 4 years of the baseline. There was a significant positive correlation between plasma P-tau181 and plasma NfL (R2 = 0.35; P < 0.0001; Extended Data Fig. 3).

Model selection for longitudinal cognition

With 4-year MMSE as the outcome in the BioFINDER model selection sample (n = 118), the model that included plasma Aβ42/Aβ40, P-tau181 and NfL as predictors (full model; R2 = 0.36; Akaike information criterion (AIC) = 684) fit the data significantly better than the basic model that included age, sex, education and baseline MMSE score (R2 = 0.24; AIC = 702; P = 0.0001 compared with the full model). However, the best-fitting model according to the AIC was the one that included plasma P-tau181 and NfL but not Aβ42/Aβ40 (R2 = 0.36; AIC = 683). In the best-fitting model, there was a significant individual effect of P-tau181 (β = −1.65; P < 0.0001) but not NfL (β = −0.70; P = 0.13) (Fig. 2a,b and Supplementary Table 1).

Fig. 2: Modeling cognitive decline using plasma Aβ42/Aβ40, P-tau181 and NfL.
figure2

Results from modeling cognitive decline in patients with MCI using plasma biomarkers. a, R2 (x axis) and AIC values (numbers in plots) for each plasma-based model, with MMSE evaluated 4 years after the baseline as the outcome, in the ADNI and BioFINDER cohorts. The dashed vertical lines show basic model performances for reference. Error bars indicate 95% confidence intervals. All models also included age, sex, education and baseline MMSE as predictors. b, Coefficients from each plasma-based model, with MMSE evaluated 4 years after the baseline as the outcome, in the ADNI and BioFINDER cohorts. Statistically significant variables, as determined by two-sided t-test on regression coefficients, are plotted with an asterisk instead of a square, and the error bars represent 95% confidence intervals. c, Estimated MMSE trajectories (colored lines), together with the estimated trajectories from the best-fitting model (P-tau181 and NfL; Aβ42/Aβ40 was not taken forward for assessing predictive performance), according to biomarker status, adjusted for age, sex, education and baseline MMSE. Error bars represent 95% confidence intervals on estimated trajectories. No corrections for multiple comparisons were performed.

Source data

With 4-year MMSE as the outcome in the ADNI model selection sample used for replication (n = 64), the full plasma model (R2 = 0.25; AIC = 310) fit the data better than the basic model (R2 = 0.15; AIC = 316; P = 0.01 compared with the full ATN model), and the best-fitting model according to the AIC again included plasma P-tau181 and NfL only (R2 = 0.25; AIC = 309). In the best-fitting model, the individual effect of P-tau181 was nearly significant (β = −0.64; P = 0.06) while the individual effect of NfL was significant (β = −1.02; P = 0.02) (Fig. 2a,b and Supplementary Table 1).

Model selection for clinical conversion

With 4-year conversion to AD as the outcome in the BioFINDER model selection sample (n = 107), the full plasma model that included all three biomarkers (area under the curve (AUC) = 0.88; AIC = 106) fit the data significantly better than the basic model (AUC = 0.70 (0.60–0.79); AIC = 140; P < 0.0001 compared with the full model). The best-fitting model according to the AIC included P-tau181 and NfL but not Aβ42/Aβ40, resulting in an AUC of 0.88 (0.82–0.95). A maximum specificity of 95.3% could be achieved given at least 50% sensitivity, and a maximum sensitivity of 96.8% could be achieved given at least 50% specificity (Fig. 3a,b and Supplementary Table 2). In the best-fitting model, there was a significant individual effect of P-tau181 (odds ratio (OR) = 5.87; P = 0.0001) but not NfL (OR = 1.73; P = 0.10). Using Cox regression (Supplementary Table 3), the model with P-tau181 and NfL was still the best-fitting model and there was a significant individual effect of both P-tau181 (hazard ratio (HR) = 2.52; P = 0.006) and NfL (HR = 2.70; P = 0.02).

Fig. 3: Modeling clinical conversion using plasma Aβ42/Aβ40, P-tau181 and NfL.
figure3

Results from modeling clinical conversion in patients with MCI using plasma biomarkers. a, AUC (x axis) and AIC values (numbers in plots) for each plasma-based model, with conversion to AD within 4 years after baseline as the outcome, in the BioFINDER and ADNI cohorts. The dashed vertical lines show basic model performances for reference. Error bars indicate 95% confidence intervals. All models also included age, sex, education and baseline MMSE as predictors. b, Coefficients from each plasma-based model, with conversion to AD within 4 years after the baseline as the outcome, in the BioFINDER and ADNI cohorts. Statistically significant variables, as determined by two-sided t-test on regression coefficients, are plotted with an asterisk instead of a square, and the error bars represent 95% confidence intervals. c, Estimated probability of not converting to AD, as predicted from the best-fitting model (P-tau181 and NfL; Aβ42/Aβ40 was not taken forward for assessing predictive performance), according to biomarker status, adjusted for age, sex, education and baseline MMSE. Error bars represent 95% confidence intervals on estimated trajectories.

Source data

With 4-year conversion to AD as the outcome in the ADNI model selection sample used for replication (n = 74), the full plasma model (AUC = 0.88 (0.80–0.98); AIC = 50) fit the data better than the basic model (AUC = 0.74 (0.52–0.95); AIC = 57; P = 0.005 compared with the full ATN model) and the best-fitting model according to AIC again included P-tau181 and NfL only. The AUC for this model was 0.89. A maximum specificity of 90.1% could be achieved given at least 50% sensitivity, and a maximum sensitivity of 100% could be achieved given at least 50% specificity (Fig. 3a,b and Supplementary Table 2). In the best-fitting model, the individual effect of P-tau181 was significant (OR = 4.58; P = 0.009) while the individual effect of NfL was not significant (OR = 2.15; P = 0.20). Using Cox regression (Supplementary Table 3), the model with P-tau181 and NfL was again still the best-fitting model, with a significant effect in this model for P-tau181 (HR = 2.22; P < 0.0001) but not NfL (HR = 1.27; P = 0.13). Receiver operating characteristic (ROC) curves for the main models are presented in Fig. 4. ROC curves for all of the models in the model selection analysis with 4-year conversion to AD dementia as the outcome are presented in Extended Data Fig. 4.

Fig. 4: ROC curves for clinical conversion across cohorts.
figure4

a,b, ROC curves from modeling clinical conversion in patients with MCI using plasma biomarkers in the BioFINDER (a) and ADNI (b) cohorts. The basic model (age, sex, education and baseline MMSE) is shown along with the best-performing model selected for further prognostic validation (basic model + plasma P-tau181 + plasma NfL (plasma TN)). Results from the full CSF model (basic model + CSF Aβ42/Aβ40 + CSF P-tau181 + CSF NfL (CSF ATN)) are also shown for the BioFINDER cohort. These results are not shown for the ADNI cohort as the full CSF model was not available (but note that the CSF TN model for ADNI was tested at the prognostic validation stage).

Effect of the APOE ε4 genotype on model selection

Including APOE ε4 genotype status (that is, ε4 carrier versus non-carrier) in the basic model did not substantially affect model selection for the primary cognitive outcome (Supplementary Table 4) or for the primary clinical outcome (Supplementary Table 5). Moreover, the full plasma ATN model still outperformed the basic model for both the primary cognitive and clinical outcomes.

Sensitivity analysis using secondary and exploratory outcomes

We performed the same model selection procedure as outlined above using the secondary and exploratory outcomes. We found that the best-fitting models identified here varied across outcomes, whereas the full plasma model including Aβ42/Aβ40, P-tau181 and NfL still always outperformed the basic model alone (Supplementary Tables 4–13).

Sensitivity analysis using an alternative plasma Aβ42/Aβ40 assay

Because plasma Aβ42/Aβ40 was not selected as a predictor in the best-fitting models for the co-primary outcomes, we tested whether this result differed when using a mass spectrometry assay (Araclon Biotech) instead of the Elecsys assay used in the BioFINDER cohort. Here, the best-fitting models according to AIC when using the mass spectrometry plasma Aβ42/Aβ40 assay still did not include Aβ42/Aβ40. Instead, the best-fitting model with 4-year MMSE as the outcome included P-tau181 and NfL only and the best-fitting model with 4-year conversion to AD as the outcome included P-tau181 only (Supplementary Tables 12 and 13).

Sensitivity analysis using plasma P-tau217 in the BioFINDER cohort

Primary analyses (that is, 4-year MMSE and 4-year conversion to AD) were repeated using plasma P-tau217 (see Palmqvist et al.18 for an assay description) instead of P-tau181. As plasma P-tau217 data are not available for ADNI, these analyses were performed using BioFINDER participants only. Findings from these analyses replicated those found when using P-tau181 (Supplementary Tables 14 and 15): P-tau in combination with NfL remained the best model for both outcomes, and R2 and AUC values were similar when using P-tau217 or P-tau181 (for MMSE, compare Supplementary Table 14 with Supplementary Table 1; for conversion to AD, compare Supplementary Table 15 with Supplementary Table 2).

Study population characteristics (prognostic validation sample)

Since the model that included both plasma P-tau181 and NfL (but not Aβ42/Aβ40) provided the best fit across co-primary outcomes in both cohorts, this model was taken forward in the prognostic validation stage. For this analysis, we therefore only required participants to have available data on plasma P-tau181 and NfL measurements and at least one of the primary or secondary outcomes. As such, 148 patients with MCI from the BioFINDER cohort (no difference from the BioFINDER model selection sample) and 425 patients with MCI from the ADNI cohort (Table 2; see Extended Data Fig. 5 for the association between plasma P-tau181 and NfL in this group) were included.

Table 2 Study participant characteristics in ADNI (prognostic validation)

Patient-level prognostic validation within cohorts

We performed an internal validation where the patient-level predictive performance of the best-fitting plasma model (the basic model plus plasma P-tau181 and NfL) was evaluated within each cohort and compared with the basic model alone and with the full CSF model (the basic model plus CSF Aβ42/Aβ40, P-tau181 and NfL).

With 4-year MMSE as the outcome in the BioFINDER prognostic validation sample (n = 118), the best-fitting plasma model improved cross-validated, out-of-sample prediction compared with the basic model alone (mean absolute error (MAE) = 3.07 points versus 3.36 points; P < 0.001; 8.5% improvement) and showed no significant difference compared with the full CSF model (P = 0.68 over 1,000 bootstrapped trials). With 4-year MMSE as the outcome within the ADNI prognostic validation sample (n = 252), the same best-fitting plasma model improved out-of-sample prediction compared with the basic model alone (MAE = 2.42 points versus 2.49 points; P < 0.001; 2.9% improvement).

With 4-year conversion to AD as the outcome in the BioFINDER prognostic validation sample (n = 107), the best-fitting plasma model improved out-of-sample prediction compared with the basic model alone (AUC = 0.63 versus 0.83 (note that these cross-validated AUCs are, as expected, lower than the AUCs from corresponding models fit on all data for model selection); P < 0.001; 31.7% improvement) and significantly outperformed the full CSF model (P = 0.002 over 1,000 bootstrapped trials; 5% improvement). With 4-year conversion to AD as the outcome in the ADNI prognostic validation sample (n = 320), the plasma model significantly improved out-of-sample prediction compared with the basic model alone (AUC = 0.66 versus 0.76; P < 0.001; 15.4% improvement). For a visual depiction of individualized predicted probabilities of conversion to AD across models, see Fig. 5.

Fig. 5: Individualized prediction of 4-year conversion from MCI to AD dementia.
figure5

a,b, Results from internal cross-validation for clinical conversion for the best-performing models, as identified in the first stage of analysis using all available BioFINDER (a; n = 107) and ADNI (b; n = 320) patients. The box plots show the minimum and maximum values (whiskers), as well as the median (central line) and interquartile range (that is, the first quartile (25th percentile; bottom edge of box) and third quartile (75th percentile; top edge of box)). The values plotted here show the predicted probability of conversion from MCI to AD dementia for each individual in the BioFINDER and ADNI cohorts, showing 32.3% improvement (AUC = 0.62 for the basic model versus AUC = 0.82 for the P-tau181 and NfL model) of the plasma-based model over the basic model in BioFINDER and a 15.4% improvement (AUC = 0.66 for the basic model versus AUC = 0.76 for the P-tau181 and NfL model) in ADNI.

Patient-level prognostic validation across cohorts

We performed an external validation whereby the patient-level predictive performance of the best-fitting plasma model (the basic model plus plasma P-tau181 and NfL) was evaluated across each cohort by first fitting the model on BioFINDER participants and then testing on ADNI participants, and vice versa. For this analysis, biomarkers were dichotomized.

With 4-year MMSE as the outcome in the BioFINDER (n = 118, of whom 28 were TN, 13 were TN+, 46 were T+N and 31 were T+N+) and ADNI prognostic validation samples (n = 243, of whom 118 were TN, 35 were TN+, 46 were T+N and 44 were T+N+), the plasma model significantly improved prediction on the test cohort compared with the basic model, both when the model was fit on BioFINDER and tested on ADNI (MAE = 3.74 versus 4.08; P = 0.0006; 8.3% improvement) and when the model was fit on ADNI and tested on BioFINDER (MAE = 4.15 versus 5.19; P < 0.0001; 20.1% improvement).

With 4-year conversion to AD as the outcome in the BioFINDER (n = 107; of whom 20 were TN, five were TN+, 49 were T+N and 33 were T+N+) and ADNI prognostic validation samples (n = 314, of whom 139 were TN, 45 were TN+, 62 were T+N and 68 were T+N+), the plasma model improved prediction on the unseen cohort both when the model was fit on ADNI and tested on BioFINDER (AUC = 0.61 versus 0.79; P < 0.0001; 29.3% improvement) and when the model was fit on BioFINDER and tested on ADNI (AUC = 0.62 versus 0.73; P < 0.0001; 18.3% improvement).

Online individualized risk prediction tool

We provide an illustrative online tool for use with the current dataset at predictprogression.com, where individualized predictions can be made for MMSE, conversion to AD dementia and clinical dementia rating scale–sum of boxes (CDR-SB) score, at 2 and 4 years after the baseline, in patients with MCI at the baseline. The tool allows the user to enter data on age, sex, baseline cognition (MMSE and/or CDR-SB) and dichotomous biomarker status for CSF or plasma Aβ42/Aβ40, P-tau181 and NfL. It is also possible to test predictions with sparse models including subsets of biomarkers. For example, for a 70-year-old female with MCI and a baseline MMSE score of 27, if all plasma Aβ42/Aβ40, P-tau181 and NfL values are known and negative, the probabilities are 6% (90% prediction interval = 2–20%) at 2 years and 16% (90% prediction interval = 5–38%) at 4 years (Extended Data Fig. 6). If all plasma Aβ42/Aβ40, P-tau181 and NfL values are positive, the probabilities change to 43% (90% prediction interval = 25–62%) at 2 years and 92% (90% prediction interval = 77–97%) at 4 years (Extended Data Fig. 6). In the absence of biomarker information, the predicted probability of conversion to AD is 33% (90% prediction interval = 23–45%) at 2 years and 69% (90% prediction interval = 56–80%) at 4 years.

Discussion

We addressed the patient-level prognostic value of plasma AD biomarkers (Aβ42/Aβ40, P-tau181 and NfL) in MCI. Plasma P-tau181 in combination with NfL best predicted primary outcomes of decline in MMSE score and clinical progression to AD dementia over 4 years. These results were robust to the time horizon (2 or 4 years of follow-up), selection of outcome (MMSE, CDR-SB, conversion to AD dementia or conversion to all-cause dementia), two different cohorts and choice of Aβ assay. In general, prognostic performance using the plasma-based models was either non-inferior or even better than when using CSF biomarkers, and better than a basic model including age, sex, education and baseline cognition. These biomarker-driven prediction models can be applied using our online tool for accurate individualized prognosis in MCI (that is, to predict, for a given patient, both what their MMSE score will be 4 years after their baseline visit and their percentage risk of progressing from MCI to AD dementia over the same time interval). This tool might improve treatment and care21 and could increase power for clinical trials for prodromal AD by only including those with a high risk of future progression.

Our study is novel in the way we address the individualized predictive value of plasma AD biomarkers, but it can be compared with previous work examining CSF and imaging biomarker-driven prognosis at the MCI stage20. Using four separate prognostic models—including age, sex, CSF Aβ42, T-tau and MMSE, as well as an ATN variant combining CSF Aβ42 and P-tau181 with hippocampal volume—van Maurik and colleagues20 looked at the likelihood of progression to dementia from MCI. While all models performed well, the highest performance was seen using the CSF ATN model. Similarly, we found that a combined ATN model (Aβ42/Aβ40, P-tau181 and NfL) in plasma outperformed a basic model with demographics and baseline MMSE score as predictors. Importantly, plasma models improved the prediction of longitudinal MMSE despite adjustment for MMSE score at the baseline, which in itself is a very strong predictor of future MMSE score.

Inclusion of P-tau181 in the best models may reflect that P-tau181 detects AD-type changes11. In contrast, plasma Aβ42/Aβ40 was not included in the best model, suggesting that plasma Aβ biomarkers do not provide additional prognostic information in MCI when an efficient plasma P-tau measure is included. This is logical, since symptoms in AD are linked to tau pathology22, and elevations in tau biomarkers appear to be dependent on Aβ pathology23,24. Findings for plasma Aβ42/Aβ40 have also been more varied than for plasma P-tau181 (refs. 10,25) and have only shown modest reductions in AD dementia (10–15% compared with cognitively unimpaired), while P-tau181 is greatly increased in AD dementia (>100% compared with cognitively unimpaired)11,12,13. However, it is possible that plasma Aβ42/Aβ40 may have added value at the preclinical stage of the disease, when it has reached pathological levels10, while tau and neurodegeneration markers continue to increase during the symptomatic stages of the disease11,26. The best models also included NfL, which is a more general marker of neurodegeneration17 and appears to give prognostic information complimentary to P-tau181.

In addition to the CSF studies by van Maurik on individualized biomarker-based risk predictions of dementia in patients with MCI19,20, recent work from our group has examined the association between plasma-based biomarkers and the risk of AD dementia11. Although similar in terms of including plasma Aβ42/Aβ40, P-tau181 and NfL, the present study differs from our previous work in a number of important ways. We now focus on identifying optimal models within the ATN framework, rather than individual biomarkers, and on patient-level (and not group-level) predictions. We also compare plasma-based models with a more basic model without biomarkers (but with baseline MMSE score) and with CSF-based models. We performed extensive internal and external validation analyses (including novel plasma P-tau181 measurements from the ADNI cohort). In terms of results, one important difference compared with the previous study11 is the finding that both NfL and P-tau181 (rather than just P-tau181) contribute to the best-performing models. Although little difference in AIC was seen between plasma models with and without Aβ42/Aβ40, we opted not to include this biomarker as we aimed to select the most parsimonious model (that is, the one performing best or as good as all the others but including the least number of biomarkers).

Although the relative importance of biomarkers may vary across contexts and intended applications, plasma biomarkers are promising due to their high accessibility and low cost. With respect to the potential future clinical implementation of plasma biomarkers, the use of binarized data (that is, abnormal versus normal) will probably prove much easier to implement as different assays will have different cut-offs; moreover, different cut-offs may even be required across different laboratories, even when using the same assay, due to local differences in how plasma is collected, handled and analyzed. That said, in the longer term, assays will need to be standardized so that the same values are obtained even when samples are measured in different laboratories; this represents a considerable task, based on the field’s experience with CSF AD biomarkers27.

The strengths of our study include the use of CSF-based ATN models as an internal performance benchmark and the focus on risk estimates at the participant level. The results were robust across different assays used to measure plasma P-tau181 in the BioFINDER and ADNI cohorts. Validation in two independent cohorts with greatly differing demographic makeup speaks to the robustness and relevance of our findings. BioFINDER patients were recruited in a consecutive fashion at three different memory clinics, with approximately 90% of these referred by primary care physicians. ADNI patients were recruited from many different clinics and may be more representative of a highly selected clinical trial population. Interestingly, the additional effect that including plasma biomarkers had on individualized prognostic performance was higher in the BioFINDER cohort than in the ADNI cohort. This phenomenon is most likely explained by the differences in the demographic makeup of the cohorts, as previously explained. Sensitivity analyses showed that our results were robust across clinical outcomes and for the method used to measure plasma Aβ42/Aβ40.

We showed that including the APOE ε4 genotype in the basic model did not dampen the effect of including plasma biomarkers. This is a promising result given the ethical difficulties in disclosing genetic status to patients, particularly as it relates to patients’ relatives. Future work will be focused on understanding whether APOE ε4 allele carriers can especially benefit from plasma biomarker measurement or whether current plasma ATN biomarkers can largely replace the need for genetic testing, as indicated by the present results.

This study has limitations. First, our sample size was relatively modest; therefore, significant differences between models are difficult to establish. We relied on selecting the most parsimonious model where differences in AIC values were small (when the models perform similarly, we believe that a less complex model with fewer biomarker predictors is preferable). Further studies on larger and more diverse populations, including in primary care, may result in more precise and generalizable models. Second, as data on plasma P-tau217 were not available from the ADNI cohort, the sensitivity analysis for primary outcomes replacing P-tau181 with P-tau217 was only performed for the BioFINDER cohort. Although plasma P-tau217 has recently been shown to outperform P-tau181 for discriminating AD dementia from non-AD conditions18, there are, as yet, no data indicating that P-tau217 is superior to P-tau181 for predicting progression to AD in MCI. As our sensitivity analysis resulted in findings that were quite similar to those using P-tau181, P-tau217 might not be a clearly better predictor of future cognitive decline than P-tau181 in individuals with MCI. Third, the finding that the plasma-based model outperformed the CSF-based model when using 4-year conversion to AD—but not when using 4-year MMSE score—as the outcome requires replication. Possibly, this finding could relate to our sample size or to changes to the blood–brain barrier, rendering plasma P-tau181 more specific for AD pathology than CSF P-tau181. It is also possible that variability in CSF production and turnover could differentially impact the performance of CSF P-tau181 versus plasma P-tau181. This too requires further study.

Lastly, as indicated above, our online prediction tool is specific to the dataset we examined and was included only to illustrate the possible future use of such a tool in clinical settings. Further studies are needed to address its performance using data from other cohorts.

In conclusion, plasma-based AD biomarkers can provide patient-level prognostic information in MCI, comparable to CSF biomarkers. Plasma P-tau181 in combination with NfL seems to best predict cognitive decline and clinical progression. Plasma biomarkers of core AD features may aid in individualized risk assessment for patients with MCI, which represents a critical step towards accessible precision medicine for cognitive diseases. Standardized assays with universal cut-offs and replication of the findings in large cohorts are needed.

Methods

This study was conducted in accordance with Standards for Reporting of Diagnostic Accuracy (STARD) guidelines. The STARD checklist can be found as part of the Supplementary Information.

Participants

In this prospective study, we studied consecutively enrolled patients with MCI from the Swedish BioFINDER cohort (clinical trial number NCT01208675; www.biofinder.se). The patients were recruited and evaluated at memory clinics in the cities of Lund, Malmö and Ängelholm between July 2008 and June 2019. They were between 60 and 80 years old and fulfilled the consensus criteria for MCI suggested by Petersen28 (including: cognitive complaints, preferably corroborated by an informant; objective cognitive impairment, adjusted for age and education; preservation of general cognitive functioning and an MMSE score of 24–30; and no or minimal impairment of daily life activities) but did not fulfill criteria for dementia, as described previously in detail29. Exclusion criteria included cognitive impairment that could better be accounted for by another non-neurodegenerative condition, severe somatic disease and current alcohol or substance abuse. After their baseline visit, all patients were seen annually, to assess clinical progression.

For validation, data were obtained from patients with MCI in the ADNI database (the convenience sample; clinical trial number NCT00106899; adni.loni.usc.edu). The ADNI was launched in 2003 as a public–private partnership led by principal investigator M. W. Weiner. For up-to-date information, see www.adni-info.org. The data included in the present study were collected between September 2005 and December 2019.

All participants gave written informed consent. For BioFINDER, ethical approval was given by the Regional Ethical Committee of Lund University. Ethical approval in relation to the ADNI cohort was given by the local ethical committees of all of the involved sites (see Supplementary Notes for a complete list). For both cohorts, no data points were excluded from the analyses. As both BioFINDER and ADNI are observational cohort studies, allocation of participants into experimental groups was not performed (that is, there was no randomization). All analyses were adjusted for age, sex, education and baseline cognition.

Outcomes

The co-primary outcomes were the global cognitive measure MMSE and clinical conversion to AD dementia, evaluated 4 years after the baseline (4-year MMSE and 4-year conversion to AD, respectively). Clinicians who evaluated cognitive decline and conversion to dementia were blinded to biomarker data. As secondary outcomes, we used 2-year MMSE and 2-year conversion to AD dementia. As exploratory outcomes, we used 2-year and 4-year CDR-SB along with 2-year and 4-year conversion to dementia due to any cause. Outcomes were selected on the basis of their clinical relevance.

In BioFINDER, clinical status of dementia due to AD or other diseases was evaluated according to Diagnostic and Statistical Manual of Mental Disorders (version 5) criteria for major neurocognitive disorder (that is, dementia). Dementia in ADNI was defined using the National Institute of Neurological and Communicative Disorders and Stroke and the Alzheimer’s Disease and Related Disorders Association criteria for probable AD30.

Predictors

All models included age, sex, education and baseline MMSE score as predictors (the basic model). We also measured Aβ42/Aβ40, P-tau181 and NfL in both CSF and plasma. These analyses were performed by technicians blinded to the clinical data.

For the BioFINDER cohort, plasma Aβ42/Aβ40 was measured using Elecsys immunoassays on a Cobas e601 analyzer (Roche Diagnostics)10. A sensitivity analysis was performed using a mass spectrometry-based plasma Aβ42/Aβ40 assay (Araclon Biotech). Plasma P‐tau181 was measured on a Meso Scale Discovery platform using an assay developed by Eli Lilly11. Plasma NfL was analyzed using a single molecule array (Simoa)-based assay17. Moreover, CSF levels of Aβ42 (used in place of Aβ42/Aβ40 due to no available Aβ40 data for the ADNI cohort) and P-tau181 were measured using Elecsys assays (Roche Diagnostics), while CSF NfL was measured using an enzyme-linked immunosorbent assay (UmanDiagnostics).

For the ADNI cohort, plasma Aβ42/Aβ40 was analyzed using an immunoprecipitation and mass spectrometry-based method31. P-tau181 was analyzed on a Simoa HD-X Analyzer (Quanterix) using an assay developed in the Clinical Neurochemistry Laboratory, University of Gothenburg, Sweden12. Plasma NfL was analyzed using the same Simoa-based assay as was used for the BioFINDER cohort.

All biomarker values were natural log transformed. Biomarkers were binarized when validating models across cohorts, whereby cut-offs were defined using Youden’s index to maximize the separation between Aβ-negative cognitively unimpaired participants and Aβ-positive patients with AD dementia from within each cohort; these participants have been described previously11,12. We derived cut-offs separately for plasma and CSF biomarkers in the BioFINDER cohort, and for plasma biomarkers only in the ADNI cohort (see Supplementary Table 16). For the BioFINDER cohort, there were 528 cognitively unimpaired participants (age = 72.5 years (s.d. = 5.5 years); education = 12.3 years (s.d = 3.6 years); sex = 57% female) and 81 patients with AD (age = 76.7 years (s.d = 5.1 years); education = 9.7 years (s.d = 3.0 years); sex = 59% female). For the ADNI cohort, there were 126 cognitively unimpaired participants (age = 71.5 years (s.d = 6.5 years); education = 16.1 years (s.d. = 2.9 years); sex = 41% female) and 106 patients with AD (age = 73.8 years (s.d = 8.3 years); education = 15.7 years (s.d = 2.7 years); sex = 58% female). Note that none of the participants used to define cut-offs were used in the statistical analysis.

Statistical analysis

In the first analysis stage (model selection), different linear regression models were fit with the cognitive outcomes described above as response variables: a basic model (age, sex, education and baseline MMSE) and plasma biomarker models (the basic model plus seven different biomarker combinations including: Aβ42/Aβ40 only; P-tau181 only; NfL only; Aβ42/Aβ40 and P-tau181; Aβ42/Aβ40 and NfL; P-tau181 and NfL; or all three biomarkers). Because APOE ε4 is the strongest genetic risk factor for AD, we tested whether the addition of APOE ε4 genotype status (represented as a binary variable split based on individuals with at least one ε4 allele) to the basic model reduced the effectiveness of using plasma biomarkers. Models were compared using R2 and AIC values (lower is better). The best-fitting model was that which included the fewest predictors among the models within two points of the lowest AIC value; this procedure is well established for selecting the most parsimonious model based on AIC values32,33. The statistical significance of different models with the same outcome variable was assessed using the likelihood ratio test. Additionally, logistic regression models were fit with the clinical conversion outcomes described above as response variables, with the same set of predictors and the same method of comparison but with AUC instead of R2 as the performance metric; a sensitivity analysis was performed using Cox regression models, to ensure that the timing of conversion to AD did not affect model selection.

In the second analysis stage (prognostic validation), the best-fitting model identified in the first stage was carried forward and its predictive accuracy was evaluated. Prognostic validation was first done separately within each cohort using 1,000 repetitions of fivefold cross-validation (internal validation), and then by fitting the model on BioFINDER participants and testing on ADNI participants, and vice versa (external validation; all biomarkers were dichotomized for this analysis, to compare across assays). For internal validation in the BioFINDER cohort, the best-fitting plasma model was compared with a CSF model that included CSF Aβ42/Aβ40, P-tau181 and NfL. Prognostic performance was evaluated using MAE with cognitive outcomes, which represents the absolute deviation in MMSE score predicted by the model (for example, MAE = 3 means that the model predicts an individual’s cognitive value to within three points, on average) and using AUC with clinical conversion outcomes.

In the model selection stage, only participants with all three plasma biomarkers available were included. In the prognostic validation stage, participants were only required to have measurements from the plasma biomarkers included in the best-fitting model. Both Q–Q plots and normality of residuals were visually inspected for primary (basic and full ATN) regression models. No statistical methods were used to pre-determine sample sizes but our sample sizes are similar to those reported in previous publications11,19. All analyses were performed using the R programming language (version 4.0.0), with significance set at P < 0.05 (two sided).

Reporting Summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

Plasma and CSF data from ADNI were downloaded from https://ida.loni.usc.edu. Anonymized data from the BioFINDER study will be shared by request from a qualified academic investigator for the sole purpose of replicating procedures and results presented in this Article and as long as the data transfer is in agreement with European Union legislation on general data protection regulation and decisions by the Ethical Review Board of Sweden and Region Skåne, which should be regulated by a material transfer agreement.

Code availability

The code that support the findings of this study is available from the corresponding author upon request. All models were built using publicly available packages and functions in the R programming language.

References

  1. 1.

    World Alzheimer Report 2019 (Alzheimer’s Disease International, 2019).

  2. 2.

    Winblad, B. et al. Defeating Alzheimer’s disease and other dementias: a priority for European science and society. Lancet Neurol. 15, 455–532 (2016).

    Article  Google Scholar 

  3. 3.

    Burnham, S. C. et al. The dawn of robust individualised risk models for dementia. Lancet Neurol. 18, 985–987 (2019).

    Article  Google Scholar 

  4. 4.

    Abbasi, J. Promising results in 18-month analysis of Alzheimer drug candidate. J. Am. Med. Assoc. 320, 965 (2018).

    Google Scholar 

  5. 5.

    Buchhave, P. et al. Cerebrospinal fluid levels of β-amyloid 1–42, but not of tau, are fully changed already 5 to 10 years before the onset of Alzheimer dementia. Arch. Gen. Psychiatry 69, 98–106 (2012).

    CAS  Article  Google Scholar 

  6. 6.

    Hansson, O. et al. CSF biomarkers of Alzheimer’s disease concord with amyloid-β PET and predict clinical progression: a study of fully automated immunoassays in BioFINDER and ADNI cohorts. Alzheimers Dement. 14, 1470–1481 (2018).

    Article  Google Scholar 

  7. 7.

    Rabinovici, G. D. et al. Association of amyloid positron emission tomography with subsequent change in clinical management among Medicare beneficiaries with mild cognitive impairment or dementia. J. Am. Med. Assoc. 321, 1286–1294 (2019).

    Article  Google Scholar 

  8. 8.

    Ossenkoppele, R. et al. Discriminative accuracy of [18F]flortaucipir positron emission tomography for Alzheimer disease vs other neurodegenerative disorders. J. Am. Med. Assoc. 320, 1151–1162 (2018).

    CAS  Article  Google Scholar 

  9. 9.

    Jack, C. R. Jr et al. NIA-AA research framework: toward a biological definition of Alzheimer’s disease. Alzheimers Dement. 14, 535–562 (2018).

    Article  Google Scholar 

  10. 10.

    Palmqvist, S. et al. Performance of fully automated plasma assays as screening tests for Alzheimer disease-related β-amyloid status. JAMA Neurol. 76, 1060–1069 (2019).

    Article  Google Scholar 

  11. 11.

    Janelidze, S. et al. Plasma P-tau181 in Alzheimer’s disease: relationship to other biomarkers, differential diagnosis, neuropathology and longitudinal progression to Alzheimer’s dementia. Nat. Med. 26, 379–386 (2020).

    CAS  Article  Google Scholar 

  12. 12.

    Karikari, T. K. et al. Blood phosphorylated tau 181 as a biomarker for Alzheimer’s disease: a diagnostic performance and prediction modelling study using data from four prospective cohorts. Lancet Neurol. 19, 422–433 (2020).

    CAS  Article  Google Scholar 

  13. 13.

    Thijssen, E. H. et al. Diagnostic value of plasma phosphorylated tau181 in Alzheimer’s disease and frontotemporal lobar degeneration. Nat. Med. 26, 387–397 (2020).

    CAS  Article  Google Scholar 

  14. 14.

    Zetterberg, H. Neurofilament light: a dynamic cross-disease fluid biomarker for neurodegeneration. Neuron 91, 1–3 (2016).

    CAS  Article  Google Scholar 

  15. 15.

    Quiroz, Y. T. et al. Plasma neurofilament light chain in the presenilin 1 E280A autosomal dominant Alzheimer’s disease kindred: a cross-sectional and longitudinal cohort study. Lancet Neurol. 19, 513–521 (2020).

    CAS  Article  Google Scholar 

  16. 16.

    Mielke, M. M. et al. Plasma phospho-tau181 increases with Alzheimer’s disease clinical severity and is associated with tau- and amyloid-positron emission tomography. Alzheimers Dement. 14, 989–997 (2018).

    Article  Google Scholar 

  17. 17.

    Mattsson, N., Cullen, N. C., Andreasson, U., Zetterberg, H. & Blennow, K. Association between longitudinal plasma neurofilament light and neurodegeneration in patients with Alzheimer disease. JAMA Neurol. 76, 791–799 (2019).

    Article  Google Scholar 

  18. 18.

    Palmqvist, S. et al. Discriminative accuracy of plasma phospho-tau217 for Alzheimer disease vs other neurodegenerative disorders. J. Am. Med. Assoc. 324, 772–781 (2020).

    CAS  Article  Google Scholar 

  19. 19.

    Van Maurik, I. S. et al. Interpreting biomarker results in individual patients with mild cognitive impairment in the Alzheimer’s Biomarkers In Daily Practice (ABIDE) project. JAMA Neurol. 74, 1481–1491 (2017).

    Article  Google Scholar 

  20. 20.

    Van Maurik, I. S. et al. Biomarker-based prognosis for people with mild cognitive impairment (ABIDE): a modelling study. Lancet Neurol. 18, 1034–1044 (2019).

    Article  Google Scholar 

  21. 21.

    Gomersall, T., Smith, S. K., Blewett, C. & Astell, A. ‘It’s definitely not Alzheimer’s’: perceived benefits and drawbacks of a mild cognitive impairment diagnosis. Br. J. Health Psychol. 22, 786–804 (2017).

    Article  Google Scholar 

  22. 22.

    Ossenkoppele, R. et al. Associations between tau, Aβ, and cortical thickness with cognition in Alzheimer disease. Neurology 92, e601–e612 (2019).

    CAS  Article  Google Scholar 

  23. 23.

    Mattsson-Carlgren, N. et al. Aβ deposition is associated with increases in soluble and phosphorylated tau that precede a positive Tau PET in Alzheimer’s disease. Sci. Adv. 6, eaaz2387 (2020).

    CAS  Article  Google Scholar 

  24. 24.

    Leuzy, A. et al. Diagnostic performance of RO948 F 18 Tau positron emission tomography in the differentiation of Alzheimer disease from other neurodegenerative disorders. JAMA Neurol. 77, 955–965 (2020).

    Article  Google Scholar 

  25. 25.

    Doecke, J. D. et al. Total Aβ42/Aβ40 ratio in plasma predicts amyloid-PET status, independent of clinical AD diagnosis. Neurology 94, e1580–e1591 (2020).

    CAS  Article  Google Scholar 

  26. 26.

    Mattsson, N., Andreasson, U., Zetterberg, H. & Blennow, K. Alzheimer’s Disease Neuroimaging Initiative Association of plasma neurofilament light with neurodegeneration in patients with Alzheimer disease. JAMA Neurol. 74, 557–566 (2017).

    Article  Google Scholar 

  27. 27.

    Kuhlmann, J. et al. CSF Aβ1–42—an excellent but complicated Alzheimer’s biomarker—a route to standardisation. Clin. Chim. Acta 467, 27–33 (2017).

    CAS  Article  Google Scholar 

  28. 28.

    Petersen, R. C. Mild cognitive impairment as a diagnostic entity. J. Intern. Med. 256, 183–194 (2004).

    CAS  Article  Google Scholar 

  29. 29.

    Petrazzuoli, F. et al. Brief cognitive tests used in primary care cannot accurately differentiate mild cognitive impairment from subjective cognitive decline. J. Alzheimers Dis. 75, 1191–1201 (2020).

    Article  Google Scholar 

  30. 30.

    McKhann, G. et al. Clinical diagnosis of Alzheimer’s disease: report of the NINCDS-ADRDA Work Group under the auspices of Department of Health and Human Services Task Force on Alzheimer’s Disease. Neurology 34, 939–944 (1984).

    CAS  Article  Google Scholar 

  31. 31.

    Ovod, V. et al. Amyloid β concentrations and stable isotope labeling kinetics of human plasma specific to central nervous system amyloidosis. Alzheimers Dement. 13, 841–849 (2017).

    Article  Google Scholar 

  32. 32.

    Akaike, H. Likelihood of a model and information criteria. J. Econom. 16, 3–14 (1981).

    Article  Google Scholar 

  33. 33.

    Burnham K. P. & Anderson D. R. Model Selection and Multimodel Inference (Springer, 2002).

Download references

Acknowledgements

Work at the authors’ laboratory at Lund University was supported by the Swedish Research Council, the Wallenberg Center for Molecular Medicine, the Knut and Alice Wallenberg Foundation, Lund University’s Medical Faculty, Region Skåne, the Marianne and Marcus Wallenberg Foundation, the Strategic Research Area MultiPark (Multidisciplinary Research focused on Parkinson’s Disease) at Lund University, the Swedish Alzheimer’s Foundation, the Swedish Brain Foundation, the Swedish Medical Association, the Konung Gustaf V:s och Drottning Victorias Frimurarestiftelse, the Bundy Academy, the Skåne University Hospital Foundation and the Swedish federal government under the ALF agreement. Data collection and sharing for the ADNI part of the study was funded by the ADNI (National Institutes of Health Grant U01 AG024904) and DOD ADNI (Department of Defense award number W81XWH-12-2-0012). The ADNI is funded by the National Institute on Aging and the National Institute of Biomedical Imaging and Bioengineering, and through generous contributions from the following: AbbVie; Alzheimer’s Association; Alzheimer’s Drug Discovery Foundation; Araclon Biotech; Bioclinica; Biogen; Bristol Myers Squibb; CereSpir; Cogstate; Eisai; Elan Pharmaceuticals; Eli Lilly and Company; EuroImmun; F. Hoffmann-La Roche and its affiliated company Genentech; Fujirebio; GE Healthcare; IXICO; Janssen Alzheimer Immunotherapy Research and Development; Johnson & Johnson Pharmaceutical Research and Development; Lumosity; Lundbeck; Merck & Co.; Meso Scale Diagnostics; NeuroRx Research; Neurotrack Technologies; Novartis Pharmaceuticals Corporation; Pfizer; Piramal Imaging; Servier; Takeda Pharmaceutical Company; and Transition Therapeutics. The Canadian Institutes of Health Research is providing funds to support ADNI clinical sites in Canada. Private sector contributions are facilitated by the Foundation for the National Institutes of Health (www.fnih.org). The grantee organization is the Northern California Institute for Research and Education, and the study is coordinated by the Alzheimer’s Therapeutic Research Institute at the University of Southern California. ADNI data are disseminated by the Laboratory for Neuro Imaging at the University of Southern California.

Author information

Affiliations

Authors

Contributions

A.L., N.M.-C. and O.H. conceived of the study. A.L. and N.C.C. performed the statistical analysis. E.S., S.P. and O.H. recruited participants and collected clinical data. H.Z., L.S., J.A.A., J.L.D., K.B., N.K.P., P.P. and S.J. were responsible for the biochemical analyses. N.M.-C. developed an online tool implementing the statistical models. A.L., N.M.-C., L.S., J.A.A., P.P., N.M.C. and O.H. drafted the initial manuscript. All authors contributed to revision and editing of the manuscript.

Corresponding authors

Correspondence to Niklas Mattsson-Carlgren or Oskar Hansson.

Ethics declarations

Competing interests

N.C.C., A.L., S.P., S.J. and N.M.-C. declare no competing interests. P.P., L.S. and J.A.A. are current employees of Araclon Biotech. N.K.P. and J.L.D. are employees and stockholders of Eli Lilly and Company. H.Z. has served on scientific advisory boards for Denali, Roche Diagnostics, Wave, Samumed, Siemens Healthineers, Pinteon Therapeutics and CogRx and has delivered lectures in symposia sponsored by Fujirebio, AlzeCure and Biogen. K.B. has served as a consultant on advisory boards or data monitoring committees for Abcam, Axon, Biogen, JOMDD/Shimadzu, Julius Clinical, Lilly, MagQu, Novartis, Roche Diagnostics and Siemens Healthineers and is a co-founder of Brain Biomarker Solutions in Gothenburg, which is a part of the GU Ventures Incubator Program. O.H. has acquired research support (for the institution) from Roche, Pfizer, GE Healthcare, Biogen, Eli Lilly and AVID Radiopharmaceuticals. In the past 2 years, he has received consultancy/speaker fees (paid to the institution) from Biogen and Roche. E.S. has no competing interests.

Additional information

Peer review information Nature Aging thanks Suzanne Schindler, Charlotte Teunissen and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Association between plasma biomarkers in MCI patients in the BioFINDER model selection/prognostic validation cohort.

This figure shows the association between the different possible combinations of ATN plasma biomarkers (left to right: a, plasma P-tau181 vs plasma Aβ42/Aβ40; b, plasma NfL vs plasma P-tau181; and c, plasma Aβ42/Aβ40 vs plasma NfL). Each filled circle represents a single MCI patient from the BioFINDER cohort (n=148); the histograms above the scatterplots show the distribution of concentration values for the plasma biomarker on the x-axis (that is panel A, Aβ42/Aβ40; panel B, P-tau181; panel C, NfL). The blue lines (solid and dashed) indicate the cutoff values for each biomarker.

Extended Data Fig. 2 Association between CSF biomarkers in MCI patients in the BioFINDER model selection/prognostic validation cohort.

This figure shows the association between the different possible combinations of ATN CSF biomarkers (left to right: a, CSF P-tau181 vs CSF Aβ42/Aβ40; b, CSF NfL vs CSF P-tau181; and c, CSF Aβ42/Aβ40 vs CSF NfL). Each filled circle represents a single MCI patient from the BioFINDER cohort (n=148); the histograms above the scatterplots show the distribution of concentration values for the plasma biomarker on the x-axis (that is panel A, Aβ42/Aβ40; panel B, P-tau181; panel C, NfL).

Extended Data Fig. 3 Association between plasma biomarkers in MCI patients in the ADNI model selection cohort.

This figure shows the association between the different possible combinations of ATN plasma biomarkers (left to right: a, plasma P-tau181 vs plasma Aβ42/Aβ40; b, plasma NfL vs plasma P-tau181; and c, plasma Aβ42/Aβ40 vs plasma NfL). Each filled circle represents a single MCI patient from the ADNI selection cohort (n=86); the histograms above the scatterplots show the distribution of concentration values for the plasma biomarker on the x-axis (that is panel A, Aβ42/Aβ40; panel B, P-tau181; panel C, NfL). The blue lines (solid and dashed) indicate the cutoff values for each biomarker.

Extended Data Fig. 4 ROC curves for all models in the model selection analysis with four-year conversion to AD dementia as outcome.

This figure shows ROC curves derived from the model selection analysis with four-year conversion to AD dementia as outcome. AUC values were the following in the BioFINDER cohort: Basic Model = 0.70, ATN=0.88, AT=0.86, AN=0.80, TN=0.88, A=0.73, T=0.86, N=0.79. AUC values were the following in the ADNI cohort: Basic Model = 0.74, ATN=0.88, AT=0.86, AN=0.79, TN=0.89, A=0.75, T=0.85, N=0.73. Full data is available in Supplementary Table 2.

Extended Data Fig. 5 Association between plasma biomarkers in MCI patients in the ADNI prognostic validation cohort.

Scatterplot showing plasma NfL versus plasma P-tau181 for the ADNI prognostic validation cohort (n=483). Each filled circle represents a single MCI patient; the histograms above the scatterplots show the distribution of concentration values for plasma P-tau181. The blue lines (solid and dashed) indicate the cutoff values for each biomarker.

Extended Data Fig. 6 Online individualized risk prediction tool.

Screenshots from our webtool predictprogression.com with predictions for a 70-year female with mild cognitive impairment and baseline MMSE 27. The upper panels (A) show predicted MMSE (left) and probability of conversion to AD dementia (right) at 2 and 4 years in a scenario where plasma Aβ42/Aβ40, P-tau181 and NfL are negative. In the lower panels (B), predicted MMSE (left) and probability of conversion to AD dementia (right) are shown at 2 and 4 years in a scenario where plasma Aβ42/Aβ40, P-tau181 and NfL are positive. In both A and B, the dark blue ribbons are 95% confidence interval of the estimates and the light blue ribbon are 90% prediction intervals. The predictions are derived from linear and logistic regression models that were established in the BioFINDER study on non-demented individuals with mild cognitive impairment.

Supplementary information

Supplementary Information

Supplementary notes, figure captions (for Extended Data Figs. 1–6) and Tables 1–16.

Reporting Summary

Source data

Source Data Fig. 2

Statistics for different ATN models.

Source Data Fig. 3

Statistics for different ATN models.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Cullen, N.C., Leuzy, A., Palmqvist, S. et al. Individualized prognosis of cognitive decline and dementia in mild cognitive impairment based on plasma biomarker combinations. Nat Aging 1, 114–123 (2021). https://doi.org/10.1038/s43587-020-00003-5

Download citation

Further reading

Search

Quick links