Review Papers Claim by Claim

Visual summary — key quantitative points first

Figure A — Fraction of proteins preserved / significantly decreased

Data source: per-paper counts: DBS +4°C: 34% of 92 proteins not significantly decreased; DBS −24°C: 76% preserved; Liquid −70°C: ~95% preserved (5/92 decreased) — see paper for statistical approach (Bonferroni-corrected linear models)Counts of significant decreases (Björkesten et al. 2017)

Figure B — Median retained abundance after 10 years

Interpretation: modest median loss at +4°C vs small loss at −24°C — emphasizes that lower storage temperature reduces cumulative decay but does not eliminate protein-specific declinesMedian after 10 years (Björkesten et al. 2017)

Figure C — Drying effect vs technical replicates (summary)

Takeaway: drying introduces a small systematic assay bias (reproducible) — compare dried samples to dried standards or correct using dried controls to avoid misinterpretationDrying reproducibility (Björkesten et al. 2017)

Critical appraisal — strengths

• Well-focused experimental question with direct, practical implications for biobanking and population studies (large longitudinal sample needs).
• Use of a high-sensitivity, multiplexed assay (Olink PEA 92-plex) validated for small-volume samples; standardized NPX outputs facilitate per-protein time-series modelingPEA assay & normalization (Björkesten et al. 2017)
• Long storage durations (up to 30 years for DBS; up to 45 years for frozen plasma) provide rare real-world longevity data that are directly actionable for biobanks.
• Appropriate statistics: per-protein linear models, ANOVA, Bonferroni correction — conservative control for false positives in multiple testing.

Critical appraisal — limitations, blind spots, and biases

1. Cohort and sample-type confounding: DBS were neonatal whole blood (cells+plasma) from newborn screening biobanks; frozen comparators were adult plasma/serum (ULSAM). Age, matrix (whole blood vs plasma), and cohort differences confound direct comparison between dry and wet storage outcomes. Authors explicitly state this limitation and adjust plasma data for age/sample type where possible, but residual confounding remainsCohort/matrix confounding (Björkesten et al. 2017)
2. Assay-dependence: results reflect PEA antibody-pair detection of (mostly intact) protein epitopes; mass-spectrometry peptide-based measurements could behave differently (e.g., proteolytic cleavage may not affect some peptide signals the same way as antibody epitopes). The authors acknowledge assay dependence; generalization beyond PEA requires empirical validationAssay dependence (Björkesten et al. 2017)
3. Paper-card and storage heterogeneity: the type of filter paper and exact storage humidity/conditions changed over the decades; authors note different card types used historically and humidity control changes, which can add uncontrolled variability.
4. Small per-timepoint N: five individuals per timepoint for DBS series (n=5 per age/time), producing 460 data points per group (92 proteins × 5 samples) — rich for per-protein time trends but limited for estimating inter-individual biological variance and for detecting low-effect-size changes in proteins with high person-to-person variability.
5. Conflict of interest: senior author has ownership in Olink Proteomics (maker of PEA panels) — disclosed by authors; increases need for independent replication using different platforms or blinded external labsCOI disclosure (Björkesten et al. 2017)
6. Lack of raw-data sharing statement: the provided text does not include a link to raw NPX matrices or sample metadata; reproducibility and reuse would benefit from deposition of per-sample NPX and metadata (paper notes anonymization constraints). The methods state anonymization but don't provide public data access details.

Where the evidence is strongest and weakest

Strong: within-sample comparisons (wet vs dried) and per-protein linear modeling across storage times — these use internal controls and conservative multiple-testing corrections, and the sample preparation and PEA protocol are described in sufficient detail to reproduce the lab pipeline. Weak: cross-cohort comparisons (DBS newborns vs adult plasma), and platform-generalization beyond PEA. Independent confirmation with alternative methods (MRM/PRM or broad-scale DIA-MS) and larger, demographically balanced cohorts would greatly strengthen claims about absolute long-term preservation.

What would falsify the paper's main conclusions?

• Independent large-cohort assays (same DBS cards) showing rapid, non-reproducible loss of detectability across most proteins within a few years — i.e., median losses far greater than reported here.
• Cross-platform replication using MS-based targeted workflows finding discordant stability patterns (e.g., proteins reported stable by PEA that show rapid peptide loss or vice versa) in the same DBS specimens.
• Demonstration that drying-induced shifts are inconsistent between batches or dependent on unreported storage variables (paper lot, humidity) making dried controls insufficient to normalize measurement biases.

Practical recommendations for biobanks and researchers

1. Store DBS at as low a temperature as feasible. The data show a marked advantage at −24°C vs +4°C (median retention and fewer significant declines) — if possible use ≤ −20°C cold chain for long-term storageTemperature effect (Björkesten et al. 2017)
2. Include dried control standards (paper controls spiked with known protein concentrations) on each card/plate to track and correct the reproducible drying shift between wet and dried matrices.
3. When feasible, collect paired samples (DBS + plasma aliquot) at the same time from a subset of participants to calibrate and translate measurements across matrices in longitudinal cohorts.
4. For biomarker discovery prioritize proteins shown to be stable across storage (per-paper lists) and validate candidate biomarkers by re-measurement in fresh samples and in an orthogonal assay (MS or independent immunoassay).
5. Deposit per-sample NPX matrices and metadata (paper type, storage humidity, temperature history) in a controlled-access repository when allowed by ethics to improve reproducibility and cross-study meta-analysis.

Suggested follow-up experiments (concise, testable)

1. Parallel-platform stability test: take archival DBS punches (same card lot) and re-analyze identical punches with (a) PEA, (b) targeted MRM/PRM MS for the same proteins, and (c) discovery DIA-MS peptides; compare protein-specific decay patterns — this will test assay-dependence.
2. Controlled storage-factor study: prepare fresh DBS from a single donor, aliquot onto identical cards, and store at multiple temperatures (RT, +4°C, −20°C, −80°C) and controlled humidities for defined intervals (1, 3, 6, 12, 24 months), then measure PEA + MS to model non-linear kinetics and humidity effects.

Confidence, novelty, and reproducibility scores (paper-focused)

Why those scores (short)

Novelty (7): long-term DBS proteome stability data are uncommon; combining decades-long archival DBS with multiplex PEA is an important contribution. Quality (8): careful methods, conservative statistics, and transparent limitations. Generality (6): results are directly relevant to PEA workflows and DBS biobanks but require cross-platform validation to generalize fully. Usefulness (8): actionable guidance for biobanks on temperature, controls, and assay planning. Reproducibility (7): methods are described but raw matrices are not publicly available in the text — replication by independent labs and cross-platform datasets would raise this further. Explanatory depth (6): provides empirical decay estimates and hypothesizes chemical/aggregation mechanisms but lacks molecular mechanistic follow-up (e.g., specific PTM mapping or peptide-level MS kinetics).

All numerical claims above derive directly from the paper (Björkesten et al., MCP 2017) — consult the paper for per-protein statistics, half-life estimates, and supplemental tables listing the individual proteins and linear-model parametersPrimary study (Björkesten et al. 2017)