In phase-1 22 , we recruited consecutive patients with 'clinically-definite' IP 32 , criteria as in Table 1. There was a contemporaneous call for healthy controls without IP: those found to have 'clinically-possible or -probable' parkinsonism 32 (Table 1, footnote^a) were not excluded. Other criteria were as shown for probands in the Table. Serum immunoglobulin concentrations were contrasted between 120 probands (12 men, 12 women per decade, 40-89 years) and 205 controls (≥12 each gender per decade, 30-89 years). The relationship of presence/absence of features of clinically-possible or -probable IP to concentration was explored in controls, as was that of a continuous measure of hypokinesia, mean-stride-length at free-waking-speed 33 . Any differential effect of Helicobacter anti-urease antibody-status was examined.

In phase-2 23 , we obtained first-visit full blood counts (FBC) from 118 other clinically-definite IP-probands (Table 1). These were contrasted with 381 routine FBCs from consecutive adults, requested by general practices from the same laboratory over the same period. Exclusions in these controls were limited to specified parkinsonism; haematological, immunological and malignant disorders; haemorrhage; and race other than Caucasian (see footnote^b). FBCs from 87 accompanying spouses/partners of probands were also studied, inclusive of clinically-possible/probable/definite IP (other criteria as for probands). Proband and 'spouse' lymphocyte subset data were contrasted with summary statistics (kit manufacturer's and published 34 ) for haematologically-normal adult donors. The following were explored as determinants of FBC and subset profiles: serum haematinic and homocysteine concentrations (probands); serum autoantibody screen (probands); Helicobacter indices and results of hydrogen-breath-testing for SIBO (probands/spouses). In a subgroup, within-proband time-trends in cell counts (yearly sampling) were contrasted between the urea-breath-test (UBT) positive for Helicobacter infection and the UBT-negative, any impact of H. pylori eradication being noted. Any modifying effect of prevalent autoantibodies on the outcome of H. pylori eradication therapy was explored. This involved duplicate gait assessments on two baseline occasions and at ≥ 6-weekly intervals post-therapy, any background anti-parkinsonian medication being constant, and usage of levodopa being an exclusion.

In both phases, any anti-parkinsonian medication and laxative use were recorded. Protocols had been approved by local ethics committees, written consent obtained from participants.

Whole blood T-cell (CD3+), T-helper (CD4+), cytotoxic T-cell (CD8+), B-cell (CD19+), and natural-killer (NK) cell (CD16+CD56+) counts were determined. The sum of the lymphocyte subsets was validated by reference to FBC lymphocyte count 35 (Figure 1).

Sera were stored at -20°C, and randomly allocated within-assay, with (except for Helicobacter serology, see below) between-assay stratification for subject-group, age and gender.

Serum ferritin was measured by two-site sandwich immunoassay with two anti-ferritin antibodies; folate, vitamin B12, and homocysteine by competitive immunoassays, using direct chemiluminescence (Bayer Diagnostics Europe Ltd, Newbery, UK.) Inter-assay coefficients of variation (CV) were 2.7, 4.2 and 3.7% at 10.8, 142.4 and 929.6 μg/l ferritin, respectively; 6.11, 7.19 and 6.36% at 1.70, 10.01 and 14.95 μg/l folate; 9.2, 2.7 and 3.0% at 178.7, 608.83 and 1343.87 ng/l B12; and 5.2, 1.5 and 3.1% at 4.9, 22.2 and 61.6 μmol/l homocysteine.

Immunoglobulin classes were measured by immunoturbidity (Cobas Mira Plus, Roche Diagnostics Corp., Indianapolis, USA). Inter-assay CV was 6.5 and 4.9% at 7.5 and 11.6 g/l IgG, respectively; 5.9 and 4.4% at 1.3 and 2.5 g/l IgA; and 4.5 and 3.4% at 0.7 and 1.1 g/l IgM. IgA subclasses were measured by radial immunodiffusion (Bind A Rid, Birmingham, UK): CV was 11% at 3360 mg/l IgA1; and 7.5% at 360 mg/l IgA2. Serum was screened for anti-nuclear, anti-smooth muscle, anti-mitochondrial, anti-gastric parietal cell and anti-liver/kidney microsomal antibodies. The additional autoantibodies listed, as well as the above, were sought before H. pylori eradication therapy and on average 1 1/2 years after, in a subgroup: anti-neutrophil cytoplasmic (ANCA), anti-thyroid peroxidase, anti-adrenal, anti-tissue transglutaminase, anti-intrinsic factor, anti-acetylcholine receptor, anti-neuronal nuclear and anti-purkinje cell. For each assay, samples were processed in a single batch, blind to sequence.

Helicobacter-status was defined by:- (i) [¹³C]urea-breath-test (INFAI Ltd., York, UK). (ii) Enzyme-linked immunosorbent assay (ELISA) of IgG-antibody directed against cell-bound H. pylori urease (SIA417A, Delta Biologicals, Rome, Italy: seropositive 'ELISA-value' >2.2). Inter-assay CV was 13.0, 8.0 and 6.0% at ELISA-values 0.8, 2.4 and 5.9. (iii) Western Blotting for serum antibodies against H. pylori antigens (RIDA Helicobacter Blot IgG, Quadratech Diagnostics Ltd., Epsom, UK: score >12 positive for current infection, 11-12 equivocal, <11 negative, irrespective of band-density). Test and quality-control immunoblots were read using a smooth curve, fitted through distance/molecular weight co-ordinates derived from the kit's developed-control strip. This identifies 11 bands: cagA product; vacuolating-toxin(Vac)A; urease-B; outer-membrane-protein; flagellin; those at 47, 33, 29, 28, 25 & 19 kDa. Magnifying strips optimized precision. Helicobacter serology was standardised between study phases by protocol, equipment and quality-control, using same investigator throughout. In phase 2, H. pylori infection was confirmed in probands by endoscopic-biopsy for histopathology, culture and molecular-microbiology 23 , eradication judged by UBT and, wherever possible, repeat biopsies.

Hydrogen-breath-tests were carried out in UBT-negative probands and spouses, using a 25 g lactulose test-dose, following 24-hour deprivation of dairy products (and medicinal lactulose) and a breakfast of 250 ml black tea/coffee or water. Hydrogen concentrations were measured (Micro Medical Ltd., Rochester, UK) pre-dose and at 15 minute-intervals for 4 hours after. A prolonged test, with non-absorbable substrate, was used because of potentially impaired gastric-empting and intestinal-transit 9 24 . Breath-hydrogen/time curves 36 37 were not bimodally distributed between-subject. Within-subject, two distinct peaks, early and 'colonic', were not present. Summary outcomes were examined: 2 hour value; maximum-value between 15 and 120 minutes; linear increase over 2 hours, when trend was to peak. Correlation coefficients were high for all contrasts, the maximum-value selected for analysis.

'Contemporaneous' reference range limits, constructed from a measurement made in 400 controls, would have a small standard deviation (approximately 0.09s, where s is standard deviation in the measurement), and be only marginally larger (0.1s) using 200 measurements 38 .

In phase-2, controls below age 46 years had anomalous age-trends in FBC indices: these (19, presumably relatively-ill, individuals) were excluded. Since phase-2 controls were not subject to the rigorous inclusion/exclusion criteria of phase-1, their white cell counts were examined for evidence of distinct Gaussian distributions (e.g. a subgroup with a higher mean count, compatible with acute intercurrent infection). There was a tendency for two distributions to describe total white cell count better (Figure 2), but no evidence that a mixture of distributions provided a better fit for lymphocytes or neutrophils. Comparisons with a 'conventional' reference range allowed robustness of any shift in distribution between subject-groups to be tested. Wilcoxon signed-rank test was used to assess a shift from the reference median, chi-square goodness-of-fit to test whether the expected 5% of measurements fell outside the range.

Many chronic diseases are multi-step and multi-factorial: explanatory models were used to examine between-group and within-subject differences in laboratory measurements. Potential confounding or effect-modifying variables were assessed, examining for interactions prior to direct effects.

Where indicated, logarithmic transformations were employed, to ensure validity of assumptions of normality of distribution and equality between groups of residual variance. The regression parameter for a log transformed outcome-variable was exponentiated, giving a relative (percentage) rather than absolute difference. That for a log transformed predictor was scaled by 1/log_e2, so that a unit change reflected a doubling.

Table 2 shows that the major differences in FBC of IP-probands (P) and their spouses (Ps) from controls (C) were in total lymphocyte count. Mean count was lower by 23.8 (95% CI: 18.7, 28.7)% in P and 17.3 (10.0, 24.0)% in Ps (p < 0.001 in each case, age & gender adjusted). Figure 3 shows the estimated proportion falling below the age-specific contemporaneous reference range: this was 12.5 (binomial exact 95% CI: 7.2, 19.8)% of P and 7.1 (2.4, 15.9)% of Ps, compared with the 2.5% expected. No proband or spouse exceeded the range. The proportion falling below the lower limit of the conventional reference range (1.3 × 10⁹/l) was 30.8 (22.7, 39.9)% of P, 10.0 (4.1, 19.5)% of Ps and 7.3 (4.9, 10.4)% of C, the frequency of 'lymphopenia' varying between groups (Pearson χ², p < 0.001). In the 37.0% of probands never exposed to anti-parkinsonian medication, mean lymphocyte count was 101.0 (95% CI: 89.5, 113.9)% of that in Ps, still significantly lower than in C. Exposure was associated with a 11.1 (0.1, 20.8)% lower count (p = 0.048), but inclusion of time-from-diagnosis (not important independently) reduced its significance to 0.1. Thus, medication status and time-from-diagnosis were intrinsically linked, possibly surrogate for a more pertinent measure of disease.

Neutrophil polymorphonuclear count tended to be higher in P (p = 0.1, gender adjusted) than in C. Haemoglobin concentration, red blood corpuscle count (RBC), packed cell volume (PCV), mean corpuscular volume (MCV), platelet count and mean platelet volume were not different in P or Ps from C. Mean corpuscular haemoglobin concentration (MCHC) was less in both than in C, by small but highly significant amounts (0.46 (0.27, 0.65); 0.33 (0.09, 0.58) g/dl, respectively, p < 0.001 & = 0.008, age and gender adjusted). Mean corpuscular haemoglobin (MCH) tended to be less in P (p = 0.06).

Subset counts in 106 of P and 70 Ps, expressed as a percentage of subset sum, are compared with the kit manufacturer's reference mean and 95% confidence limits in Table 3 and Table 4. Figure 4 summarises direction of effects. In P, CD4+ and CD8+ distributions were platykurtic, greater proportions than expected falling both above and below reference ranges. In Ps, there was a significant rightward shift in mean CD4+, and a significant leftward shift in CD8+ with respect to lower reference limit. A significant leftward shift in mean CD19+ was seen in both P and Ps. That to-the-right in mean CD16+56+ characterised P alone. Subset shifts were not explained by anti-parkinsonian medication.

No difference in CD4+/CD8+ ratio was found between P and Ps, after adjustment for higher values with age (p = 0.001) and lower values in males (p = 0.004). Overall, ratios complied with the inter-quartile reference range for gender by decade 34 . However, in younger males (41-50 years; 14P, 2Ps), the mean ratio (1.26 (0.92, 1.61)) was markedly below the expected 2.29 (p < 0.0001). In the 14 younger male probands, the lower the ratio, the higher the CD8+ count (adjusted r ² = 40%, p = 0.01), but not the lower the CD4+.

Regarding within-couple concordance (70 pairs), CD3+, CD4+, CD8+ and CD19+ counts were lower in probands, by 14.2 (2.0, 24.9), 18.2 (5.1, 29.5), 17.1 (2.1, 29.8) and 31.9 (14.4, 45.8)% respectively (paired t-tests: p = 0.03, 0.01, 0.03 & 0.002). Probands' CD16+56+ count was higher by 23.6 (7.7, 42.0)% (p = 0.004).

Figure 5 illustrates that the long thin duodenal enterocyte mitochondria found in IP are not necessarily associated with abnormal lymphocyte mitochondrial morphology.

Mean nutrient intakes were estimated from 3-day food diaries in 23 of the phase-2 probands: they complied with UK National Reference Values.

In all 126 probands (Table 2), serum homocysteine was above the desired-maximum of <16 μmol/l in 43.2 (binomial exact 95% CI: 34.1, 52.7)%. Figure 6 shows that the expected proportion had a serum B12 below reference range (<180 ng/l), but 16.1 (10.0, 24.0)% had values within the 'equivocal range' (180-250 ng/l). The variance in homocysteine explained by B12 was only 12.1% (adjusted r ², p < 0.001). Including relevant covariates, time-from-diagnosis (p = 0.001) and gender (p = 0.01) in the model increased variance explained to 27.0%, B12 maintaining its significance. Time-from-diagnosis was independent of age, which itself did not contribute to variance explained. Homocysteine was 1.8 (95% CI: 0.8, 2.8)% higher per year post-diagnosis, 12.5 (2.5, 22.5)% higher in females.

Six percent of probands were both newly diagnosed and untreated: their median homocysteine concentration, 13.4 μmol/l, was below the desired maximum. Receiving anti-parkinsonian medication other than levodopa (40.3%) was not associated with increased homocysteine, compared with being untreated (37.0%). Probands receiving levodopa (22.7%) had a median total daily dose of 300 (interquartile range 300, 450) mg), a mean time-from-diagnosis of 7 years. Their homocysteine was higher than in the remainder by 24.2 (95% CI: 9.6, 40.8)% (p = 0.001, adjusted for B12 & gender). However, exposure to levodopa was intrinsically linked to time-from-diagnosis, and the estimated 7-year increase in homocysteine for probands never exposed to levodopa (19.6 (7.7, 31.6)%) was comparable.

Probands' serum folate distribution was platykurtic, 7.6 (binomial exact 95% CI: 3.5, 14.0)% having concentrations above the reference range (3-13 μg/l), 13.6 (8.0, 21.1)% below. Variance in homocysteine explained by folate (p = 0.007) increased from 6.1 to 21.0% by including time-from-diagnosis (p < 0.001) and gender (p = 0.005) in the model, but the significance of folate decreased to 0.1. Folate did not contribute to variance explained by B12, gender and time-from-diagnosis.

Serum ferritin was below the reference range (20-300 μg/l males; 20-200 μg/l post-menopausal females) in 6.8 (3.0, 12.9)%. It contributed to explaining probands' MCHC and MCH: estimated increase of 0.17 (95% CI: 0.01, 0.33) g/dl in MCHC and 0.30 (0.06, 0.54) pg in MCH with a doubling of ferritin (p = 0.04 & 0.02, respectively).

Neither homocysteine nor these haematinics contributed to explaining total lymphocyte count in IP.

Anti-nuclear antibody (ANA) was present in 25.3% of the 126 phase-2 probands, anti-intrinsic factor in 15.1%, and anti-gastric parietal cell in 7.6%. Gastric autoantibodies were not associated with a lower serum B12. Anti-smooth muscle antibody was present in 2.6%, anti-mitochondrial and anti-liver/kidney microsomal negative throughout. In the more comprehensive screen in 21 probands, ANCA was present in 3, but none had antibodies against myeloperoxidase or proteinase-3 targets. Total lymphocyte count was not associated with autoantibody-status.

Table 5 shows that, in phase-1, serum immunoglobulin concentrations in the 120 IP-probands and 205 contemporaneous controls were similar.

Figure 7 shows that presence, in controls, of hand-bradykinesia (16%), or postural abnormality (9%), was very strongly associated with a higher IgG and IgA. Stride-length was shorter by 59.3 (95% CI: 1.9, 116.8) mm if IgG concentration doubled (p = 0.04), but gait was not associated with IgA. Resting rigidity in arms was rare (1%), but rigidity evoked by movement of the contralateral limb, the activation phenomenon, common (70%) and associated with IgA (Figure 7 footnote). Resting tremor was also rare (1.5%), postural tremor common (48%) but not associated with immunoglobulin classes. Of these features, only hand-bradykinesia was associated with IgM, by contrast a strong negative relationship. All associations with IgA were attributable to IgA1.

Immunoglobulin concentrations, in probands, were unrelated to the presence/absence of a cardinal feature or stride-length. Concentrations were not associated with any class of dopaminergic medication. However, constipation, as denoted by regular laxative use (22 of the 100 probands so questioned), was associated with increased IgA (odds ratio 2.1 (1.03, 4.4) for concentration being double if taking medication, p = 0.04). Antimuscarinics can, of course, constipate: the odds ratio remained similar (2.30 (1.23, 4.33), p = 0.01) when antimuscarinic use (14 more probands) was also taken to denote constipation. Again, relationships depended on IgA1.

Crude prevalence rate estimates for 'Parkinson's disease' in 34 European studies range widely, from 65.6 to 12,500 per 100,000 39 . The estimate lay within the binomial exact 95% CI for the spouse cohort (9,195 (4,045, 17,317) per 100,000) in only one study. All 8 spouses with clinically-definite IP had 'frank' UBT-positive Helicobacter infection, only 8% of the other spouses did (P < 0.0001). Table 6 shows that frequency of positive Helicobacter-status, by breath-test or serology, tended to be less in the spouse cohort than in presenting cases.

In the earlier IP-cohort, prevalence of seropositivity was not significantly different from that in the socially-similar controls (Table 6). The odds of a proband's sample being immunoblot-positive and ELISA-negative, compared with vice versa, became 69.1(95% CI: 4.9, 978.3) times greater in phase-2 than, 10 year earlier, in phase-1 (p = 0.02). In the later IP-cohort, ELISA-positivity was shown to represented frank infection, in that its prevalence concurred with that of UBT-positivity.

Maximum breath-hydrogen between 15 and 120 minutes (median 41 (interquartile range 17-73) ppm) was not influenced by age, gender or subject group (40 probands, 15 spouses without clinically-definite IP, in phase-2). The meter manufacturer's diagnostic cut-point (20 ppm increment) was exceeded at least once within the 2 hours in 65% (irrespective of subject-group), by two consecutive readings 37 in 60%.

Although the maximum was not associated with total lymphocyte count, it was with the CD4+ subset count (estimated increase 9.2 (95% CI: 1.2, 17.8)% with a doubling of breath-hydrogen, p = 0.03, gender-adjusted, irrespective of subject group). It contributed to explaining probands' serum ferritin, and MCHC in probands and spouses (estimated decreases of 20.2 (8.5, 30.4)% in ferritin and 0.20 (0.02, 0.39) g/dl in MCHC with a doubling of breath-hydrogen, p = 0.03 & 0.02, respectively). Probands' folate, B12, homocysteine and autoantibody-status were unrelated to maximum breath-hydrogen. Probands' and spouses' CD4+/CD8+ ratios were not explained by breath-hydrogen (nor UBT-status or immunoblot-status).

Helicobacter infection appeared to keep SIBO at bay. Breath-hydrogen was higher (Spearman rank correlation, p = 0.003) in those of the currently UBT-negative probands and spouses who had been UBT-positive on presentation. It tended to be higher (p = 0.06) where the immunoblot score had been positive/equivocal. Hydrogen-breath-tests were performed, in two subsequent probands, before and serially after biopsy-proven eradication of Helicobacter (initially detected by molecular-microbiology on two culture-negative biopsies). One had become positive (two consecutive values criterion) at 8 months, the other converted between 12 and 18 months: even low-density infection appeared protective.