Condensed tannins

Proanthocyanidins (PACs), also called condensed tannins, are chains of flavan-3-ol units such as catechin and epicatechin, the same building blocks found in tea and cocoa. They are widespread in plants: grape seeds, apples, cocoa beans, tea, many berries. They give the dry, puckering feeling of strong tea or young red wine, because they bind to proteins in saliva. When heated in acid they release coloured anthocyanidins, which is where the name comes from.

The units of a PAC chain are usually joined by a single carbon–carbon bond between carbon 4 of one unit and carbon 8 (or 6) of the next. This is called a B-type link. In an A-type link there is a second bond as well: an ether bond, through an oxygen, between carbon 2 of the upper unit and the oxygen on carbon 7 of the lower one. The double link makes the joint between the two units more rigid and changes the shape of the molecule. Procyanidin A2, a dimer of two epicatechin units joined by an A-type link, is the simplest example.

Many plants contain some A-type PACs; what is unusual about cranberry is that they make up a large share of its PACs.

How bacteria hold on

Bacteria that cause bladder infections have to resist being washed out by urine. Uropathogenic E. coli do this with fimbriae (also called pili), hair-like protein fibres tipped with adhesins that bind specific sugar structures on the cells lining the urinary tract. P-fimbriae bind to galactose-containing glycolipids. The proposed mechanism for cranberry PACs is physical-chemical rather than antibiotic: the PACs, or compounds derived from them, interfere with the binding of fimbriae to those receptors, so that fewer bacteria attach, without killing the bacteria. The A-type link is thought to help, perhaps by changing how the molecule interacts with the fimbrial proteins, but the exact interaction is still under study.

Measuring PACs: BL-DMAC

PACs are difficult to measure because they are mixtures of chains of different lengths, and different methods give very different numbers for the same sample. A common method for cranberry is based on 4-dimethylaminocinnamaldehyde (DMAC), which reacts with flavan-3-ols in acid to give a green-blue colour measured near 640 nm.

The reason the standard matters can be illustrated with a simple model. If each chain gives roughly the same amount of colour regardless of its length, as would happen if the reaction mainly involves one end of each chain, then a gram of long chains gives less colour than a gram of dimers. A product whose PACs are longer than the dimer standard will then read low.

# A simple model of why the choice of standard matters in the DMAC assay.
# Assumption (a simplification): each chain gives about the same colour per
# molecule, so colour per milligram falls as the chain gets longer.
# Procyanidin A2, the BL-DMAC standard, is a dimer (DP 2).

def apparent_share(mix):
    """mix: {degree of polymerisation: share of PAC mass}. Returns the
    measured PAC mass, as a share of the true mass, read against A2."""
    return sum(share * 2 / dp for dp, share in mix.items())

mixes = {
    "mostly dimers":              {2: 0.6, 3: 0.2, 4: 0.2},
    "tetramers and pentamers":    {2: 0.1, 3: 0.1, 4: 0.4, 5: 0.4},
    "longer chains (DP 4 to 10)": {4: 0.2, 5: 0.2, 6: 0.2, 8: 0.2, 10: 0.2},
}
for name, mix in mixes.items():
    a = apparent_share(mix)
    print(f"{name:27s} reads {100 * a:3.0f} % of true mass "
          f"(true/measured = {1 / a:.1f})")
print("Feliciano et al. (2012): cranberry-PAC standard gave values 2.2 × those with A2")
# mostly dimers               reads  83 % of true mass (true/measured = 1.2)
# tetramers and pentamers     reads  53 % of true mass (true/measured = 1.9)
# longer chains (DP 4 to 10)  reads  34 % of true mass (true/measured = 3.0)
# Feliciano et al. (2012): cranberry-PAC standard gave values 2.2 × those with A2

The middle case, mostly tetramers and pentamers as Foo and colleagues described, gives a factor close to the 2.2 reported for cranberry products. The lesson for anyone reading a specification is that “PAC content 36 mg/g” means nothing without the method and the standard, and that two suppliers’ numbers can be compared only when both are stated.

From the test tube to the clinic

A finding in a test tube does not show that a food prevents disease. Anti-adhesion in vitro shows a mechanism that could work; whether it does work in people is a question for randomised trials, and their results are mixed.

A risk ratio does not say how many people benefit. That depends on how common infections are without cranberry.

# Relative risks from the Cochrane review (Williams et al. 2023), cranberry
# products against placebo or no treatment, symptomatic culture-verified UTI.
reviews = {
    "women with recurrent UTI": (0.74, 0.55, 0.99),
    "children":                 (0.46, 0.32, 0.68),
    "pregnant women":           (1.06, 0.75, 1.50),
}

def per_100(rr, baseline):
    """Infections per 100 people over the study period, and the number who
    would need to take cranberry for one fewer infection (if any)."""
    treated = 100 * baseline * rr
    fewer = 100 * baseline - treated
    nnt = 100 / fewer if fewer > 0 else float("inf")
    return treated, fewer, nnt

baseline = 0.30   # an assumed risk without cranberry, for illustration only
print(f"assumed risk without cranberry: {100 * baseline:.0f} in 100")
print("group                     RR (95% CI)        change per 100     NNT")
for group, (rr, lo, hi) in reviews.items():
    t, f, n = per_100(rr, baseline)
    _, f_lo, _ = per_100(hi, baseline)
    _, f_hi, _ = per_100(lo, baseline)
    nnt = f"{n:.0f}" if n != float("inf") else "none"
    print(f"{group:25s} {rr:.2f} ({lo:.2f}–{hi:.2f})  "
          f"{-f:+5.1f} ({-f_hi:+.1f} to {-f_lo:+.1f})  {nnt:>4s}")
# assumed risk without cranberry: 30 in 100
# group                     RR (95% CI)        change per 100     NNT
# women with recurrent UTI  0.74 (0.55–0.99)   -7.8 (-13.5 to -0.3)    13
# children                  0.46 (0.32–0.68)  -16.2 (-20.4 to -9.6)     6
# pregnant women            1.06 (0.75–1.50)   +1.8 (-7.5 to +15.0)  none

With these assumptions, about 13 women with recurrent infections would need to take cranberry for one fewer infection, but the confidence interval runs from a large benefit to almost none. For pregnant women the interval includes harm as well as benefit and is centred on no effect.

Processing and the berry

Cranberry PACs and the anthocyanins that colour the fruit are both damaged by heat, oxygen and light, and PACs bind to proteins and cell-wall material during processing, which makes them harder to extract and measure. How a berry is dried affects what is left and what a test can find. Chapter 5 turns to the engineering of drying.